This form does not yet contain any fields.

    Tin Whiskers

    Tin whiskers form when tin plating on component leads is under pressure.  The tin squeezes out to form filaments as thin as a human hair.  The whiskers grow long enough to potentially cause a short circuit.  This concern has come about since the European Union adopted restricition of hazardous substances, which has led manufacturers to start using tin instead of a mixture with lead, the objectionable element. 

    The issue of tin whiskers is less likely than solder joint thermal failure or mechanical disk drive failure, but the cause can be hidden and diagnosing the issue can be a huge expense.  There are a few undertakings by the industry to mitigate this problem, such as mixing the tin with other materials or using thin coatings, such as a layer of nickel to prevent copper from diffusing into the tin. 

    While there is a little a facility can do to prevent a problem caused inside these components, keeping a clean data center can help with this and other failure issues.  Keeping a data center clean helps remove contaminants that can enter computer equipment and potentially dislodge tin whiskers that cause short circuits. 


    Data Center Infrastructure Management (DCIM) Tools

    DCIM: What used to be a much simplier subject a few years ago has blossomed into a fast-developing part of the data center industry.  Today it seems like there are many qualified tools and packages on the market to help with managing your white space, but also much beyond.  This has mainly come from the outside in - facilities support software tools were developed and applied to help with the IT infrastructure instead of IT management tools growing beyond the white space to track more than IT.  We'll be following the many tools as compared by Gartner and other third party reviewers as well as asking and reviewing more from the providers themselves. 


    CRAC vs. CRAH

    I had this come up recently... what is a CRAC (pronounced 'crack')?  What is different from it than a CRAH (pronounced 'craw' or 'cray')?

    CRAC = Computer Room Air Conditioner

    A CRAC is often thought of as having an internal compressor, thus not needing the support of a centralized chilled water system.  But this also means it has to reject heat somewhere - like to a coupled rooftop drycooler.  A CRAC is also thought of as having refrigerant internal to the system, and with those extra motors and parts they need more maintenance care and have higher failure rates.

    CRAH = Computer Room Air Handler

    A CRAH is essentially a specialized air handling unit for computer room environments.  Most are a simple fan arrangement with an internal coil supported by a centralized chilled water system.  

    Clients may decide to go with one nomenclature or another, but knowing the difference can be important when working with design engineers, operations personnel and clients.


    What do we want from CFD?

    Computational Fluid Dynamics (CFD) analysis of data centers has been going on for some time, but how does it help us?  For an existing data center, do you want a CFD model to show you things you already know, such as hot or cold spots? 

    90% of the time we are using a CFD report to help with data center arrangements, comparing scenarios and layouts to best suit a given space.  Sometimes this is for new data centers and sometimes this is for existing spaces that are looking to add equipment – and in either case the data center operations team wants to know what will happen. 

    Does a CFD report tell the truth?  To a degree, the truth is given, but not absolutely.  Despite even the best of efforts, there are many, many nuances in a real data center than just cannot be completely captured in a virtual data center.  But we can come close.  Leakage rates, factors for under floor (or overhead) obstructions can be incorporated, bank-off panels can be found, and more and more to make a CFD model closer to the actual.  But we reach a point of diminishing returns, where adding more detail does not affect the depicted performance. 

    What a CFD model doesn’t do is solve our problems for us.  We need to establish what we can change and what we can’t, then take those scenarios and crunch the numbers to see what improvements we might be able to achieve.  A CFD report can definitely help with finding the best choice for potential layouts, but only from the scenarios given.  A model will not start to generate best and worst options for adding equipment to our data centers… at least not quite yet. 

    So we are back to making our own decisions.  With a CFD modeling report to help us, we can make more informed decisions, which often lead to the capabilities of allowing more equipment to be added and in a better location while also helping to plan for future equipment refreshes in our data centers. 


    Cooling Tower Water Storage

    When operating your data center, often there are backup measures to keep things running during a power outage.  Generators may be the first thing that comes to mind, as they bring power back during an outage, but along with that the cooling systems also need operate.  As the heat is rejected through the mechanical systems, the last item that rejects the heat to the atmosphere is often the cooling towers.  To operate continuously, they need to maintain enough water through a loss of make-up water, as 1% or more of the water is evaporated or lost to drift conditions.  In addition to this, the blow down cycle may keep operating during a power outage, which may double that water loss. 

    Ok, so perhaps 2% losses don’t seem so bad… but added up over a day or more of operation the water levels in the towers will run low.  If they run too low they don’t work… and therefore no more cooling for your data center. 

    Would the typical make-up water flow be sufficient?  It just may be – but sometimes when a location loses power for an extended period, the utility companies may not be able to provide an adequate flow.  Sometimes this happens right away, sometimes the pressure and/or flow begins to taper down. 

    Or perhaps there are other issues with the water supply that were unforeseen.  It could even happen without a power outage, such as upstream construction accidently taking down your only supply pipe from the utility.  What’s the plan for that? 

    Many cooling towers have an integral sump that collects the cooler water at the bottom of the tower.  This sump depth has a minimum to feed the condenser water pumps, but not necessarily too deep, as this adds material cost and weight.   Other towers might be field-built, with a separate sump beneath the tower, and yet more might be a combination of a factory tower with a field built sump. 

    All cooling tower sumps provide enough water for the chiller and cooling tower to operate for some time without make-up water, but the volume of the sump will dictate how long.  Increasing the water that can be stored in the sump not only increases the amount of operating time from hours to days, it also helps to create a more uniform water temperature entering the chillers.  Another means would be to provide a separate water source, such as a tank, on the site. 

    How big would a tank or how deep should the sumps be?  A good plan would be to match what the generators can do, so let’s go back to the generators and see what they are doing.  For example, if they are supporting 50% of the IT power load (and the other power needs to support that load) for 2 days, then the cooling plant should also be able to support 50% of the IT cooling load (and the other cooling needs to support that load) for 2 days.  By stepping backward through the load needed and what would be expected of the cooling towers, a total water volume can be estimated during a peak period. 

    Page 1 ... 2 3 4 5 6 ... 7 Next 5 Entries »