Key Criteria for Evaluating Edge Colocation v

KEY CRITERIA/MARKET LANDSCAPE

Key Criteria for Evaluating Edge Colocation v

NED BELLAVANCE

TOPIC: EDGE INFRASTRUCTURE

CREDIT: PHOTOCREO MICHAL BEDNAREK

https://gigaom.com/analyst/bellavance-ned/

Key Criteria for Evaluating Edge ColocationTABLE OF CONTENTS

Summary1

Edge Colocation Primer2

About the Key Criteria Report3

Report Methodology4

Decision Criteria5

Evaluation Metrics6

Impact of Key Criteria on Evaluation Metrics7

Analyst’s Take8

About Ned Bellavance9

About GigaOm10

Copyright11

Key Criteria for Evaluating Edge Colocation v 2

1. SummaryThe edge market has been growing by leaps and bounds, despite the fact that many people remainuncertain about what the edge is, and which technology components enable edge applications. Withthe rollout of 5G wireless technology and the explosion of Internet of Things (IoT) devices, the marketfor edge enablement technology is projected to expand exponentially. Industries and municipalitieshave taken notice of the potential use cases of edge applications, and are starting to explore possibledeployment models. Edge applications need a place to run, and that place will be some iteration of theinfrastructure edge.

Understanding the opportunity and building an edge computing strategy today will help technologyleaders to improve processes and increase competitiveness. Edge computing is a fundamentalcomponent for modern infrastructures because it allows data to be directly processed and consumedwhere it is created, while sending only the necessary information to the network core and cloud. Edgeinfrastructure has specific characteristics that enable it to respond to unique requirements whilecreating an improved experience for customers. Power consumption, space utilization, ease of use,reliability, and more are all crucial aspects that determine the success or failure of the overall strategy.

Thinking about edge infrastructure as an extension of the existing infrastructure, rather than as aseparate entity, will increase the speed of new application development while improving day-to-dayoperation. Total cost of ownership (TCO) and overall simplicity should always remain among the toppriorities for edge computing deployments.

In this report we will focus on one aspect of an edge deployment: the physical infrastructure that hoststhe hardware and software running an edge application. We will consider the table stakes, key criteria,and evaluation metrics for judging whether a physical infrastructure solution is a suitable match foryour use case. Not all edge applications will have the same requirements when it comes tocommunication, scalability, security, and more. By judging edge colocation solutions against your edgeapplication requirements, you can be confident that the solution meets your needs now and in thefuture.


2. Edge Colocation PrimerThis report looks at Edge Colocation solutions that exist in the market, through the lens of key criteriaand metrics. While our focus is on Edge Colocation, the broader market of Edge-focused solutionsremains immature and defined in di!erent and often conflicting ways. So, first we look at what wemean by Edge Infrastructure.

Defining Edge InfrastructureWe define edge as a location and not a specific technology. The edge is any computing environmentthat is outside of a typical data center or cloud provider. Edge locations tend to be resourceconstrained in terms of power, cooling, and network connectivity when compared to a traditional datacenter environment.

The Linux Foundation Edge (LF Edge) alliance has worked to establish common terminology with itsOpen Glossary of Edge Computing. It also has produced a landscape for edge, available on GitHub,which defines common categories for edge solutions. We will be using many of the definitions laid outby the LF Edge alliance. Notably, the most relevant solutions for this report are those in theinfrastructure, platform, and real estate categories.

Edge encompasses the technology stack that sits outside of cloud providers, on-premises datacenters, and enterprise o"ces. In this report, we are creating a framework for reviewing solutions thatreside in the LF Edge category of real estate, and not infrastructure or platform (for more informationon those two LF Edge categories and edge infrastructure as a whole, please review the Key Criteria forEvaluating Edge Infrastructure report). While these labels are useful in the context of the LF Edgelandscape, for this report, we are instead going to use the following categories (see Figure 1):

• Physical Infrastructure: Solutions that cover the physical infrastructure space will include thephysical resources for an edge solution. This includes physical space, power, cooling, andconnectivity. These are resources that a typical colocation facility might provide but are found in anedge context.

• Core Systems: Solutions in this category provide the base layer for an edge application and includethings like physical hardware, operating systems, and hypervisors. A core system solution sits in thespace between the basic physical infrastructure and the application deployment and managementlayer. A cluster of physical servers running virtual machines could be considered a core system.

• Platform Systems: Solutions in this category provide a well-defined platform for the deployment ofedge applications. They support the management of the core systems, orchestration of applicationdeployment, and monitoring and metrics of application performance. A customized version ofKubernetes with a built-in service catalog and dashboard would be considered a platform system.


https://github.com/lf-edge/glossary

https://landscape.lfedge.org/

https://gigaom.com/report/key-criteria-for-evaluating-edge-infrastructure/

https://gigaom.com/report/key-criteria-for-evaluating-edge-infrastructure/

Figure 1: Edge Applications Depend On Physical, Core and Platform Systems

This report will focus on the physical infrastructure component, also referred to as edge colocation.

Relating Edge Infrastructure to Edge ColocationAs with all emerging markets, vendors are attempting to di!erentiate by o!ering specialized services;in this case, on top of traditional colocation. By establishing a concrete set of categories and astandard set of criteria, readers can better determine whether those solutions meet the requirementsof their use case.

The edge can be thought of as layers, with each layer defined by its compute capacity, powerrestrictions, and network connectivity. The device edge is the outermost component, and it includesend-user devices and sensors located in the ‘last mile’ of the network. The next layer is the


infrastructure edge, connected to the device edge by an access network. The infrastructure edgetypically sits in close proximity to the device edge but has characteristics of a traditionally hosted datacenter. The infrastructure edge may, in turn, be interconnected across many sites using an aggregationedge layer of networking, or through a core network to a public cloud provider or regional datacenter. The available access network, aggregation layer, and core network are all considerations whenselecting an edge colocation provider.

Infrastructure edge represents a sweet spot for many edge applications. While it would be nice to runapplications on the device edge, that approach has limitations in many cases. It can lack su"cientcomputational power, data storage, or electrical power. Device edge devices may have upgrade cyclesof 10-15 years in the industrial space, making them too static a platform for application development. Itis also possible to run edge applications on a public cloud provider or traditional on-premises datacenter. In that case, there may be too much network latency or insu"cient bandwidth to the deviceedge for an application to be e!ective.

The infrastructure edge sits in the middle, with su"cient network bandwidth and lower latency than anupstream solution. It will also have greater computation power, more data storage, and dependableelectrical power. While infrastructure edge equipment will not be as easy to upgrade as a traditionaldata center, it is still simpler to upgrade than thousands or millions of device edge endpoints.

Edge Colocation vendors are currently looking to deliver across these layers. While the market ishighly dynamic right now, it is becoming clear that di!erentiation will come down as much to serviceso!ered, as to infrastructure covered. We use this factor to guide what we see as key di!erentiatingcriteria in the sections below.

Opportunities at the EdgeThere are hundreds, if not thousands, of edge use cases that provide an opportunity to improvee"ciency, enhance customer experiences, reduce waste, and curtail shrinkage. The following are afew sample use cases in di!erent industries.

Manufacturing

A well-known manufacturing company uses IoT sensors in its assembly line to augment the existingsensors built into the original equipment. The new IoT sensors measure vibrations of the equipment athundreds of di!erent points. Data from those sensors is merged with information from the originalequipment sensors to build a machine health model. Anomalies can be detected with greater accuracy,which leads to less downtime and higher production line e"ciency. Sensor data is sent to aninfrastructure edge device for analysis and model building and then to the cloud for long-term storage,aggregate analysis, and modeling across multiple locations.


Shipping and Logistics

Shipping companies can encounter challenges with the proper arrangement, loading, and contents ofpallets for shipping goods. One solution put into practice merges machine vision technology fromsmart cameras with data from the enterprise resource planning (ERP) system to track the packing ofpallets in real time. Workers receive immediate feedback on their packing process, and the companysees reduced shrinkage, improved packing e"ciency, and lowered shipping anomalies. Aninfrastructure edge server performs the integration of machine vision with ERP data and theaggregation of data for analysis.

Gaming

Hatch, a subsidiary of gaming company Rovio, has been leveraging edge computing provided in micro-data centers by Packet. Hatch uses specialized hardware in Packet’s edge locations to run local,multiplayer games in real time, providing an enhanced experience for consumers on lower-endAndroid devices that may not have the necessary memory or CPU to run higher-end games. Bystreaming the game from a local edge point in sub-millisecond proximity to their device, gamers enjoya rich gaming experience on an inexpensive device. Making an edge deployment model available tothem has given Hatch a competitive advantage in emerging markets across the globe.

Due to the immature and evolving nature of edge solutions, it is not straightforward for the end user tograsp where the lines of demarcation are, and how the di!erent solutions compete with orcomplement each other. End users should be able to take the key criteria listed below and apply themto existing solutions and new vendors to construct a viable solution based on their unique needs.


3. About the Key Criteria Report


4. Report MethodologyA GigaOm Key Criteria report analyzes the most important features of a technology category to help ITprofessionals understand how the features impact an enterprise and its IT organization. These featuresare grouped into three categories for purposes of our examination:

• Table Stakes

• Key Criteria

• Emerging Technologies

The goal of the Key Criteria report is to help organizations assess capabilities and build a mid-to-long-term infrastructure strategy. With mature technologies, solutions in the Key Criteria Report are often(but not always) graded on two things: Their applicability across target market categories (such asenterprise or small-medium business) and deployment models (such as cloud or on-premises). Ofcourse, it is up to individual organizations to assess options based on their specific users’ needs, andnot solely on the organization’s vertical or market sector.

Table Stakes: Assumed ValueTable stakes are system characteristics and features that are important to consider when choosing theright solution. They include architectural choices that depend on the organization’s size, requirements,expected growth over time, and types of workloads planned. Table stakes are mature, and theimplementation of these features will not add any business advantage nor significantly change the totalcost of ownership (TCO) return on investment (ROI) of the infrastructure.

Key Criteria: Di!erentiating ValueKey Criteria features serve to di!erentiate one solution from another. Depending on specific userneeds, they can have a significant impact on one or more of the metrics mentioned. Therefore,implementation details are essential to understanding the benefits relative to the infrastructure,processes, or business. After considering table stakes and key criteria, aspects like architectural designand implementation should be analyzed in great detail.

Over time, the di!erentiation introduced by a feature becomes less relevant. It falls into the ‘tablestakes’ group, while new system capabilities introduce new benefits or address new needs, with apositive impact on metrics like e"ciency, manageability, flexibility, and so on.

Key Criteria bring several benefits to organizations of all sizes with di!erent business needs. Thiscategory is organized to give the reader a brief description of the specific functionality or technology,its benefits in general terms, and what to expect from a successful implementation.


Emerging Technologies: Future ValueIn this section, we analyze the most compelling and intriguing technologies on the horizon, trackingthose expected to materialize over the next 12 to 18 months. Some emerging technology features mayalready be present in the market in some form, but usually as part of niche products or designed toaddress very specific use cases. At this stage, implementations are not mature enough to be regardedas key criteria and should be considered mostly for their potential downfield value, where they canmake a di!erence in the long term.

Over time, emerging technologies become key criteria and key criteria become table stakes, as shownin Figure 2. The Key Criteria report reflects the dynamic embedded in this evolution, helping ITdecision makers track and assess emerging technologies that may significantly impact their value to anorganization.

Figure 2: Evolution of Features Over Time

Critical Impact of Features on Evaluation MetricsAfter enumerating the features and capabilities o!ered by a candidate solution, IT decision makersmust consider two types of evaluation metrics:

• Technical parameters such as scalability, flexibility, and reliability

• Business-oriented parameters such as return on investment and total cost of ownership


Evaluation Metrics in this report are broad, top-line characteristics of a product or service -- forexample, characteristics like scalability or interoperability or cost e!ectiveness. These issues can beeasily understood across the business organization. By evaluating how key criteria and other featuresimpact these metrics, it’s possible to better understand the value a solution can have to theorganization. For example, robust integration and extensibility features can directly impact a technicalparameter like flexibility, while also impacting a business parameter like total cost of ownership.

The goal of the GigaOm Key Criteria report is to structure and simplify the decision-making processaround key criteria and evaluation metrics, allowing the first to inform the second, and enabling ITprofessionals to make better decisions.


5. Decision Criteria

Table StakesThe features in table stakes are those things that buyers will take for granted. Much like a combustionengine in a car, the feature is presumed present and functional. Di!erentiation comes from a disruptionof that assumption; for instance, the use of an electric motor to provide propulsion.

Connectivity

There are few things more important to an edge solution than connectivity. Edge applications tend toneed low latency connectivity to an aggregation point, and each aggregation point needs connectivityof some type to the next level in the stack, whether that is a corporate-owned data center or a publiccloud provider. There is also the consideration of interconnection between the edge aggregationpoints and central o"ces, sometimes referred to as the aggregation network.

Figure 3: Edge Devices Require Aggregated Connectivity to the Cloud

As depicted in Figure 3, Physical infrastructure solutions must provide basic connectivity acrossmultiple media, including fiber optic and radio-based technologies. While this does not necessarilymean 5G (see the Emerging Technologies section), it does mean providing connectivity to thedownstream components of an edge application, such as mobile phones, IoT gateways, or smartcars.Many of the downstream components will use cellular connectivity or WiFi, and the physicalinfrastructure solution must support that type of connectivity.

Likewise, the solution must support connection to upstream components, like a private data center orpublic cloud provider. That will mean 10Gb fiber, at the very least, with 100Gb fiber on the horizon.


Usually, there will be multiple edge colocation facilities within a geographic region. Theinterconnection of these sites provides multiple benefits to the user, especially if they adopt adistributed application model that can be housed in each site. For example, consider a gamingapplication consumed on mobile devices. As the end users travel around the region, the connectioncan be seamlessly handed o! between edge collocation points to minimize latency. A robustaggregation network also can provide failover and high-availability to applications that are designedwith multiple points of presence in mind. The aggregation network should form a resilient andperformant mesh across the region that it supports.

Security in Depth

Physical edge infrastructure is deployed often in non-traditional and di"cult-to- secure locations.Unlike a traditional data center or cloud-based deployment where physical access is tightly restricted,recorded, and regulated by on-site personnel, edge deployments may include remote locations withno human presence. Any solution stored in such a location must be able to withstand both a physicalattack as well as cyber intrusion.

At the physical infrastructure layer, an edge solution should provide physical security: video cameras,motion sensors, personnel check-ins, and physical isolation of di!erent tenants, if required. In manyways, the physical presence for an edge location is no di!erent than a traditional colocation datacenter, with the exception that often the location will be unsta!ed and possibly in a remote area.

Integrated Monitoring

Physical infrastructure at the edge and in colocation facilities will be largely unsta!ed, meaning thatremote electronic monitoring will be especially critical. Each vendor must have a robust monitoringsystem that tracks the physical status of the location, including power lines, temperature, humidity, andair quality, as well as the status of the physical hardware installed at each location. Proper monitoringgoes hand-in-hand with a robust standard of security; visibility being a key component of any securityprotocol.

In addition to monitoring the facility for internal processes, the vendor should also make theirmonitoring information easily accessible to end users. Ideally, the monitoring stack will have an APIfront-end that can be integrated easily with the customer’s existing monitoring platform.

Remote Support

As mentioned previously, edge solutions may be running in remote, unsta!ed locations that are di"cultto access. A physical infrastructure solution needs to be able to provide remote operational supportand hands-on assistance when required. The service-level agreement (SLA) for hands-on support mayvary depending on location, but some level of hands-on support is a must-have for any physicalinfrastructure solution. Ideally, the vendor will have the ability to remediate most issues without rolling atruck to the location, thus reducing downtime and cost for the end user.


For those cases where the customer prefers to manage their equipment directly, the vendor shouldo!er remote support expertise to help the customer access and navigate the facilities. The spaceo!ered by the vendor probably will not resemble a traditional datacenter, and the customer will needto be guided on the proper operational procedures and standards.

Regional Redundancy

In a traditional datacenter, there is an assumption of redundancy for all components. Depending on therating of the datacenter, it may follow an N+1, N+2, or even N+N formula for redundancy ofcomponents. Due to the distributed nature of edge colocation facilities, individual sites may be thoughtof as isolated failure domains. Some vendors may consider this a reason to have lower redundancy ofcomponents at the site-level, as long as complete redundancy is available at the regional level.

Customers should understand the approach to redundancy for each vendor in the space, and makesure that the application and deployment model aligns with the level of redundancy available from thevendors being considered. For instance, a highly distributed application running across multiple sitesmight be resilient enough to tolerate a single site failure, while another application with a moremonolithic model may not be able to tolerate a site failure gracefully.

Industry Certifications

Some edge applications will be industry specific or deal with sensitive data that is regulated. Physicalinfrastructure vendors will need to have the appropriate level of certification to host these applications.The required certificate will be heavily dependent on the customer application and industry, but thereis an expectation that the vendor will have achieved common data center certifications like SSAE 16 orPCI-DSS.

Key CriteriaThe table stake items listed in the previous section represent features that all solutions should includeat a minimum to be considered fit for use. Key criteria, on the other hand, are those items thatrepresent points of potential di!erentiation between the solutions from di!erent vendors.

Dense Regional Coverage

Coverage can be considered through the lens of sites in a region, regions in a geography, and globalpresence. Depending on the use case of the customer, the necessary level of coverage will varygreatly. Dense regional coverage refers to the number of sites in a region and the average latency to asite from any point in the region. Edge applications tend to need low latency connections back to anedge colocation site. More sites in a region reduces latency, improves resiliency, and increases thepotential bandwidth between sites.

As potential customers evaluate physical infrastructure solutions for their target markets, the density of


regional coverage could a!ect their purchasing decisions greatly. Although one provider may havethree larger locations in a region, another provider may have ten or more smaller locations in the sameregion providing lower latency to the device edge.

Globally Available Footprint

Organizations have grown accustomed to the public-cloud consumption model, in which adding a newlocation does not require working with a di!erent public cloud. The end user simply selects a region inthe target geography and deploys their application. As edge applications grow in popularity,organizations will expect a similar level of service from the physical infrastructure vendor. Solutions candi!erentiate themselves by being globally available, or by forming strategic partnerships with vendorsin separate geographies to provide global coverage.

Having multiple regions in a geography or a global presence reduces the administrative overhead forthe customer. If the same provider can be used across an entire country or continent, then thecustomer doesn’t need to negotiate as many agreements or learn the processes of multiple providers.For customers that require a global presence, it may make sense to work through a provider withglobal reach for the same reasons.

Consumption-Based Pricing

Renting space in a typical colocation facility usually involves a multiyear commitment for a minimumamount of space, for example, two cabinets for three years. In addition to those long-termcommitments, there is usually a cost associated with Internet lines, tra"c volume, hands-on assistance,additional monitoring, and any extra services. In stark contrast is the cloud operational model that hasminimal upfront costs, no annual commitment, and metered billing for everything.

Some consumers will be better suited to cloud-like, consumption based pricing, and others may preferthe colocation economic model. It is likely that most edge colocation providers will follow a typicalcolocation cost model. Therefore a vendor may distinguish themselves by o!ering capacity on aconsumption-based model. Special consideration should be given to the impact of burst scaling andrapid, permanent growth on any existing cost models, since they can skew the preferred economicmodel towards operational expenditure or capital expenditure (OpEx vs CapEx).

Rapid Rollout Capability

Edge colocation facilities need to keep pace with the rapid growth of edge applications. Whenevaluating a vendor, customers should keep in mind both current capacity and future growth. Becausephysical infrastructure vendors are dealing with real estate and physical constraints, it is important tounderstand how quickly they can roll out a new site once the property is secured.

Vendors can distinguish themselves by showing a proven track record of turning up new sites in ashort period of time and rapidly adding capacity to existing sites. This capability is a function of bothphysical site design and well-defined process.Key Criteria for Evaluating Edge Colocation v 15

Flexible Consumption Model

Due to the unknown nature of edge application design, architecture, and growth, it is di"cult to predictexactly what physical resources are required for an application in the long-term. Especially for new usecases, it is critical that the edge colocation provider have su"cient flexibility in their o!ering toaccommodate shifts in how the application is deployed, managed, and monitored. This flexibilityincludes aspects such as network connectivity and capacity, physical space consumption, and powerdensity.

A customer may start out wanting a mesh network among all sites that is managed by the edgecolocation provider. As time passes, the customer may want to take over the logic of the mesh networkto provide additional redundancy or streamline the performance of their application based onobserved behavior. The provider will need to have the flexibility to adapt to the new networkrequirements of the customer. This could be true also for the addition of denser server hardware withnew power requirements, or the introduction of GPUs that increase the power consumption of a singleserver.

Automated Issue Resolution

There can be no doubt that issues will arise in remote edge locations that require assistance from thevendor. Some issues can be solved remotely, while others will need sta! to be physically present at thesite or sites. Since many customers will not have local sta! at the edge sites, they will rely on thevendor to provide remote assistance in a timely manner to resolve issues. Mean time to resolution(MTTR) is a metric measuring the average time it takes to resolve an open issue with the vendor.

Vendors can distinguish themselves by implementing proactive and automated issue resolution, inwhich problems are detected before they impact service and are remediated through automatedprocesses.

Emerging TechnologiesThe physical infrastructure being deployed to support edge applications is evolving at a rapid pace,with new entrants and ideas emerging every few months. There are several up and comingtechnologies that should be considered when selecting an edge colocation solution to use for edgeapplication. These near-term, game-changing technologies are not yet ready for production. Perhapsthey are already deployed in a limited fashion or available as an alpha feature. As they mature, theymay significantly impact the metrics in the previous section.

Several emerging technologies have the potential to disrupt or alter the current state of physicalinfrastructure edge solutions. Some of these technologies have already been deployed in a limitedcapacity, but are not yet widely available.


5G Wireless

When 4G/LTE finally arrived on the scene, it revolutionized the experience many users had on theirmobile devices. It enabled scenarios and applications that were previously unimaginable on 3Gspeeds. The advent of 5G will bring a similar revolution to the IoT and edge devices. It will drasticallyalter the amount of data processed by an edge location as well as the required bandwidth capacityfrom the edge location to the internet at large.

One of the main features of 5G is the concept of the local breakout, a standard that prevents thetromboning of data across the local network to a remote data center and back to the local networkagain. Imagine two IoT devices in the same region that need to communicate using 4G. Unless specialbreakout technology was put in place, the communication from one device would travel through theaccess layer to an aggregation layer that determines where the other device is, and back downthrough the access layer to the second device — the total round-trip-time (RTT) measured in 10s or100s of milliseconds. With 5G, the communication would not leave the local edge location, and thus thetotal RTT could be reduced to 100s of microseconds.

Local breakout can be facilitated greatly by the services available at edge locations.Telecommunications providers can leverage the enhanced connectivity available at edge locations tofurther reduce the distance of travel and the number of jumps needed to move between source anddestination. Some physical infrastructure providers have already planned out their mesh network andsolution to include providing these enhancements, and they have partnerships with telecomcompanies to assist with improved local breakout.

Public-Cloud Edge

Two of the public-cloud leaders have already started the process of o!ering services at the Edge.Microsoft has its Azure Stack suite of products, and AWS recently announced Local Zones based ontheir AWS Outposts product. CloudFlare already has a robust edge presence across the globe ando!ers Workers to run serverless applications at the edge. In the short-term, edge applications are likelyto be driven by specific use cases, rather than being available as general-purpose platforms to buildnew solutions on. The public-cloud providers are aggressively attempting to expand their services intothe edge to provide a consistent platform across the spectrum of environments. In the long term, thiscould disrupt the ecosystem of edge solution providers that exist today as they may be supplanted oracquired by the public-cloud giants.

In most cases, the public cloud providers will look to work with existing physical infrastructureproviders, rather than building their own dedicated facilities. While it might make sense for the bigpublic cloud providers to build their own datacenters, it does not make economic sense to build outthousands of edge locations and deal with the issues of real estate management and construction.Physical infrastructure providers may see increased consumption from public cloud providers, whichwill limit their capacity for other customers in the short term. In the long term, significant investmentfrom public cloud providers may accelerate the construction of edge colocation facilities, and provide asimple on-ramp for small, and medium businesses that do not wish to manage their own hardware.


Remote Work

The recent COVID-19 pandemic has changed network usage patterns drastically and increased theprevalence of remote work for information workers. These changes may become permanent in thewake of the pandemic, as employers and employees alike discover that their presence at the o"ce isnot strictly necessary. The shift in tra"c patterns and need for distributed VPNs, remote desktops, andcontent caching may increase the need for edge colocation facilities that are closer to thoseperforming the remote work. Such an increase may spur the construction of new edge sites orexpansion of existing edge sites to satisfy the additional demand. Edge applications and networksspecialize in distributed and decentralized usage patterns, and remote work is a prime example ofboth.


6. Evaluation MetricsOur assessment of the solution space continues with an exploration of evaluation metrics, which areused here to determine the impact that edge colocation services might have on an organization. Inmany cases, these metrics will be consistent across technology sectors, reflecting fundamental aspectslike flexibility, ease of use, and total cost of ownership. For purposes of this exploration, we considerthe following evaluation metrics:

• Total cost of ownership

• Time to value

• Capacity and scalability

• Partner ecosystem

• Roadmap and growth

Total Cost of Ownership

Any edge solution will require an investment from the consumer to ensure the deployment issuccessful. TCO attempts to evaluate all the explicit and implicit costs that go into making a particularsolution viable. A dedicated colocation space may be more expensive, but also meet regulatory needs.

Time to Value

Once a customer signs an agreement with a vendor, the clock starts ticking. Vendors are evaluated ontheir onboarding process and time until the customer is able to start using the edge location to hosttheir application. This includes the installation of core systems, network configuration, and deploymentof the application in production. While some factors are outside the purview of the physicalinfrastructure vendor, they can assist in streamlining the process of onboarding.

Capacity and Scalability

For the purpose of this report, scalability refers to two distinct measures. One is the ability of thephysical infrastructure solution to scale a single location. As edge applications become more popularand grow in scope and scale, the vendor should be able to scale their existing facilities to meetincreased customer demands. Commonly this would be referred to as vertical scalability, and in somecases may be a literal vertical scaling of the facility to add more capacity.

The second aspect is horizontal scalability, which in this case refers to the ability of a vendor to scaletheir solution in a region by adding sites. In order to scale e!ectively, there is an assumption thatworkloads can be rebalanced, and that the network aggregation layer is capable of scaling to expandthe mesh between sites.Key Criteria for Evaluating Edge Colocation v 19

Similar to the dimensions of scalability, there is a similar consideration of capacity at both the individualsite and the regional level. Each individual site will have a certain amount of power, cooling, networkcapacity, and physical space. While each of these factors can be increased to a certain degree—adding more network circuits or additional cooling units — there is a reasonable upper limit ofcapacity available to an individual customer. For consumers that have an established growth model fortheir application, they must verify that the vendor can provide su"cient capacity in both the short andlong term.

Partner Ecosystem

The provider of edge physical infrastructure will not be an island unto themselves. They need to workwith partners to create a viable solution. There are several di!erent partnerships to consider whenassessing each vendor: upstream, downstream, stack.

Upstream partners are those partners that provide upstream services to the edge colocation provider.These primarily include internet exchange points, ISPs, and public cloud providers. They may alsoinclude partnerships with large colocation providers hosting private clouds as an alternative to publiccloud providers. For a customer already using one of the public cloud providers, it may be critical thattheir edge colocation has a partnership with that provider to enhance the network connectivity orpossibly supply localized versions of the public cloud services at the physical infrastructure location.

Downstream partners represent those vendors that provide connectivity to end user devices. Thiscould be ISPs, radio access network (RAN) providers, or telecommunications companies. In order toprovide the best low latency and comprehensive connectivity to downstream devices, physicalcolocation facilities at the edge will need to integrate with the providers of those networks. Dependingon the use case of the customer, they may require a wired connection to edge devices, a wirelessconnection using a proprietary radio network, or a 5G connection through a telecom.

Stack partners are those partners that augment the existing solution by providing services associatedwith the core systems and platform systems in the edge infrastructure stack. Some examples would becontent distribution networks (CDNs) that place their equipment in the edge locations and work closelywith the edge colocation provider to enhance the functionality of their solution. In some cases thestack partners will be the primary point of contact for consumers, so they do not have to deal with thephysical infrastructure layer. In other cases, the stack partners create an economic incentive to usetheir combination of services with the edge colocation provider, while not obscuring the relationshipbetween the edge colocation provider and customer.

A robust ecosystem of partners can provide an end-to-end solution that is validated and pre-configured. Obviously, this lowers the barriers to entry and lessens the adoption curve for customers,allowing them to focus on the development of their application.


Roadmap and Growth

Just as important as where a vendor is today is where they are planning to go tomorrow. The edgeinfrastructure space is still relatively new, and the key players are still being established. Customersthat are looking for a long-term relationship should be ready to evaluate the financial stability ofpotential vendors, as well as their roadmap for features and growth. For instance, a potential customerthat is doing a pilot deployment in Dallas, Texas today may plan to expand out for a full productiondeployment across the US next year. When evaluating the potential vendors, they will need to find onethat is suitable for their current geographic location, and one that already has locations in other keyareas or has plans to expand in the near future.

Unlike other aspects of technology, acquiring the rights to use real estate and bringing in the relevantresources for networking and power is not an automated, instantaneous process. Leases need to benegotiated, permits must be granted, and construction needs to happen. All of these processes requiremonths or years of planning, and are capital intensive in nature. The viability of a vendor’s plans mustbe seen through the lens of real estate management and construction, rather than a lens of technologywhere things occur rapidly in a software defined paradigm.


7. Impact of Key Criteria on Evaluation MetricsIn this chapter, we provide guidance on how the key criteria features described in the previous chapterimpact the metrics defined in the Evaluation Metrics chapter earlier in this document. Table 1 helps thereader understand:

• The impact that each feature or capability has on each evaluation metric

• The solutions in the market with the best impact on a specific metric at the moment of writing

• The implementation that o!ers the best impact on a specific feature or system capability

Table 1: Key Criteria and Their Impact on Evaluation Metrics


To better understand the impact of Key Criteria on Evaluation Metrics depicted in the graphic above,we have expanded on each metric below.

Total cost of ownership: The total cost of ownership will be most impacted by the cost model of thevendor. Consumption based pricing will keep the initial TCO fairly low as the customer ramps up asolution from development to proof of concept, and finally production. Additionally, the key criteria of aflexible consumption model and automated issue resolution will play a factor. If a customer is able toscale their consumption of space, power, and cooling dynamically, then the TCO model could becomeOpEx friendly. Otherwise, there may be a large CapEx component that will need to be recouped overseveral years. Also playing a factor is the mean time to issue resolution. Downtime means lost revenue.A solution with a strong automated issue resolution component can minimize downtime and improveTCO calculations.

Time to value: Time to value is meant to measure how long it takes for a customer to start seeing areturn on their investment. While much of the time to value relies on the operational e"ciency of thecustomer, the edge colocation facility can expedite the process by providing excellent remote support,leveraging the partner ecosystem, and simplifying the cost model for rapid adoption. Vendors that o!erconsumption based pricing and flexibility in their consumption model allow a customer to rapidlyprototype a solution at a minimal cost, and iterate over multiple designs until they reach one that isprofitable.

Capacity & Scalability: The biggest factors in capacity and scalability are the density of regionalcoverage and global coverage. The more sites there are, the greater the capacity and scalability of agiven region. More regions mean that a highly successful application can scale across the globe withexcess capacity available as needed. Even if the required density is not available today, customersshould assess the capability of a vendor to roll out additional sites or more capacity rapidly at aparticular site.

Partner Ecosystem: Partners looking to work with the vendor might be interested in working with acompany that has flexible consumption models and the ability to roll out additional capacity in serviceof the partner’s solution. Many of the key criteria for a vendor solution may be reliant on partnerintegrations, and thus there is a symbiotic relationship between the partner ecosystem and thestrength of a vendor solution. Some vendors may even partner with each other to increase regional orglobal presence.

Roadmap and Growth: The roadmap for development of new features and the growth of the vendor isgoing to be highly reliant on their ability to roll out additional sites and capacity. Dense regionalcoverage also provides the ability for a physical infrastructure vendor to provide innovative features ontop of their platform.


8. Analyst’s TakePhysical infrastructure edge solutions are still in their adolescence, as vendors determine whatcustomers want in their solutions and how to build out a general-purpose physical infrastructure edgeplatform. The goal of this report is to provide some clarity around the di!erent layers that compose theinfrastructure edge stack—physical infrastructure, core systems, and platform systems—with a focus onedge colocation facilities that provide physical infrastructure.

The Table Stakes section outlines what features should be included in any viable solution, and do notrepresent points of di!erentiation for any specific solution. The Key Criteria section outlines whatfeatures present opportunities for vendors to di!erentiate their solutions from others in themarketplace. Key Criteria have an impact on di!erent metrics that may inform a purchasing decision bya customer. Lastly, there are a host of emerging technologies that are potentially game-changing forhow infrastructure edge will evolve over the next three to five years.

Our short-term recommendation is for CIOs and CTOs to educate themselves on the edge colocationsolutions that exist in the market today, and look for opportunities where edge could assist in achievingcompany objectives or initiatives. Edge technologies are going through a discovery and growth periodat the moment, with few clear winners and an uncertain path. With the full realization of edgetechnologies being five to10 years away, it makes sense to adopt a ‘wait and see’ approach with highlyfocused investment.

For specific use cases, we encourage practitioners to start with a hosted o!ering to determine whethera project is viable, and then bring certain components of the solution in-house where the internal teamcan enable di!erentiated benefits. Most organizations will lack the resources to manage hundreds ofinfrastructure edge locations physically, so it makes sense to find a strong edge colocation partner thatcan meet your current needs, and in the long term meet the growing needs of your edge deployments.

With the establishment of criteria and metrics to evaluate physical infrastructure edge solutions, futureGigaOm reports will focus on applying this methodology to specific vendors.


9. About Ned BellavanceNed Bellavance

Ned is an experienced IT practitioner with experience in thefield. Ned has worked with Fortune 500 companies andSMBs across multiple verticals, developing and deployingboth on-premises and cloud-based architectures. Nedhas authored two books on the Azure Kubernetes Serviceand HashiCorp Terraform and holds several industrycertifications from vendors including but not limited toMicrosoft, VMware, AWS and Citrix.



10. About GigaOmGigaOm provides technical, operational, and business advice for IT’s strategic digital enterprise andbusiness initiatives. Enterprise business leaders, CIOs, and technology organizations partner withGigaOm for practical, actionable, strategic, and visionary advice for modernizing and transforming theirbusiness. GigaOm’s advice empowers enterprises to successfully compete in an increasinglycomplicated business atmosphere that requires a solid understanding of constantly changing customerdemands.

GigaOm works directly with enterprises both inside and outside of the IT organization to apply provenresearch and methodologies designed to avoid pitfalls and roadblocks while balancing risk andinnovation. Research methodologies include but are not limited to adoption and benchmarkingsurveys, use cases, interviews, ROI/TCO, market landscapes, strategic trends, and technicalbenchmarks. Our analysts possess 20+ years of experience advising a spectrum of clients from earlyadopters to mainstream enterprises.

GigaOm’s perspective is that of the unbiased enterprise practitioner. Through this perspective, GigaOmconnects with engaged and loyal subscribers on a deep and meaningful level.


11. Copyright© Knowingly, Inc. 2020 "Key Criteria for Evaluating Edge Colocation" is a trademark of Knowingly, Inc..For permission to reproduce this report, please contact [email protected].


https://www.knowingly.com/

https://www.knowingly.com/

mailto:[email protected]

Gigaom Research

MARKET RADAR

GigaOm Radar for Edge Colocation v1.0

NED BELLAVANCE




GigaOm Radar for Edge ColocationTABLE OF CONTENTS

Summary1

About the GigaOm Radar2

Market Categories and Deployment Types3

Key Criteria Comparison4

GigaOm Radar5

Vendor Insights6

Analyst’s Take7


About GigaOm9

Copyright10

GigaomResearch

GigaOm Radar for Edge Colocation v1.0 2

1. Summary

The edge market has been growing by leaps and bounds, despite the fact that many people remainuncertain about what the edge is, and which technology components enable edge applications. Withthe rollout of 5G wireless technology and the explosion of the Internet of Things (IoT) devices, themarket for edge enablement technology is projected to expand exponentially. Industries andmunicipalities have taken notice of the potential use cases of edge applications, and are starting toexplore possible deployment models. Edge applications need a place to run, and that place will oftenbe some iteration of the infrastructure edge.

Edge infrastructure is an extension of existing infrastructure found in the public cloud or on-premisesdata center, located in close proximity to edge devices and IoT gateways. Applications running onedge infrastructure can perform operations such as:

• Data aggregation and summarization

• Machine learning analysis

• Localized breakout

• Real-time data processing

The infrastructure edge can be broken down into three categories established in our Key Criteria forEdge Infrastructure report: physical infrastructure, core systems, and platform systems. This radarreport focuses on physical infrastructure, also referred to as edge colocation, using the key criteria andevaluation metrics found in the Key Criteria for Edge Colocation report.

Physical infrastructure addresses the physical resources for an edge solution, encompassing resourcesthat include physical space, power, cooling, and connectivity. These are resources that a typicalcolocation facility might provide, but are found in an edge context.

While the use cases for the edge are varied and still unfolding, the need for an edge colocationsolution is clear. The transformation of edge applications relies on suitably provisioned locations thatsupport their compute, network, storage, and security requirements. The radar used in tandem with theKey Criteria report can assist you in evaluating which solutions meet the unique requirements of youredge application.

GigaomResearch


2. About the GigaOm Radar

GigaomResearch


3. Market Categories and Deployment Types

For a better understanding of the market and vendor positioning (Table 1), we categorized solutions foredge colocation by the target market segment (small-medium, large enterprise, and ISP/MSP) and bythe expected deployment model (access edge or regional edge).

Market Segment

• Small-medium enterprise: In this category the best solutions will meet the needs of very smallbusinesses that can grow up to address those of medium-sized infrastructures. They can also besolutions adopted by large enterprises for departmental use cases. Solutions that do well in thiscategory can be adopted by organizations with limited resources and skill sets.

• Large enterprise: Usually adopted for larger and business-critical projects. Solutions that excel inthis category have a strong focus on scale, security, and experience working with larger companies.Globally available footprints and low TCO will be major differentiators.

• ISP/MSP: Edge colocation facilities will be heavily utilized by service providers in thetelecommunications and cloud services industries. The networking, coverage, and flexibility of eachsolution will be critical for meeting the unique needs of these industries. Many of these rollouts willneed to scale beyond the current facilities, and solutions can differentiate on their regional densityand rapid rollout capabilities.

Deployment Model

In addition, we recognize two deployment models for solutions in this report: access edge or regionaledge.

• Access Edge: Vendors in this deployment model are focused on building a rich network of smalledge sites in major metropolitan areas. They tend to be focused on a single geography with highdensity in select metropolitan areas. They may choose to partner with data center providers foradditional capacity and network exchange points to enable public cloud and private data centerconnectivity.

• Regional Edge: Vendors in this deployment model already have a robust number of large datacenters in major metropolitan areas. They are focused on building out the capabilities of theseexisting properties for edge use cases and applications. They tend to have a global presence andreduced density at the local level. They may choose to partner with or acquire access edgevendors.

GigaomResearch


Table 1: Vendor Positioning

GigaomResearch


4. Key Criteria Comparison

Following the general indications introduced with the “Key Criteria for Evaluating Edge Colocation,”Table 2 summarizes the performance of each vendor included in this research, in the areas that weconsider differentiating and critical in this sector. The objective is to give the reader a snapshot of thetechnical capabilities of different solutions and define the perimeter of the market landscape.

Table 2. Key Criteria and Evaluation Metrics Comparison

In some cases, the vendor’s pricing model data was not shared, so the Consumption-Based Pricing rowhas been left blank. Note also that a higher ranking in the Total Cost of Ownership field (TCO) indicatesthat the TCO is higher, not that the solution is a better value.

By combining the information provided in Table 1 and Table 2, the reader should be able to get a clearidea of the market and the available technical solutions.

GigaomResearch


5. GigaOm Radar

This report synthesizes the analysis of key criteria and their impact on evaluation metrics to inform theGigaOm Radar graphic in Figure 1. The resulting chart is a forward-looking perspective on all thevendors in this report, based on their products’ technical capabilities and feature sets.

Figure 1: GigaOm Radar for Edge Colocation

The GigaOm Radar plots vendor solutions across a series of concentric rings, with those set closer tocenter judged to be of higher overall value. The chart characterizes each vendor on two axes--Maturityversus Innovation, and Feature Play versus Platform Play--while providing an arrow that projects eachsolution’s evolution over the coming 12 to 18 months.

GigaomResearch


As you can see in the Radar chart in Figure 1, access edge vendors tend to be more of a feature play,whereas the regional edge vendors are focused on creating a platform. We believe that Vapor IO islooking to expand its capabilities into more of a platform play through internal development andpartnerships.

Innovation is a rejection of sticking with the status quo, so it’s no surprise that the access edge vendorsare strong innovators in the edge colocation sphere. Digital Realty also seems to be trying to reinventitself and its approach as an edge colocation provider.

We believe that the regional edge vendors will continue to mature their offerings and attempt to build abroader platform. We also expect that new challengers will arrive on the scene, especially around theaccess edge. As more edge projects are approved, the demand for access edge will increase both inthe U.S. and abroad.

GigaomResearch


GigaomResearch


6. Vendor Insights

CoreSite

CoreSite is a data center operator in eight U.S. markets with 23 purpose-built colocation facilities. Thecompany focuses on providing colocation, cloud, and interconnection to a community of over 1,350customers. It should be considered a regional edge provider, offering excellent networking capabilitiesand partnerships with the major public cloud providers like Amazon Web Services and Microsoft Azure.

CoreSite is currently shifting its focus to edge use cases, such as autonomous vehicles, genomics R&D,streaming data, and gaming. CoreSite offers colocation services from a single cabinet to a private cageor even a custom suite, giving customers a high degree of flexibility and the ability to grow as theirapplication demands it. The company’s size and scale give it a high score on capacity and scalabilityand make it an attractive option for regional edge services in areas where the company has asignificant presence. It does not yet have edge-specific services or partnerships with access edgeproviders to expand its reach.

CoreSite also specializes in carrier-neutral peering, and Internet exchange sites. Peering, cloudexchanges and inter-site connectivity from CoreSite enable organizations to set up a more efficientnetwork architecture and support multi-cloud connectivity by accessing multiple vendors through asingle connection with the CoreSite Open Cloud Exchange (OCX). Access edge deployments will needa way to connect back to the public cloud or private data center. CoreSite can provide this connectivitywhile serving as a centralized point for additional data processing and routing of applicationcommunications.

Strengths: CoreSite has a strong background in data center operations and managed services, whichfits nicely with the skills gap at some small-to-medium businesses (SMBs). It has multiple data centersin each region, providing a high level of connectivity.

Challenges: CoreSite is growing its edge presence currently, and may not have sufficient coverage foredge applications with low-latency requirements. It also lacks a physical footprint internationally.

GigaomResearch


Digital Realty

Digital Realty is a global company with hundreds of data centers across the world. While it is a majorplayer in the colocation space, it has not focused on building a platform until recently. The companyannounced the PlatformDIGITAL brand in an attempt to move beyond basic colocation services andprovide edge-to-core-to-cloud solutions around networking, operations, storage, and security. Thisnew and directed focus has given Digital Realty an excellent rating for roadmap and growth.

PlatformDIGITAL is a software-defined platform working at a global scale across all of Digital Realty’sdata centers. It is composed of the Network Hub, Control Hub, Data Hub, and SX Fabric. Eachcomponent helps to manage a different aspect of the data center stack across multiple locations. Manyedge applications will make use of hundreds, if not thousands, of sites. Creating a platform capable ofmanaging multiple sites with software-defined networking, storage, and access controls will beespecially useful for the operators of those edge applications.

Digital Realty would be considered a regional edge player, due to the presence of data centers inmultiple major metropolitan areas. It may not have sufficient density in a given market to provideaccess edge services, but it can certainly work with access edge partners to create a comprehensivesolution. To that end, Digital Realty has partnered with Vapor IO to extend PlatformDIGITAL to theaccess edge by integrating Vapor IO’s Kinetic Edge Exchange offering into the Network Hub. DigitalRealty has plans to continue expanding its partnership ecosystem with other providers, mobile networkoperators (MNOs), and tower companies.

Strengths: Digital Realty boasts a global footprint that will appeal to particular companies andworkloads. It’s emerging platforms-centric services promise to enable rapid and flexible deploymentsin its data centers.

Challenges: Digital Realty has a large number of data centers, but lacks a presence at the accessedge. The company hopes its partnership with Vapor IO will fill the gap to support low-latency edgeapplications.

GigaomResearch


EdgeConneX

EdgeConneX offers an interesting mix of both hyperscale and what the company likes to call‘hyperlocal’ operating facilities, ranging from 10 kilowatts (kW) to 100-plus megawatts (MW). With such alarge breadth of offerings, EdgeConneX received high marks for having a flexible model forconsumption. The company currently operates data centers deployed in over 30 markets across threecontinents. The publicly available data centers are characteristic of a regional edge provider, with oneor two facilities in a given market.

The EdgeConneX Far Edge product line offers single-tenant, turn-key solutions that augment theprovider’s existing public data centers. Far Edge can refer to access edge facilities within 10 kilometersof the device edge or a micro-data center at the device edge itself. The EdgeConneX Far Edge productseems to be targeted at the latter, due to its single-tenant nature and proximity to edge devices.

EdgeConneX has a robust monitoring and self-service system, EdgeOS, that enables users to view thestatus of all their sites—from hyperscale to access edge—from a single UI. A unified managementsolution from a single vendor such as this one, for both regional edge and access edge deployments,improves the TCO for running an edge application.

Strengths: The mix of both regional and access edge offerings allows customers to work with a singlevendor for all their edge needs. EdgeConneX has great flexibility in ways to access edge facilitiesdesigned and deployed in the field.

Challenges: Access edge facilities appear to be single-tenant and not a colocation arrangement. Thecompany’s wide focus across hyperscale, regional edge, and access edge may lead to deficiencies inexecution.

GigaomResearch


EdgeMicro

EdgeMicro operates micro-sized data centers providing leasable space, power, and cooling in acompact, EIA-standard, prebuilt form factor. Each unit is pre-fabricated and built offsite, then installed ina prepared location. EdgeMicro focuses on deploying sites across the United States and is increasingdensity based on customer demand. The form factor lends itself to the rapid rollout of new sites toincrease capacity, with a mobile deployment option available to create immediate capacity.

EdgeMicro is laser-focused on building robust and secure physical locations with the proper power,cooling, and network options, without interfering with the customer. The corporate leadership ofEdgeMicro has extensive experience with carriers, ISPs, and content providers, which are the primarytarget for EdgeMicro’s products and services. Due to its high level of focus, the company has forgedmultiple partnerships to create a complete solution for clients, earning EdgeMicro high marks in thepartner ecosystem category.

EdgeMicro solutions are not opinionated in terms of network and application design. This providesflexibility for the primary target customers, but less sophisticated customers may prefer a moredirected and managed approach.

Strengths: EdgeMicro can deploy new capacity rapidly to an existing or new region. Its sites haveespecially robust networking resources, making it ideal for telecom and CDN solutions that requirehigh levels of upstream and downstream bandwidth.

Challenges: The offering from EdgeMicro seems to focus on highly technical customers and nottraditional SMBs. It also provides a low level of site density in most regions where it is deployed.

GigaomResearch


Equinix

Equinix is one of the world’s largest data center providers, with over 210 data centers in 55metropolitan areas across 25 countries. Such an extensive and unmatched global presence hasearned it the highest marks for a globally available footprint. The company has traditionally focused onrobust connectivity options for Internet exchange and network peering points. The company’s largeglobal presence in metropolitan markets enables it to function as a regional edge provider. Unlike anaccess edge provider, the smallest data centers at Equinix are around 5,000 square feet.

Equinix has unparalleled peering connectivity to all the major and minor carriers, as well as the publiccloud. They have developed an offering called the Equinix Cloud Exchange Fabric (ECX Fabric),enabling businesses to build a virtual fabric across all Equinix sites. Theoretically, such a virtual fabriccould be extended to both edge sites and branch offices, providing a uniform control and managementplane for all networking services. A unified network fabric for edge sites would greatly simplify theadministration and deployment of edge applications.

Equinix also has a mature monitoring platform for data center infrastructure management (DCIM) calledIBX SmartView. As workloads approach the edge, locations tend to become more diverse. Equinixplans to offer IBX SmartView as a regular part of operations for colocation services to implementproper monitoring and alerts.

Strengths: The ECX Fabric offering will enable businesses to build a virtual fabric across all Equinixsites in close proximity to edge locations. Equinix also has a robust self-service portal and monitoringsolutions with multiple integrations, including ServiceNow.

Challenges: Equinix lacks a true access edge presence (a problem for the most demanding real-timeenterprise and low-latency applications). At the same time, its large data center footprint can slow theexpansion of sites into new or existing regions. Equinix could augment its offering by partnering withan access edge player.

GigaomResearch


Vapor IO

Vapor IO focuses on deploying micro data centers at the access edge layer, each providing networkconnectivity and up to 150 to 180 kW of power. Vapor IO works closely, though not exclusively, withCrown Castle International—a real estate investment trust—for the sourcing of deployment sites. Thisrelationship enables Vapor IO to quickly procure real estate and networking for new sites across thecountry. Vapor IO locations provide space, power, and cooling in a compact, stackable form factorcalled Vapor Edge Modules (VEM) that are assembled off-site.

Vapor IO combines multiple sites in a region to form what it calls the “Kinetic Edge.” From the KineticEdge, it can offer distributed colocation and networking services, presenting multiple sites as a single,virtual data center. Currently, coverage is limited to regions in the United States, where Vapor IO plansto deploy facilities in 36 cities by the end of 2021, focusing on site density and networking for eachdeployed region to create a virtual data center with extremely low network latency. Its plans forscalability and growth have earned Vapor IO excellent marks in the roadmap and growth category.

Vapor IO also provides a monitoring and management system using the Synse open-source protocol.This integrates well with third-party systems and enables customers to monitor and manage, in waysthat complement and enhance current management toolsets, the resources it has deployed in VaporIO data centers. Vapor IO also offers the Kinetic Edge Portal, a web-based service providing Vapor IOcustomers with capacity planning, site configuration, monitoring, and troubleshooting.

Vapor IO recently announced a partnership with Digital Realty, a major data center operator, to offerthe Vapor IO Kinetic Edge as a network service in Digital Realty’s PlatformDIGITAL solution. This addsfunctionality to the Digital Realty portfolio and introduces Vapor IO to a large swath of existing DigitalRealty customers.

Strengths: Vapor IO’s robust blend of modules and managed networking appeals to service providers,CDNs, and industries deploying edge infrastructure and hardware across a metropolitan area. Vapor IOalso does an excellent job of providing monitoring.

Challenges: Vapor IO is focused strictly on cities in the U.S. for the moment; therefore it lacks a globalpresence.

GigaomResearch


7. Analyst’s Take

As demand for low-latency, high-bandwidth edge applications increases, so does the need for physicalspace to host those applications. The edge colocation marketplace divides into two distinct categoriesof providers: access edge and regional edge.

The access edge providers focus on smaller individual site deployments and building a mesh networkout of those sites for a given region. The strength of this approach is the ability to achieve ultra-lowlatency to the device edge and continue to maintain the same network profile as a device changes itslocation. Not all applications will require this type of disaggregated solution, and those that do need topay a premium for the lower latency and smaller hosting footprint.

Regional edge providers generally are using their existing colocation facilities to offer edge-friendlysolutions. The focus for these providers appears to be creating a platform for managing networkingand security around an edge deployment. Whereas access edge providers are supplying space andconnectivity, the regional edge providers want to build a platform that expands beyond physicalinfrastructure into the realm of core systems—the second of three tiers in our edge infrastructuretaxonomy.

As the edge market grows, we believe that the two categories will continue to differentiate along thesame lines. Access edge providers will look to build a dense network of local sites to provide the bestpossible performance for edge applications. By contrast, regional edge providers will attempt to createa multi-tier platform service that overlays the cloud, data center, and edge. We expect to see additionalpartnerships formed between players in the two categories, and partnerships formed between theaccess edge providers and the major cloud hyperscalers.

We believe that most customers will consume an edge colocation service through a partnership with acore system or platform system provider. Readers who are planning to deploy their own physicalinfrastructure in an edge colocation facility should evaluate the requirements of their edge applicationsacross the dimensions of networking, security, and operations to determine which category of physicalinfrastructure edge provider is best for you.

GigaomResearch


8. About Ned Bellavance

Ned BellavanceNed is an experienced IT practitioner with experience in thefield. Ned has worked with Fortune 500 companies andSMBs across multiple verticals, developing and deployingboth on-premises and cloud-based architectures. Nedhas authored two books on the Azure Kubernetes Serviceand HashiCorp Terraform and holds several industrycertifications from vendors including but not limited toMicrosoft, VMware, AWS and Citrix.

GigaomResearch



9. About GigaOm

GigaOm provides technical, operational, and business advice for IT’s strategic digital enterprise andbusiness initiatives. Enterprise business leaders, CIOs, and technology organizations partner withGigaOm for practical, actionable, strategic, and visionary advice for modernizing and transforming theirbusiness. GigaOm’s advice empowers enterprises to successfully compete in an increasinglycomplicated business atmosphere that requires a solid understanding of constantly changing customerdemands.



GigaomResearch


10. Copyright

© Knowingly, Inc. 2020 "GigaOm Radar for Edge Colocation" is a trademark of Knowingly, Inc.. Forpermission to reproduce this report, please contact [email protected].

GigaomResearch


Gigaom Research

VENDOR PROFILE

Vapor IO v1.0

Key Criteria of Edge ColocationNED BELLAVANCE




Vapor IOKey Criteria of Edge Colocation

TABLE OF CONTENTS

Summary1

About the Vendor Profile2

Key Criteria Analysis3

Evaluation Metrics Analysis4

Bottom Line5


About GigaOm7

Copyright8

GigaomResearch

Vapor IO v1.0 2

1. Summary

Vapor IO offers edge colocation and network solutions for customers running edge applicationsadjacent to the access edge (0-10ms latency). Its Vapor Edge Module (VEM) design is deployed atmultiple locations within a metropolitan area to create what the company calls the Kinetic Edge. TheKinetic Edge enables edge colocation for physical resources, converging wireless and wirelineconnectivity in a Kinetic Edge Fabric, along with software-defined exchange points for peering andInternet through the Kinetic Edge Exchange. Vapor IO’s facilities serve as the connectivity and computehub between the access edge and the nearest regional data center or IX point.

The Vapor IO solution includes four key components: physical location, colocation, interconnection,and transport. Vapor IO operates and maintains the physical sites that host its edge modules, offering amulti-tenant colocation solution in quarter-rack increments with dedicated security, networking, andpower. All sites within a market are interconnected in a mesh topology that forms the Kinetic Edge,allowing customers to deploy a multi-site distributed application architecture without worrying aboutthe underlying networking. Vapor IO also offers low-latency connectivity to local Internet exchanges,simplifying customer networking challenges.

To assist with management and monitoring of the Kinetic Edge solution, Vapor IO offers the KineticEdge portal and the Synse open-source protocol for monitoring operational technology within adatacenter. The portal assists with capacity planning and monitoring of systems through a simple UI,and the Synse protocol provides APIs for integration with other monitoring and ticketing systems aswell as automation and orchestration platforms.

GigaomResearch

Vapor IO v1.0 3

2. About the Vendor Profile

GigaomResearch

Vapor IO v1.0 4

3. Key Criteria Analysis

Dense Regional Coverage

Vapor IO’s concept of the Kinetic Edge is predicated on many interconnected sites working together toprovide compute and networking services. Vapor IO has been pursuing this strategy as it expands tonew metropolitan areas, resulting in dense regional coverage.

Placing a large number of sites in a region has two primary effects: it lowers the average latency ofdevices running edge applications and it improves the overall resiliency of the system. Much like anyother distributed system, a single node or site failure will not result in a catastrophic outage of thewhole network. Vapor IO also uses redundant and diversified fiber routes, further increasing theresiliency of the solution.

Globally Available Footprint

For the time being, Vapor IO is focused on rolling out its solutions within the United States, with a planto have 36 Kinetic Edge cities deployed by the end of 2021. Due to the focus on U.S.-onlydeployments, Vapor IO does not have a globally available footprint.

Customers that require global deployment of an edge solution in the immediate future would need tofind vendors offering similar services in other countries. Vapor IO is participating in the Kinetic EdgeAlliance, which may include members that function outside of the U.S. The company also has plans toexpand to a global presence, and is actively partnering with companies like Digital Realty to increasethe footprint of its services.

Consumption-Based Pricing

The Vapor IO pricing model combines leased space and network access across the Kinetic Edge for afixed monthly cost with a three-year commitment. The fixed monthly fee includes power and cooling,Kinetic Edge networking, Kinetic Edge portal, and telemetry from Synse data feeds.

Power and cooling costs are baked into the pricing, with no upcharges for variation in consumption.Likewise, networking for the Kinetic Edge includes both traffic across the Kinetic Edge in the regionand transport on and off the Kinetic Edge to nearby Internet exchanges and cloud on-ramps.

Network transport costs are a major concern for public cloud deployments, so it is especiallyinteresting that Vapor IO decided to go with a flat fee for network traffic rather than metering cross-zone or egress traffic. This decision makes the cost of the Vapor IO solution much more predictablethan a public cloud deployment.

GigaomResearch

Vapor IO v1.0 5

Rapid Rollout Capability

As demand for edge applications grows, colocation providers need to keep pace with that demand.Vapor IO’s VEM model allows it to add additional capacity to an existing site with relative ease. Sincethe Vapor Edge Modules are built and configured off site, standing up a new site can also happenrapidly.

Vapor IO is planning to expand its national footprint from four markets at the beginning of 2020 to 36markets by the end of 2021. That level of accelerated growth indicates that Vapor IO is able to roll outnew sites rapidly.

Flexible Consumption Model

Many edge application projects will start as pilot programs or proofs of concept and then grow intosomething bigger. Vapor IO offers options for users to start at a single site and expand their footprint,both at the individual site and across multiple sites as their needs grow.

The Kinetic Edge Exchange provides a flexible networking platform that can grow and shift with theuser’s changing needs. Additional capacity can be added on-demand and new peering relationshipscan be deployed through software-defined policies, without the end user needing to visit the Vapor IOsites.

Automated Issue Resolution

With so many unmanned sites under Vapor IO’s management, the company needed a solid monitoringand remote management solution. Vapor IO has adopted Synse, an open-source solution formonitoring all operational technology within a datacenter. The Synse API is openly available tocustomers, so they can monitor the status of any given site and make appropriate decisions for theiredge applications based on that data.

Synse can monitor and perform remote operations on devices in the data center, including makingoperational adjustments or restarting a device. The solution is extensible through a plug-in modelincluding an SDK for internal and third-party development, allowing it to be flexible and future-proof.

GigaomResearch

Vapor IO v1.0 6

4. Evaluation Metrics Analysis

Total Cost of Ownership

Vapor IO provides a low total cost of ownership (TCO) compared to a company deploying its own edgesites to create sufficient coverage. Since Vapor IO has a flexible deployment model, it is possible tostart small and grow the solution over time, keeping the cost of the solution in proportion to itsdeployment footprint. Vapor IO’s fixed cost structure provides a highly predictable pricing model, socustomers need not worry about network egress costs or variable power consumption.

Vapor IO also has a lower TCO versus deploying in a regional edge provider. To achieve the same levelof regional coverage, a customer would need to contract with multiple regional edge providers,thereby increasing complexity and administrative overhead. Even when doing so, it is likely that acustomer would not be able to replicate the features and functionality available from Vapor IO.

Time to Value

Vapor IO can provide rapid time to value given its simple setup and onboarding process. Of particularnote are the Kinetic Edge networking and exchange systems. Rather than build a mesh network andconfigure internet breakouts, customers can leverage the Kinetic Edge to quickly get connected acrossKinetic Edge locations and the broader Internet. The Kinetic Edge also gives the customer a presencein IX facilities, where they can negotiate for additional cross-connects easily.

From an operations perspective, there will be some delay as Synse is adapted. Although it is a flexibleopen platform, it also requires customization if a customer wishes to integrate the OperationalTechnology (OT) information into their existing monitoring platforms. Vapor IO may want to considerdeveloping standard integrations for common monitoring and ticketing systems.

Capacity and Scalability

Scalability can refer to both the individual site (vertical scale) and the regional deployment (horizontalscale). Vapor IO’s VEM model allows it to design a deployment that is optimized for each site, with theoption to scale each site vertically if additional capacity is required. The Kinetic Edge model enablesthe company to scale horizontally by adding more sites to the connected edge.

As with all edge data centers, there is a reasonable upper limit on the capacity of any given site. VaporIO has a high-density design allowing for 150-180 kW of IT equipment, as opposed to more commondesigns of 80 kW from other vendors. Customers should plan for limited capacity on a per-site basis,and design their applications to take advantage of the Kinetic Edge as a whole for both scale andresiliency.

GigaomResearch

Vapor IO v1.0 7

Partner Ecosystem

The partner system can be evaluated by looking at downstream, upstream, and stack partners thathelp create complete solutions for Vapor IO customers. The company has a number of essentialpartnerships along each front.

Vapor IO is working with various telecom partners to build out 5G radio networks for downstreamdevice connectivity. It also has announced a partnership with Cloudflare to serve content todownstream consumers. Vapor IO’s partnership with Digital Realty helps it expand its upstreampresence, enable bursting to the regional edge, and improve connectivity between regions and to thepublic cloud.

The Kinetic Edge Alliance, started by Vapor IO, has over 50 member companies across all aspects ofthe ecosystem, including silicon providers, land owners, and software companies. The Kinetic EdgeAlliance looks to accelerate edge adoption by creating reference architectures, crafting joint solutionsacross vendors, and building a global community to discuss and plan the future of edge infrastructureand computing.

Roadmap and Growth

Competing in the nascent field of edge colocation requires a focus on growth and planning for newfeatures. Vapor IO recently secured $90M in series C funding to fuel its aggressive plans for expansionto 36 markets by the end of 2021.

Vapor IO’s participation and leadership in the Kinetic Edge Alliance brings it to the forefront ofdiscussions around edge architectures, while being active in the LF Edge working group will giveVapor IO insight into and influence over the shape of the edge to come.

GigaomResearch

Vapor IO v1.0 8

5. Bottom Line

Vapor IO has pursued an access edge strategy with a heavy emphasis on the networking aspects ofedge colocation. Its Kinetic Edge strategy provides a robust connectivity and colocation solution forcustomers wanting to avoid the cost and complexity of building something in-house. It is a clear pointof differentiation, and one of the reasons why regional edge providers like Digital Realty are eager topartner with Vapor IO.

Another main point of differentiation is Vapor IO’s form factor and ability to scale rapidly. Vapor IOdesigns its own Vapor Edge Modules, and can adapt the design to the unique requirements of aparticular site, while maintaining high-density capacities of 150-180 kW for IT equipment. The edgemarket is poised to expand dramatically in the next 24 to 36 months, and based on Vapor IO’sroadmap and growth targets, it seems more than capable of meeting future demand. This is especiallygood news for potential buyers who wish to minimize the number of edge colocation vendors theywork with.

A final major point of differentiation is Vapor IO’s monitoring solution Synse. Each VEM is heavilyautomated and sensorized, including in aspects like temperature, air pressure, humidity, and vibration.All of this data is combined with information from OT systems such as HVAC to present a completepicture of the data center through the Synse protocol. The open-source protocol provides an excellentpoint of integration with existing customer solutions, and it can be customized and extended via a plug-in system to meet specialized requirements.

Vapor IO has an impressive access edge solution offering exceptional connectivity through the KineticEdge and robust management capabilities across all sites. The company’s plan for growth andexpansion of the partner ecosystem is an excellent signal that Vapor IO is a leader in the edgecolocation space.

GigaomResearch

Vapor IO v1.0 9

6. About Ned Bellavance

Ned BellavanceNed is an experienced IT practitioner with experience in thefield. Ned has worked with Fortune 500 companies andSMBs across multiple verticals, developing and deployingboth on-premises and cloud-based architectures. Nedhas authored two books on the Azure Kubernetes Serviceand HashiCorp Terraform and holds several industrycertifications from vendors including but not limited toMicrosoft, VMware, AWS and Citrix.

GigaomResearch

Vapor IO v1.0 10


7. About GigaOm

GigaOm provides technical, operational, and business advice for IT’s strategic digital enterprise andbusiness initiatives. Enterprise business leaders, CIOs, and technology organizations partner withGigaOm for practical, actionable, strategic, and visionary advice for modernizing and transforming theirbusiness. GigaOm’s advice empowers enterprises to successfully compete in an increasinglycomplicated business atmosphere that requires a solid understanding of constantly changing customerdemands.



GigaomResearch

Vapor IO v1.0 11

8. Copyright

© Knowingly, Inc. 2020 "Vapor IO" is a trademark of Knowingly, Inc.. For permission to reproduce thisreport, please contact [email protected].

GigaomResearch

Vapor IO v1.0 12

Key Criteria for Evaluating Edge Colocation v

Documents

Transcript of Key Criteria for Evaluating Edge Colocation v