Making Effective And Informed Decisions With Data Center Modeling

By Author: Jamie Hanson
Total Articles: 3534
Comment this article

Knowing just where you are within the data center you work every day would seem something you would obviously know. However, most of us are so consumed with the current problem at hand that we really just don't have the time to step back a take a look at the bigger picture. We all intuitively know that putting out the brush fires in front of us without really understanding how the forest in general is affected by each change can be dangerous at best. We ask ourselves, "Am I really addressing the most urgent problem or just the one that raised itself today?" What is the answer; how can I get the view to determine where I am and where I need to go? The solution is, well at least in terms of your data center, modeling.

So, how do I get a model of my data center? As with any problem you want to solve there are two parts; one is to understan the problem in sufficient detail, and secondly understanding exactly what an ideal solution would look like. We've already discussed the problem, so let's move on to the solution. First, let's decide what data to collect, where to store it, how it shal be retrieved, analyzed, and presented.

To ...
... have an accurate and all encompassing picture of your data center you'll need to collect:

- An detailed rendering of the data center and all physical obstructions

- Dimensional and placement information for all equipment

- Weight of racks, cabinets, devices, and floor mounted equipment

- Power draw and thermal generation characteristics of all active devices

- Power infrastructure of your data center

- Connectivity of devices to infrastructure

Once you've collected all of this data, a non-trivial task to be sure the question becomes where should it be placed to your be advantage. Often this information ends up in an Excel spreadsheet. Although spreadsheets are convenient, this is completel non-optimal for any analysis nor provide automated update. Data stored in spreadsheets without extreme diligence becomes obsolete after the first change is made to the data center. This is likely to occur before you even entered the data into your spreadsheet. So, by the time you're finished, your work is already invalid on the spreadsheet. To avoid this pitfall, you'll wa to store your data in a relational database and interface it to a change management system. This approach will ensure your data is retrievable in any combination desired and will remain accurate and up to date. You can build this yourself or take th less laborious path of purchasing a product designed for data center management which includes a relational database and integrate it to your corporate change management software.

Now that we've got our model data in a database and know that it is assured accurate, we can start displaying and analyzing our model. Visual representations of models provide an effective high level overview of what is taking place in your data cen at a glance. For instance, a top down view of the data center with configurable overlays to display space availability, therma hot spots, and areas overweight are invaluable for identifying big issues needing attention. Without visualization, these woul likely be buried in the data and be invisible until they cause major issues or an outage in data center operation.

Other areas of concern, such as utilization of the power distribution infrastructure lends itself less to visual display, but are critically important to understand in effectively operating your data center. A key example is to understand individual utilizat of power circuits both from balanced use perspective and redundancy for critical devices. Often, data centers that employ th phase power have limited or no visibility into how effectively they are using the legs of those circuits. This leads to unused capacity or overtaxed circuits. Both can have a large impact to your data center. Data center managers may often think the have redundant power supplied to their mission critical devices, but analysis through modeling will show that both device pow supplies are ultimately connected to the same circuit. If the A circuit should fail, the B circuit would be overdrawn at 120% yielding no redundancy at all in this case. Without modeling, it would be very difficult to uncover either of these example problems.

Modeling can also help communicate things that are otherwise difficult to explain. For instance, a simple walk through the da center might indicate that there is plenty of open space for new equipment, but that simple visual inspection belies the fact th open space is not the only consideration for the addition of new equipment. Data center equipment requires a great deal of supporting infrastructure to operate, i.e. power distribution and backup, cooling, signal and bandwidth availability and adequa physical clearance as required by safety and other standards. Modeling software can produce displays and reports that show areas served for infrastructure components, such as HVAC systems and/or PDU's and offset clearances for these required standards. Such a visual display quickly makes it clear, although there may be floor space it is not suitable for immediate placement of new equipment.

What other sort of benefits should be expected from modeling? With a model, you can create anything desired without the effort and expense of actually building and deploying, but can determine in advance what plans will be required. A simple example is designing a rack based solution for a new service offering. With modeling, you'll quickly be able to determine if it will fit, how much power it will consume, how much heat it will generate, and how much it will weigh. All this information is very important to know, especially if you're going to place this solution at a remote location or co-location facility that charge on the above metrics. Also, a good modeling solution should be able to produce an assembly report to ensure the solution yo designed is the one that gets implemented.

But, what about future planning? Once you start using your modeling solution it will not only know where you are now, but know where you've been and at what rate of change you are traveling forward. Thus, your modeling software should be able predict where you are going and when you'll get there, somewhat like a GPS. Trending reports can help you determine your current growth rate and when you'll exceed your current infrastructure. Since new facilities or upgrades to current ones take months to plan and budget and even longer to execute, knowing when you hit critical thresholds is of the utmost importance. In a related manner, being able to use your model to consider "what if?" scenarios is also extremely valuable.

Again, without actually doing anything in the physical world you can see how new equipment purchase decisions would play o in your data center. You can determine if packing more capacity into less rack units really is the panacea the vendor is extoll or have you really only transferred expense from space consumption to cooling because you're now packing more heat generation into the same space. Also, if there are savings, when does the business break even on that expense? Well designed modeling software should do clarify these questions.

In summary, with an accurate model of your data center and software that provides analysis, reporting and meaningful displays, you can quickly and easily determine the big picture and have confidence the decisions you are making are the correct ones.

Rackwise data center manager software is an enterprise class application for data center management. http://www.rackwise.com