|Автор: Edward Henigin|
A data center that draws utility power from two separate substations provides less risk of extended power outages than a data center powered by one substation. How a data center brings those separate substation feeds into a facility directly affects its costs, risk level and control. There are three different ways data centers set up locations of a throw-over switch from two diverse utility feeds into a building: per-site, per-facility and per-UPS. In this article, we examine the pros and cons of each option from a facility owner’s perspective.
Per-site is the utility company’s typical solution for bringing power from two substations into a facility. Both feeds are routed into a single throw-over switch, which has a single feed to one or more utility transformers. Each utility transformer then feeds into a single building. Since the utility company retains control of the throw-over switch, no maintenance is required by the data center operator, making this the least expensive option.
Not having to perform maintenance, however, means the company forfeits control of the throw-over switch to the energy company and relies on them for proper maintenance. Typically, utility companies have a “run-to-failure” maintenance plan, meaning they do not perform preventative maintenance of the switch. Instead, they wait for it to fail – taking all buildings on the site down – and then replace the switch under emergency conditions. This is cheaper for the utility company, but substantially increases the risk if prolonged downtime.
Additionally, with only one power line per facility, if anything happens to that line, the facility will have to rely on generator power indefinitely. If any future construction is planned on or near the site, the impact of accidental line cutting between the transformer and facility is greater than having two lines to each building.
The next level of configuration is to have dual utility feeds to each building and a dedicated throw-over switch per-facility. In this configuration, each facility has the option of being powered by one substation or the other.
Again, the utility company maintains control of the throw-over switch and decides which feed powers each individual facility on the site. The risk of line cutting is smaller than the per-site option, since there are two individual feeds deployed into each building.
However, switching the power source between substations is still conducted on a large scale: per-facility. This means that at any given time everything within one facility runs on power from a single source. There is no ability for a facility to spread its load across multiple power feeds, and, similar to the per-site configuration, there remains a single point of failure for the entire building. Because of this, failure due to maintenance is still a large concern in the per-facility model.
The best configuration brings multiple utility feeds to each UPS lineup within the building. Each UPS lineup has the option of utilizing power from either one of the substations. There is no single point of failure in this configuration, and power loads within the facility can be shared.
Typically, facilities using this model will run a portion of their UPS systems on one substation and the remaining portion on the other. In the event that one substation or power feed fails, an entire facility will not fail, and there is a lower load to be switched over to the redundant power source.
Data Foundry has a per-UPS configuration at its Texas 1 data center. The facility’s six UPS units share the load 50/50.
Operators have more control with this configuration, however, it is more expensive to implement and only found in premium data centers. The facility’s staff also assumes additional responsibility for proper maintenance and for execution of the throw-over from one power source to another when a substation fails.
When deciding how to take two diverse utility feeds into a facility, factors such as cost, risk and control must be considered. In light of these factors, each company should carefully consider its resources in the design phase of a construction project to decide which of the three configurations works best for their specific case.
About the author:
Edward Henigin serves as CTO of Data Foundry and is also one of the company’s shareholders. He has a physics degree from Bradley University.