The Power Protection Blog

The decision to create a Tier III Data Centre is a strategic one, usually as as result of the necessity of creating an extremely fault tolerant system, either for customer demand (website hosting for example) or for business continuity reasons (credit card processing, military, financial etc.).

The Tier III data centre was conceived when computer systems were introduced that had dual power supplies. The basis of the Tier III centre is that there is one power path with redundancy, and an alternate power path:

Tier III Data Center Power Flow

Note that essential cooling is now added to the UPS output, whereas with Tier I and II it was expected that the site could cope with short downtime of cooling during which time the generators would power up and restart the system. In Tier III this is not allowed and cooling is continuous. This needs to be borne in mind when selecting the type and size of UPS. It is for this reason that you will often find rotary UPS systems (with their ability to handle mechanical loads better than static systems) used for such applications, although this is by no means a requirement.

The computer systems are dual powered (commonly referred to as A & B inputs). If a computer system is utilised without dual inputs it is expected that a local Static Transfer Switch (STS) is utilised.

In the diagram above we will assume that the primary power path is A, and the secondary power path is B. The Static Transfer Switch (STS) also has primary and secondary inputs, and the primary input is also taken from Path A. The computers are therefore supplied by two power inputs, all of which is provided by the UPS systems. Should the primary power fail anywhere along Path A, then the Static Transfer Switch will revert to its secondary input and continue to supply power along Path B to the load.

NOTE: Static Transfer Switches are capable of switching within μsecs between their A & B inputs provided the two sources are synchronised. If they are not then there will be a switching delay. This is particularly important when you consider the issue of selectivity, i.e. the ability of the source to clear a fault. In order to achieve this selectivity, UPS are synchoronised with their bypass input. Should a fault occur they can switch to bypass instantaneously (quickly anyway) which then will allow a greater fault current allowing the fault to be cleared quickly (ie pop a fuse or trip a breaker) without causing disruption to other equipment on parallel circuits. This means that the primary and secondary sources should be synchronised. Make sure this can be done! Another factor of having unsynchronised inputs is for the potential of having 400V (not 230V) AC within the computer room cabinet.

As you can see, a Tier III is inherently more robust as it will allow failure along the entire path without power being lost. This is what classifies a Tier III system – it is basically a Tier II system with an alternate power path, derived from a seperate source. A Tier III centre has an availability of 99.982% which equates to 1.6 hours of downtime per year.

So how does this help the average computer room user? Well, Tier III is probably way over the top for an SME computer room. I have known small financial companies that require the fault tolerance of the Tier III infrastructure, however Tier III is more strategic and therefore the site is designed from the beginning with Tier III in mind. It is difficult to post fit a Tier III system without severe disruption to the existing business. However, if you required just a little more protection against unplanned outages, it may not be too difficult to install a secondary power path and an STS to feed your computers. Or another alternative may be to look at a halfway house for Tier IV…….

Tier II centres encompass all the features of Tier I centres with the addition of redundant critical power and cooling components.

Tier II Data Centre Power Path

Each component must be capable of operating if the other component fails. This is typically achieved with n+1 redundancy. What this means is that if ‘n’ eg 2, modules are required to support the load then install ‘n+1′, i.e. 3. There is a lot of debate as to what determines true redundancy. For example, some manufacturers have seperate UPS modules, controlled by a single controller. If the controller fails then the system fails, so this is not true redundancy, although they may argue that the controller is designed with redundancy built in.

Some UPS modules don’t have an internal static bypass and instead rely upon a wrap around static bypass. The argument here is that one large wrap around is more robust than several smaller ones, as smaller systems may blow up one by one due to a race condition in the event of a fault. My belief is that the latter is unlikely as static switching can occur in μseconds, probably about a thousand times quicker than the time needed to damage the single static switch. In any case, the static switch is usually rated many times higher than the nominal load current to accommodate fault currents. However, the wrap around is now a single point of failure in the system – although if it fails, this will only cause a problem if the system needs to bypass, so is this a problem?. The debate will continue to rage on.

Historically, UPS were configured in what was known as a “hot standby” configuration to achieve redundancy. In this instance two UPS are fed from the utility, but the output of the standby UPS is fed to the bypass input of the primary UPS. The primary UPS provides power and if it fails (and bypasses), the standby unit will then provide UPS power to the load via the bypass of the primary. Works in principle, and can be used with mixed manufacturers and ratings of different UPS systems, however, the primary UPS output is a single point of failure. In addition, should the primary UPS fail, the secondary UPS will instantaneously be expected to deliver from 0 to 100% load immediately. Shouldn’t be a problem, but sometimes it is!

The modern method of achieving redundancy is to share the load equally amongst the UPS modules (this is how all our Kehua Parallel Systems operate). The UPS talk to each other through redundant communications ports and no one UPS is master over the others. If any UPS should fail, the UPS is isolated from the others automatically.

Enough about redundancy, the Tier II centre is more robust than the Tier I centre however still has one power path and therfore there are times during faults or planned maintenance that the computers have to be powered down. As a result a Tier II centre has availability of 99.741% equating to 22.7 hours of downtime per year. Not much better than Tier I on the face of it (28.8 hours), however Tier II is more robust against unplanned outages.

So how does this impact the normal computer room? As said under Tier I then it depends upon the financial impact of downtime to your business. For SME’s that rely on computer systems, but will all go to the pub if the power is off, then there is perhaps no need to keep the computers running for hours on end, simply shut down gracefully and that’s that. Where you don’t want to have to shut down the system to perform maintenance on the UPS, and don’t want to leave the system vulnerable to power cuts or surges when using an external bypass switch, then you will need redundancy. Like all things in life, it’s a choice based upon your needs and wants.

It’s worth stating that Tier Standards are only a guide as to the robustness of a site against outages, there is no standard or law dictating that this is the way it should be. Use the information as a guide to what is best for your business.

Tier I data centres provide a dedicated site infrastructure to support IT Systems and include:

• A dedicated space for IT systems
• A UPS to filter power spikes, sags and protect against momentary outages
• Dedicated Cooling Equipment
• A Backup Generator to protect against prolonged outages

There is a single power path delivering power to the load and redundancy is not required. As a result any component or distribution path failure will impact the computer systems.

Standby Power Flow Schematic for a Tier I Data Center

During normal operation the UPS is providing clean power and protecting the load. A short term power outage will see the UPS continue to provide power to the computers but the cooling system will be shut down. During extended outages the generators will activate allowing continuing operation of the computers and the cooling system will restart. Any planned work will more than likely require the computer systems to be shut down.

Tier I data Centres have an availability of 99.671%, which equates to over 28.8 hours downtime per year (Planned and Unplanned).

For a typical computer room, a Tier 1 set up is more than likely adequate, with the addition perhaps of a redundant UPS module (see Tier II). The use of a generator is optional and dependent upon the impact of downtime to the business. The IT equipment can be configured to shut down gracefully in the event of a extended power failure, and the fact that lost data has been avoided is probably acceptable for many businesses.

Over the next week we’ll be blogging about how UPS are required and configured to achieve Tier I, Tier II, Tier III and Tier IV levels.

The Tier system has been developed by The Uptime Institute as an objective basis for comparing the capabilites of one particular design topology over another or to compare groups of sites.

We will be showing that (at least as far as power is concerned) how we can have move from Tier to Tier level depending upon the requirements of the data center, and how best to utilise these classifications into everday computer rooms. After all, not every business will need the stability of a modern data center.

Watch this space.