Data Center Power: Ignore At Your Peril

Wednesday Feb 23rd 2005 by George Spafford
Share:

Underlying all of IT's fancy servers, networking gear and other devices is the need for clean, reliable power.

Underlying all of IT's fancy servers, networking gear and other devices is the need for clean, reliable power.

This basic foundation is often taken for granted and not sufficiently reviewed. As a result, IT may be forced to deal with excessive amounts of equipment failure, corrupted data and overall service-level issues. This topic will be covered in two articles. This first one will focus on some basic project elements that are needed; the second will cover issues for consideration.

As with any project, a basic plan is needed. This helps facilitate improved communications and better results. Within the basic project plan, the following task categories and suggestions are given for review:

Identify Stakeholders

Don't forget that other groups in your organization may be very familiar with power issues and be great resources to utilize. This includes groups such as facilities maintenance, manufacturing operations, manufacturing engineering, and so on. All parties who can significantly contribute to the solution or be impacted by the result should be considered a stakeholder.

Talk With The Utility

A good starting point is to contact your local power company and have a meeting with your customer representative. Topics to discuss include past outages, typical time to repair, highest risks and what their service levels are.

Ask for a map of the power grid in your area to understand what heavy users are on your segment. Point blank, ask the representative what risks you should prepare for. What are the longest outages on average? What other issues are common in the area in terms of spikes, sags, harmonics, etc. Then, ask how the power company can help. There are things they will do for free and others for a fee.

Site Survey

Either through the power company, or a consultant, install a power line monitor to collect data about spikes, sags, harmonics and so on. The duration of the survey should be based on a combination of what the IT team suspects and external recommendations. If a foundry is next door and they turn the furnaces on at 6 a.m. and off at 6 p.m., then be sure to monitor the lines during those times. This is predicated on you understanding the power grid you are on and who you share the line with.

Interviews

Talk to IT, manufacturing, management and other organizations in your area to understand what they have experienced and avenues they have taken to deal with risks associated with the power. Move beyond IT to better leverage the experiences of others both in terms of the state of power quality as well as potential solutions that they employed.

Analyze the Results

Take the results of the utility meeting, site survey and interviews and review them with the IT team to understand threats and see what possible compensating solutions the team can come up with. Having good representation from people who understand issues surrounding power quality can be very beneficial. Objective third-party consultants can be very helpful in terms of the unbiased recommendations that they can offer.

Develop a List of Requirements

The next step is to generate a list of requirements for power to the data center. This is shaped, in part, by service level commitments, power consumption, planned expansion, sensitivity of the electronics to power quality, etc. Armed with a list of requirements, it is time to open talks with vendors.

Engage Vendors

Part of the interview process should be a discussion of vendors that other groups have used. This should be coupled with internal experience and a review of solutions in the industry trade magazines and websites.

Aim to get at least three vendors involved and be sure to research them. The old saying of "trust but verify" must always echo in your thoughts. Require formal written proposals so the solutions can be compared, questions assembled and vendors held accountable. If a legal issue ever arises, you'll want everything in writing to prove what was said and done.

Additional Vendor Selection Criteria

In addition to internally developed criteria, consider including the following elements as well:

  • Installing the system -- Does the vendor have experience and/or good plans in terms of how to install the new system with our creating unplanned disruptions and minimal planned disruptions to operations?
  • Testing the system -- How will the vendor test the system before it is put into production? Does it seem like a valid plan? How will the system be tested in the future to make sure that it is reliable and functioning as planned?
  • Cutover planning -- How do they plan to do the actual change-over from the existing to the new system? In steps at night during the week? Over a weekend? Over several weekends? What power protection will be in place during the cutover?
  • Experience -- Does the vendor have sufficient experience? Look at the knowledge of the equipment they are selling, the implementation and support services as well. Do they have enough experience in your industry to understand your needs?
  • Longevity -- In addition to a vendor's experience, also look at their life expectancy. It defeats the purpose to buy the best solution if the only vendor who can support it goes out of business.
  • Liability -- Who will be accountable if there are problems? Should liability be pushed on the vendor? Should some be accepted? Is insurance needed? The point is that the liability of a power failure and failure of any backup systems needs to be considered. Once again, contingency plans should be driven on the basis or risk.
  • Maintenance -- Power systems aren't a "buy and forget" investment. They need to be routinely tested, batteries replaced, fuel checked and/or swapped, engines cleaned, etc. There are far too many examples of people relying on huge UPS (uninterruptible power supply) systems and when the power failed, the UPS failed due to the batteries not being routinely replaced.

    For big systems, resist the temptation to do internal maintenance. These systems can be very complex and the better left to trained technicians. Similarly, if they are maintained in-house, the liability related to failure is fully absorbed by the organization.

  • Support -- Look at how the vendors answer questions and/or how quickly they can dispatch a technician in the event of a problem. 24x7 phone support is one thing 24x7 4-hour on-site response with parts is quite another.
  • Vendor Selection

    Take the requirements list and establish a weighting for each point. Those items that are most important get a weight of five and then go down the scale to least important which get a one. Then score each vendor with a one for a low score to a five for a high score. As a result, the vendor with the highest score should be the best, the next best vendor will have the second highest score and so on. This allows for fairly objective ranking of the vendors on the basis of merit. Stakeholders should weigh in on each and an average for each score developed.

    Summary

    IT systems can't run without power and their reliability and availability are directly linked with the quality of the power supplied. Power can't be taken for granted and if it hasn't been reviewed, now is an ideal time to start such a project. In the next part of our article, we will review topics for consideration.

    Share:
    Home
    Mobile Site | Full Site
    Copyright 2017 © QuinStreet Inc. All Rights Reserved