Data center co-location electrical system design

Overview

This article is an expert piece by Kenneth Kutsmeda published in Consulting-Specifying Engineer in 2018. Its subject is the comparison of the electrical redundancy typical of enterprise, colocation and cloud facilities, and how shared redundancy affects tenants in multi-tenant sites. The piece sets side by side the redundancy expectations that change with the business model.

The author first defines N with a worked example: with a 1,800 kW load and 1,000 kW UPS modules, N is 2. He then contrasts catcher and distributed ('3 to make 2') UPS schemes. These definitions help prevent the same term from being used in different senses in redundancy discussions and give the related design calculations a common starting point.

The typical arrangements by business model are as follows. A typical financial enterprise has 2N utility, 2N UPS, N+1 or N+2 generators and 2N distribution with maintenance bypass. Colocation uses N utility and N+1 or N+2 UPS and generators, with one redundant unit per 4-10 base units. Cloud has redundancy mainly in software, with UPS at N or N+1 (or server-level batteries), one redundant unit per 12-16 base units, and the network kept at 2N.

As one of the lessons learned, the article stresses the consequence of shared redundancy: a unit taken out for one tenant's maintenance may be unavailable during another tenant's failure. It also lists due-diligence questions about maintenance exposure and service level agreement (SLA) terms. As a limit, the article describes typical practice; for a specific project the redundancy decision must be assessed separately against risk tolerance and contract terms.

Key points

1. N example: a 1,800 kW load with 1,000 kW UPS modules gives N = 2.
2. Typical financial enterprise: 2N utility, 2N UPS, N+1 or N+2 generators, 2N distribution with maintenance bypass.
3. Colocation: N utility, N+1/N+2 UPS and generators; one redundant unit per 4-10 base units.
4. Cloud: redundancy mainly in software; UPS N or N+1 (or server-level batteries); one redundant unit per 12-16 base units; network kept 2N.
5. Shared redundancy means a unit out for one tenant's maintenance may be unavailable during another tenant's failure.
6. Catcher and distributed ('3 to make 2') UPS schemes are contrasted.
7. Due-diligence questions about maintenance exposure and SLA terms are listed.

Using it in design

• In business-model templates (enterprise, colo, cloud), set default redundancy and base-to-redundant ratios.
• In the reliability check, compute exposure when one redundant unit is in maintenance.
• Warn if the ratio exceeds about 1:10 for colocation.
• Assess tenant contract maintenance windows and SLA terms together with the redundancy design.

Who should read it

Engineers designing electrical redundancy for colocation and multi-tenant facilities, and technical managers doing due diligence on the tenant side.

Publication details

Authors

Kenneth Kutsmeda

Publisher

Consulting-Specifying Engineer

Year

2018

Document language

English

Source type

Expert / community

Access

Free to read

Licence

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

Citation