Edge Technology Options & Deployment Models – Light Reading

With hype around all things “edge” on the rise, Heavy Reading launched a research study to gain a realistic understanding of how edge computing will affect the future of network connectivity. Conducted by Heavy Reading and co-sponsored by Arista, CoreSite, Fujitsu and Infinera, the Strategies for Connecting the Edge study is anchored in a global survey of 91 network operators and enterprises.

In this blog — the second of two readouts of the key findings — we focus on technology options and deployment models.

Location, Location, Location
The first blog tackled a basic question: “What is the edge?” This blog focuses on another basic question of intense discussion and debate: “Where is the edge?” To get a better sense of industry views, Heavy Reading asked survey respondents to select the locations for their initial edge computing deployments.

For the group as a whole, there was strong agreement on the primary locations for edge computing, namely, multi-tenant data centers (with dense network connectivity), large regional data centers, and corporate/in-house data centers. Figure 1 separates the results by telecom operator respondents and enterprise respondents.

Figure 1: Locations for Initial Edge Computing Deployments

N=59 telecom, 23 enterprise (Source: Heavy Reading)

N=59 telecom, 23 enterprise
(Source: Heavy Reading)

There are some differences in how telecom operators and enterprises prioritize these three locations. For telecom operators, multi-tenant data centers are expected to be the top location, followed by large regional data centers and then in-house locations. The majority of enterprises, however, selected corporate/in-house locations, followed by multi-tenant data centers and then regional data centers.

Interestingly, telecom operators and enterprises are also largely in agreement on where edge computing will not be housed initially: in street cabinets either at aggregation points or at cell sites. This finding is bad news for those speculating that edge connectivity must be placed at every macro site and small cell in order to deliver on the 5G promise of ultra-reliable low latency communication (URLLC) and massive machine-type communication (mMTC) applications. Service providers and vendors closest to the market have been cautious for some time about how far the edge will spread, and the survey results indicate this cautious viewpoint is becoming more widespread.

Technology options
Just like the centralized cloud, the edge also requires efficient network connectivity to be valuable. Heavy Reading asked respondents to select the most important network equipment features and functions for edge data center interconnection (DCI) over the next three years. From a list of features/ functions provided, Figure 2 shows the results in ranked order from highest to lowest priority.

Figure 2: Most Important Network Equipment Features/Functions for Edge DCI

N=90 Score is a weighted calculation, where items ranked first are given a higher weight (Source: Heavy Reading)

Score is a weighted calculation, where items ranked first are given a higher weight
(Source: Heavy Reading)

Consistent with overall concerns about data security, Layer 1 encryption/FIPS certification topped the list, followed by single multi-rate transponders, integrated optics on routers (or IPoDWDM), and automation features. With the exception of security, the prioritized functions are aimed at saving capex and opex. IPoDWDM, for example, saves capex by eliminating transponder shelves and placing long reach optics directly in the routers themselves. Automation is a major theme across networking overall and saves on operations by reducing/eliminating manual interactions and speeding up processes.

Single multi-rate transponders, meanwhile, hold the potential to save on both opex and capex by reducing card inventory and simplifying ordering. However, Heavy Reading cautions that relative pricing was not included in the survey question and that cost premiums on single-rate transponders could erode some of the potential benefits.

Respondents — particularly the telecom operators — identified integrated optics on routers, or IPoDWDM, as an important enabling technology for edge DCI applications. Digging deeper, we asked respondents to estimate how extensively IPoDWDM will be used in their edge networks. Figure 3 shows results for the 59 telecom operator respondents only.

Figure 3: Expected IPoDWDM Connectivity for Edge Networks (Telecom Only)

N=59 (Source: Heavy Reading)

(Source: Heavy Reading)

IPoDWDM has been a long time coming, but results indicate that technology and market conditions are finally ripe to make a move — and edge connectivity is a key application. At nearly one-third, a plurality of telecom respondents expect that 10% to 29% of their edge connections will be IPoDWDM, while an additional quarter (25%) expect 30% to 50% of edge connections to use IPoDWDM. Thus, over half of telecom operators surveyed expect IPoDWDM to account for 10% to 50% of their edge DCI connections.

Readers interested in in additional data points and analysis from this survey can access the full white paper, Strategies for Connecting the Edge: 2019 Heavy Reading Survey. They can also access the related archived webinar, Connectivity for the Edge Computing Era, in which Heavy Reading and study sponsors expand on these edge topics and more.

This article is sponsored by Arista, CoreSite, Fujitsu and Infinera.

— Sterling Perrin, Principal Analyst, Heavy Reading

(0)  | 

This UrIoTNews article is syndicated fromGoogle News