The UPS Story Continues

A “Simple” Upgrade Becomes a “Project”

power-station-kw by octal, on Flickr

power-station-kw by octal, on Flickr

As I mentioned in A UPS Story, we needed to upgrade our UPS system in our small datacenter. I contacted the vendor of our existing system (vendor A) and had a basic bill of materials drawn up for the replacement of our 40kVA system to a new 100kVA system. It is a modular system that can reuse some of our existing components. About this time, my boss brings in a consultant to manage the project. Her instincts tell her we’ll need it and I have other projects to work on. We’ve worked with him before and he specializes in datacenters. One of the few consultants (it seems) who is actually useful. We needed to get the equipment delivered by year end to get it in the 2010 budget, but for due diligence purposes he suggests we talk to another UPS vendor (vendor B). We already have an investment in vendor A, but it’s worth looking at what’s available to be sure we’re doing the right thing.

Due Diligence

We are looking at a 100kVA system with a runtime of ~25 minutes that takes up three racks in our datacenter. That’s the same footprint as the existing system (that does 40kVA for ~20 minutes), though the incongruity of this doesn’t fully hit us at the time. After a conference call with vendor A, we have a lot of questions and our consultant comes in for an onsite discussion and to take some measurements. We get a mostly comparable quote from vendor B for a conventional (as opposed to modular) UPS. It is markedly cheaper than vendor A’s solution. It is substantially larger as well. Yet the runtime on this is only 11 minutes at our theoretical full load. The existing UPS is inside the datacenter and we can’t squeeze this alternate system in there. Also, the difference in size and runtime between the two vendors seems excessive so much head scratching ensues. We have another conference call with vendor A and realize the system that was quoted with 14 minutes of runtime, but has additional expansion options.

A Weighty Matter

After some discussion, we decide that 10+ minutes is enough time. This is allows for our systems to shut down cleanly in the event our generator does not kick in during an outage. Adding enough battery to give someone time to troubleshoot and get the generator running, let alone time to arrive on site, is unreasonable.

So now we’re deciding where we’re going to put the UPS. Our consultant likes to be thorough, so we start checking the physical infrastructure. Electrical is no problem. The UPS doesn’t put out enough heat to affect the HVAC. However, these lead acid batteries are pretty heavy. Some math is done and we determine that we’re looking at a floor load of over 300lbs/sqft for fully loaded battery cabinets. A structural engineer is brought in to take a peek at things. He wants to know the manufacturer of the raised floor and to see building plans.

Turns out, we don’t know who makes the floor. No idea about its load, though it’s assumed to be at least rated for 250lbs/sqft.

Turns out, the datacenter is in a different location on the building plans. The floor where the datacenter is now located shows a 125lbs/sqft rating, but 250lbs/sqft where it was originally planned. OK, we’ll look at the as-built plans.

Turns out, we don’t have any. The architect firm that designed the building doesn’t, either.

This complicates matters. We end up with the idea of putting the battery cabinets on the first floor of our building in an office space that’s already being remodeled. This avoids our potential floor loading issues. Our electrician has no issues with making this work. Facilities looks into it a bit and suddenly things blow out of proportion really fast. Next thing we know, we have a meeting with a structural engineer, an electrical engineer, our consultant, our facilities director, one of our construction coordinators and an architect.

Details, Details, Details…

How complex can something get? Try converting an office space into a space that is within code for a large lead acid UPS.

  • We still may need to shore up the floor to hold the weight (but much easier on 1st floor than on the upper floor where the datacenter is located)
  • There are spill containment rules that seem to require the ability to contain the acid should all the batteries spontaneously empty themselves onto the floor
  • The room must have a clean agent for fire suppression (such as FM200), most likely with water backup
  • Sealing the room to contain said clean agent
  • New doors & removing windows
  • Potential HVAC requirements
  • Duct work changes, mechanical engineer may be needed
  • 1 hour fire separation requirement, which means fancy ceiling and sheet rock, maybe more

These are some of the issues that came up in our meeting. They are not necessarily an accurate reflection of building codes in your city (or mine). When all is said and done, a project we were hoping to keep under $100k is looking at $500k just for the changes to this room.

Redo From Start

Now we’re back to looking at our options for systems that will fit in the datacenter. The conventional UPS system is out and we’re evaluating rack mount modular designs. We’re thinking about per rack systems, per row systems, or just replacing the existing unit with a larger one. A somewhat expanded scope from where we initially started, but essentially we’re starting all over again. We still don’t know our floor’s load rating, but hopefully the structural engineer will eventually be able to tell us that. We’re hoping it doesn’t turn out to be 125lbs/sqft.


Leave a Reply