Why Cloud Computing is Requiring us to Rethink Resiliency at the Edge

Overview

The paper argues that as enterprise applications move to the cloud, the few on-premise racks left at branch sites become more critical per rack, because they host proprietary applications and the network path to the cloud. It contrasts the Tier III/IV practices typical of centralised and regional facilities with the Tier I reality of many local edge rooms: no security, cable clutter, 1N power, reliance on comfort cooling and no monitoring. Localized or micro data centers range from 1-2 MW down to 10-20 kW, trending toward a couple of racks in a room or closet.

From the user's point of view, series availability is dominated by the weakest edge site. In the example, a Tier III central site at 99.98% (1.6 h downtime) times a Tier I edge at 99.67% (28.8 h) gives 99.65%, or 30.7 hours of downtime. With ten Tier I edge sites of 100 people each plus one Tier III central site, 28,820 of the 30,400 people-hours of annual downtime come from the edge sites. The proposed scorecard multiplies annual downtime hours by a 1-5 severity rating based on people impacted and function, an FMECA-style method used to rank sites; a site at 98.60% availability (122.64 h/yr) with severity 5 scores 613.2, which is 65.5% of the overall 935.6.

Edge best practices are a locked room or enclosure with access control, 24x7 environmental and video monitoring, centralized or remote monitoring, redundant power paths, critical circuits on generator and a second network provider. The example of a Tier III micro data center is a prefabricated single 42U enclosure with redundant UPSs and power distribution. The paper offers a reliability method, but the scorecard rests on assigned severity ratings, and the record gives no more detail on how those ratings are set.

Key points

1. Localized/micro data centers range from 1-2 MW down to 10-20 kW, trending to a couple of racks.
2. Series availability: Tier III central 99.98% (1.6 h) x Tier I edge 99.67% (28.8 h) = 99.65%, or 30.7 hours of downtime.
3. Ten Tier I edge sites of 100 people plus one Tier III central: 28,820 of 30,400 people-hours of downtime come from the edge.
4. Scorecard: annual downtime hours x severity (1-5), an FMECA-style ranking.
5. Example: 98.60% availability (122.64 h/yr) with severity 5 scores 613.2, 65.5% of 935.6.
6. Typical edge room: no security, cable clutter, 1N power, comfort cooling, no monitoring.
7. Example Tier III micro data center: prefabricated single 42U enclosure with redundant UPSs and power distribution.
8. Best practices: locked access-controlled room, 24x7 monitoring, redundant power, generator-backed critical circuits, second network provider.

Using it in design

• In a portfolio/edge module let users model multiple sites with availability (or tier), headcount served and severity (1-5), compute series availability for dependent sites and rank sites by downtime x severity.
• Offer micro data center presets (single 42U enclosure, 10-20 kW) with options for redundant UPS, dedicated cooling, locked enclosure and environmental sensors.
• Warn when an edge room relies only on building comfort cooling, has 1N UPS at a high-severity site, lacks generator-backed circuits or has a single ISP.
• Flag open racks without access control.

Who should read it

IT managers and designers prioritising edge-room investments across many branch sites.

Related standards

• Uptime Institute Tier Standard: Topology
• ANSI/TIA-942
• EN 50600-1
• IEC 60812

Publication details

Authors

Kevin Brown; Wendy Torell

Publisher

Schneider Electric

Year

2021

Document number

WP 256

Document language

English

Source type

Manufacturer

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation