Data Center Relocation - Computer Room Design

ABR'S E-NEWSLETTER
A PERIODIC PUBLICATION BY THE ABR CONSULTING GROUP, INC./
NO. 01 VOLUME 01 - May 11, 2001

Wait Times for Long-Lead Environmental Equipment

What a difference a year makes. This time last year, there was a 62-week lead time for the most popular Tate raised floor tiles. Can you believe 62 weeks!!?. The most popular Liebert A/C units had a 8-10 month lead time. Want a new data center? Wait 15 months. It wasn't a shortage, it was red-hot demand. Reportedly, 3 million sq.ft. was tied up by two major telecom carriers. In June of last year, we had a large project that was reduced in size thus producing a surplus of 15,000 sq.ft. of Tate raised floor components. At the time, these components were so hard to get, we were seriously considering putting them up for bid on EBay. We didn't. We even had one major customer accept wood-core flooring because they needed their data center in 6 months and wood-core was all that was available.

Today, the Lieberts are down to the time to build them (60-90 days). The Tate raised floor components are in stock.

Emergency Generator Testing and Rolling Blackouts

Big problem here. With the onset of rolling blackouts in California, some data centers are quickly finding out that their emergency backup systems are not functioning as they expected. In fact, some are not functioning at all. Generators are not energizing or, when they do, electrical connections are not correct. The result is the sudden silence of A/C units, servers and mainframes. Hint: You never want to hear a pin drop in a functioning computer room.  The cause - improper testing, lack of testing and/or lack of proper maintenance.

You must periodically test the emergency generator systems regardless of any threat of blackouts. This is especially important for essential services sites such as hospitals, police stations, County centers, 911 operations, etc. You never know when a nearby accident or other circumstance will take out your power.

Hopefully, your periodic testing is a method that includes the UPS/battery as battery failure is a common culprit for emergency system failures. Emergency systems should be tested monthly at best and quarterly at worst. Once a year testing is not acceptable and invites failures in the system. Lack of maintenance on the emergency equipment will also lead to failures, especially the UPS battery. Unfortunately, rolling blackouts are forcing the issue.

About Testing Emergency Backup Systems

Most data centers are designed so that the emergency standby generator supports the data center and possibly other IT areas but not the general office spaces in the buildings in which these data centers reside. This design restricts the complete testing of the emergency generator system in that you cannot cut municipal power to fully simulate a real loss of power. Dropping municipal power will be fine for testing the systems for the data center but all power to the general office spaces will be lost during this type of generator test. Not a good trade off. Instead, many data centers rig their tests so that (1) municipal power is not interrupted, (2) the generator is energized by tricking it to sense a loss of power, and (3) general office spaces are not disturbed.

This is not our preferred method of testing a generator and its capability to keep the data center alive for two principal reasons. First, a test of the generator by not dropping municipal power leaves open the remote possibility that when the real municipal power is lost, the generator will not energize properly. Second, the generator is tested without the load of the data center it is designed to support. Testing the generator without a load is not good for the generator.

A third reason for some data centers is that the generator testing is so independent, that the remainder of the backup system (UPS, battery load, automatic transfer switches, etc.) are not included in the test.   This increases the risk that some portion of the data center emergency backup system will not operate properly during a sudden loss of municipal power.

Our preferred method of testing a data center backup takes a bit of courage but, in our opinion, is the one true method of making sure you have the best possible chance of operating fully during a loss of power. It's very simple. Once a month, go to the main power panel and drop the municipal power. Let the entire data center sink or swim on the results. Most data centers are not setup to run this type of test without bringing down the rest of their building and for those that can, they're chicken to pull the switch. Most will opt for just testing the generator. We have witnessed CIOs and Operations Managers that demand such a strict test actually conduct these tests on a monthly basis. They believe, as we do, that if the data center won't work under controlled emergency conditions, it won't work under a real loss.

Saving Money When Relocating Data Centers with IBM Mainframes

We have several hints for you when relocating data centers containing IBM mainframes and large IBM peripherals. First, try to time your move event at the same time you need to upgrade your mainframe. It's not as hard as you may think. IBM provides a good deal of installation and training services as part of the purchase or upgrade. Money to be saved here as many of these services dovetail with your relocation planning. The best news is that you don't have to move any mainframes. You can install the new mainframe, completely test the system and have it ready for the software migration all without serious time pressures. The older mainframe is left behind as part of the relocation event for IBM to de-install. That's part of the deal too. Second, consider a serial number swap. That's where IBM will pre-install a like system to the one you need to move in your new data center and will de-install your existing system after the move. IBM will then swap the serial numbers on the two systems so that your software connects with the serial number and your ongoing warranties remain as is. This is especially useful for complex systems that take 2-4 days to de-install and the same to re-install. An IBM 3495 tape storage system would be a prime example where this type of swap would work.

Lastly, a serial number swap can be valuable when you cannot afford the equipment downtime. Think about a bank and its ATMs. The ATMs can blink as you cut them over but you can't take them down for a day or two. In this type of move, temporary or new systems are set up in the new data center and brought online. This is referred to as seed equipment. This can be a very expensive, but necessary method of keeping your business online to the customer. If you're using IBM equipment, you can save money with serial number swaps.

Cat 6 Cable Certification On The Horizon?

Getting closer. The June, 2001 meeting notes of the TIA committee 42, sub-committee TR-42.7, the sub-committee (the committee that deals with Cat 6), indicates that the sub-committee is trying to resolve final issues with Cat 6. They still need to finalize and send a ballot to all members of the sub-committee. Still looks a ways off. A fully developed Cat 6 cable will permit the use all 4-pairs of the cable. Full duplex transmission will take place. If you wish to follow this sub-committee or any activity of the TIA, go to www.tiaonline.org.

How Much Equipment Yard Do I Need For My Data Center?

On data centers that are 10,000 sq.ft. and above, you need roughly 1 sq.ft. of equipment yard for every 2 sq.ft. of data center spaces (includes NOC, printer areas, etc.)

-------------------------------------------------------------------------------------------------------

Periodically, ABR Consulting Group, Inc. will e-mail you condensed information for the ever-changing and every-dynamic world of designing, planning and relocating data centers, server rooms, labs and other key IT/IS spaces. All you have to do is place your e-mail address below and click on subscribe. You can unsubscribe at any time by entering your e-mail address and clicking on remove.