If the power goes down, you are down!! Total power outages and brown outs result in severe switch and router systems failures causing massive disruptions, not only to local networks connecting end customers to the internet, but to all back-end private network connections, backbone links to the internet itself and overall system failures. Customers' servers within these data centers will experience the same problems including; hard drives, processors, memory and motherboard failures. Therefore, maximum power uptime is the most critical of all data center systems! Good power systems keep you up and protect your equipment.

Mistake #1 - Generator Problems

Generators are the primary source of back-up power during a power outage. It is important that the generators are well maintained, of sufficient size to handle 1.5 times the expected total building load, and is tested under load at least once each month.

One of the main problems is that some colocation facilities that started small, or that were not built from the ground up as a colocation facility, will have multiple small generators. This is because, as they grew, the first generator installed was not large enough to support the entire facility. As their colocation sales increased, the need for more generator power and a lack of funds created a situation in which they purchased more small generators and divided the power grid into multiple sections each supported by an individual generator. This is often sold by the sales people at these facilities as redundant generators. In reality all the generators will need to start and transfer switches will need to transfer to bring up the entire facility. In this situation, instead of additional redundancy, what they really have is multiple points of failure. Therefore, beware of small, under funded facilities with multiple generators. You should ask questions if you see multiple generators and they are claiming generator redundancy. Ask questions such as, 'Show me the equipment that transfers the power from one generator to another in the event of a generator failure also known as the paralleling gear. If they can't show you, they do not have redundant generators.

Mistake #2 - Transfer Switch Failures

Small facilities often have mechanical transfer switches that transfer the power from the utility to the generator in the event of a power outage. Larger, more reliable facilities will typically eliminate the transfer switch with a system of special circuit breakers that transfer the load during an outage. Furthermore, better facilities will have two systems of circuit breakers, or redundant power grids, in case there is a failure of a circuit breaker. The circuit breaker type of system is considered much more reliable than a transfer switch. Many facilities have experienced transfer switch failures, including one of the author's first small facilities. So ask your colocation sales person to show you how they transfer power in the event of a power failure. Look for circuit breakers instead of transfer switches, and look for redundant methods to transfer the power from utility to generator. This is one of the most common points of failure in a colocation facility. Make sure that there is redundancy in the power transfer system.

Mistake #3 - Non redundant UPSs (Uninterruptible Power Supplies)

Redundant UPS systems are absolutely critical to ensure maximum uptime. UPS systems are not just a stack of batteries that hold the load while the generator comes online in 15 seconds. They are very complicated and have systems that clean and monitor the power. There is a computer within them and they all have multiple points of failure. Regardless of the manufacturer's claims, whether it be APC, Liebert, Powerware or others, these UPS systems do fail, and do so fairly frequently. Make sure that your colocation facility has redundant UPS systems. When I say redundant, I don't mean two UPSs. You need to verify that they are set up in a n+1 configuration, which means that, if there is a failure of one UPS, the other one will handle the load immediately without interruption of power. Have your colocation salesperson explain in detail how the system works and how one system fails over to the second system. Have them show you the hardware that makes this happen. True redundant systems are considered paralyzing redundant systems. That is, they share the overall load. But any one of the online units should be able to take 100% of the full system's demand in the event of a single unit failure.

Two UPSs often means that the first UPS was not large enough to handle the load and they bought more UPS systems as they grew just like the generator scenario above. But they are not redundant. Furthermore, even if the there are multiple UPS systems running in an n+1 or redundant configuration, you need to make sure that there is sufficient capacity on the redundant UPS system to handle the full load of the entire facility in the event of a failure. In a situation in which a colocation facility has two UPSs and claims redundancy, are they running both UPS systems below 50% capacity? If not, during a failure of the first UPS, the second UPS needs to handle the load. If both UPSs are running on a common bus, which is typical in a redundant situation, and they are both running at 60% capacity, be aware of what happens when one of the UPSs running at 60%, fails over to another UPS running at 60%? You then have a total of 120% load going to the second UPS resulting in a second UPS failure that will cause the entire facility to go down. This also applies to facilities with more than two UPS systems. If the facility has three, they should be running at a maximum of 66% capacity. And if there are four, - - 75% capacity and so on.

You can ask questions such as "What percent of capacity are you running your UPSs?" Have them show you the display on the front of the UPS that shows the load. Ask to inspect the power room, the power diagram or, on the white board, how their redundant UPS systems operate.

Mistake #4 - No Redundant power grids

Typically, if you find a high-end facility with all this power redundancy built- in, they will have multiple PDUs (power distribution units) around the facility connected to the redundant UPS systems and power grids. Ask to see the redundant PDUs, and if you need maximum uptime, request two power circuits - one from each power grid, or PDU, to your colocation cabinet or cage. This way, if your facility has a power failure, it will typically be across just one grid and you will still have power. You can plug one server power supply into one circuit, and the second to the second power circuit. This may seem like overkill, but if your business is online, and "going down" will cost you money, it may be well worth the extra $200 bucks per month so you can sleep at night. Don't forget if you server's power supplies share the load and you have multiple servers you must run your redundant circuits at less than 50% capacity or you will blow the breaker on the redundant circuit in the event the primary circuit fails.

In rare cases some data centers will have redundant power connections to the utility from two different power substations. This while rare can add one more layer of redundancy to ensure that a power outage will not affect the data center. Instead of starting generators this facilities can simply switch between power grids if there is an outage.

Power redundancy should be a primary concern when choosing a colocation facility. Make sure that your data center actually has it. Don't believe everything you read or hear. Verify all information with a facility tour, and make sure that you actually see the equipment and have them explain to you how it operates. Ask to have the data center manager attend the tour so that he can explain directly to you, in detail, how the redundant systems operate. If they can't demonstrate their redundant systems, move on. There are typically several facilities with this level of redundancy in a major metro area that don't charge any more for it than those without it. If you are buying from a large, nationally known provider, do not assume that they have all the redundancy built-in. Many of those large providers bought other small providers that did not have the millions it takes to build this type of redundancy, and they still operate these small, non-redundant facilities.