The last two weeks have likely been a challenging time for a server room manager to sleep at night.  With one snow storm, and one ice storm reliable power is the hot topic!

As power is slowly restored to the affected areas, it is interesting to see many empty parking lots along the Route 202 corridor.  Likely some of these businesses are having second thoughts about their IT preparedness strategy.  Losing power for a few hours is one thing, but being without power for days is a whole different challenge.

If you have a generator it probably took the load before your UPS batteries went dead.  For managers relying on rack-mounted UPS systems to carry them through probably experienced a few unwelcome surprises – shorter battery run-times than expected and unplanned server shutdowns, i.e. crashes.  Depending on the type of UPS technology you have, your lack of run-time may have you looking for better technology.  What type you have now doesn’t really matter because going forward, you want to only buy On-Line Double Conversion units, not Line Interactive.

UPS units – even rack-mounted ones – are designed to protect against fluctuations in the utility power quality as well as provide battery backup in the event of an outright failure.  Line interactive UPS units are very commonly used.  The problem is that they rely on the battery for more than just utility power loss.  Frequency or voltage variations trigger the load to shift to the battery, and even though it may only be energized for a short period of time these frequent ‘hits’ curtail their life.  Double-conversion UPS units, as the name implies converts the utility AC power feed to DC, and then regenerates a clean AC feed to the connected load (where it is them internally converted back to DC!).  The only time the battery pack is energized is if there is an outright power failure.  The benefit is longer battery runtimes. More Information Click Here.

UPS batteries however have a finite life.  Just like your flashlight or car battery, they degrade with time and load.  With all the tasks IT Managers have, maintain an updated UPS battery replacement schedule likely isn’t high on the list.  After experiences like this week – maybe it gets moved up but that is hindsight. 

After utility power is restored the next issue to deal with is replacing the batteries.  Replacing them will of course mean another shutdown. Here is an idea – invest in a Liebert MicroPod which allows you to externally bypass the UPS and not have to shutdown anything.  Here is a link to this remarkably inexpensive, highly useful UPS add on.

All this UPS discussion is good for minutes – maybe an hour of battery backup.  After this storm you may be more interested in hours rather than minutes.   For critical devices– like network switches now is the time to consider a DC UPS power solution.  Before telephony went IP, DC power was the de-facto power standard in the industry.  The small windowless telephone buildings you see were built to provide local switching for a DC-powered network.  It is tried and true technology that did not migrate quickly to the server room.  Now however nearly all manufacturers offer DC-powered equivalent products – particularly the switch manufacturers like Cisco.

The biggest advantage for making your network switch DC powered is you can economically power it for HOURS without spending a fortune on batteries or real estate.  IT Managers with disaster experience will deploy this technology if they do not have a backup generator.  The DC option is often overlooked by network designers when selecting a power system.

The Emerson Netsure family of products offers a rack-mounted solution that has the same form factor as an AC-powered UPS. The DC systems offer the significant advantages of scalability and redundancy.  And a string of DC batteries at -48Vdc, is on tenth the number of jars used for an AC UPS with a 480Vdc battery system!  This means that you can get a lot of bang for your buck, meaning longer run times of 4 to 8 to 12 hours (depending on your load) without buying a lot of batteries.  And the rectifiers that convert the AC source to DC, are modular and hot swappable.  That means that you can have enough power to support your load, plus one rectifier for redundancy, and open slots in the chassis to add more rectifiers in the event that your load grows over time.

The DC Plant, distribution breakers, and the batteries can all fit into one rack!

And if some of your loads are AC only, a rack mounted inverter can be used to power those loads off of the DC plant (using the same extended battery backup).

DVL offers a complete line of Emerson Netsure DC systems ranging from 8 watts to 200 watts per rectifier and systems that range from 10 amps to 10,000 amps.  If you’re interested in learning more please download this Emerson Network Power whitepaper “DC Power Systems for Convergent Networks”

What better time to talk about raising the temperature than this frigid and gloomy month of January!  I am not talking about your house or office temperature, but your server room temperature, and specifically the return temperature of the air entering your cooling system.

Much has been written on improving efficiency of server room cooling system by raising the return air temperature.  What is not so well written is how to do this in smaller rooms that may not have the picture-perfect layout of a hot-aisle/cold-aisle design.  Let’s face it, there are many rooms that are oddly shaped, some with partial ceilings and even some with an exterior window wall.  These layouts are never properly addressed in articles we’ve seen.  The short answer may be DON’T, but if your room meets the following criteria you CAN.

1. Understand the only temperature that matters is the temperature of the air entering your servers.  The hot-aisle (exhaust air) will always be hot, that’s physics.  It is important that that air stream is returned to the cooling system as directly as possible. 

2. Sufficient cooling capacity in your dedicated cooling system to support the existing cooling load. Redundant cooling systems – which is always recommended is ideal. 
   What you need to determine is the capacity of your system at the conditions you desire to operate, and then compare it to your applied load.  The applied load should not be guessed!  You can start by looking at the load on your UPS system and then add-in non UPS-connected devices, external heat sources (like floor-mounted transformers that may have ended up in your room) and other seasonal or cyclical loads.  This can be a big task where you should engage available resources to help.  Hint – DVL engineers are happy to assist.  Please contact us if you would like our help.

3. The building system is not heating your server room while your dedicated cooling system is trying to cool it!  The heat introduced by the building heating system should actually be part of the equation in determining your cooling load; but since we’ve seen it too many times it deserves special attention.  It is worth checking any supply air grilles in your room that are not part of your cooling system.

4. The supply air is delivered, ideally by ductwork, to the front of your servers.  This ensures the coldest air is introduced to the servers that need it.  If you have a non-ducted system with either a floor, wall or ceiling-mounted system, you are cooling the space not just the cold-aisle and likely seeing a more even temperature distribution throughout the room.  Raising the room temperature is still a good idea if you meet the first criteria.  Keep in mind the temperature guidelines for data processing equipment as published by ASHRAE extend above 70 degF.

5. You have sufficient temperature monitoring in place that will provide immediate feedback on the temperature entering your most critical servers.  Rack-mounted power distribution strips with temperature sensing capability is a good investment.  The Liebert MPH and MPX family of products have multi-node temp sensor that will allow you to profile your rack from top to bottom.  If you have dual power strips (always a good idea) to power dual-corded devices, you can use one sensor for the front, and one for the back.  The ability to track the temperature difference over time will confirm what your temperature operating range can be.

There is on caveat to all of this, and this applies to the rooms with the perfect layout and design:  If you lose your dedicated cooling system for whatever reason, your temperature safety cushion is much shorter.  Server room temperatures will rise dramatically without cooling.  Maybe you’ve experienced it so you understand the time to get to unacceptable temperatures is shorter, maybe much shorter than before.  Hopefully you didn’t skip over the comment about redundant cooling in #1 above!

Our next posting will continue this conversation with a thermal model of a small server room.

Check back soon to the DVL website, we will be posting two useful tools in helping you understand and calculate temperature rates and efficiency ratings.  Please visit www.dvlnet.com under our TOOLS section. To reach a Data Center Engineer, call us at 215-785-5950.