FACILITIES capacity
During a facility’s design phase assumptions are made about the types of equipment that will be deployed throughout its lifecycle. Though usually based upon the best available data from the IT department, these predictions are often rendered obsolete even before the data centre opens, as the industry changes technological tack once again.
“Think of the problems encountered when blades first appeared,” continues co-author David King, Consultant Engineer at Future Facilities. “Such change makes it inevitable that original design assumptions will be broken – in fact a data centre with a 10-15 year lifespan will experience up to three complete refreshes, any of which could result in design capacity being lost if not managed carefully.”
What is capacity and how is it lost? Data centre designers apply a formula for a new build that adds the design capacity, which is the total amount of IT equipment it is due to house (measured in kW or MW), to an element of redundant capacity that will allow for at least a degree of future change or expansion.
All the key specifications, such as cabinet layout, power density, network connectivity and cooling system, will be calculated using the design capacity as a basis.
“The problem is that while many capacity planning and forecasting tools assume that unused capacity is readily available, in reality a significant proportion won’t be. We’ve looked into each of the four critical resource components – space, power, network and cooling – to determine why this happens and how it can be prevented,” continues Dr Warner.
Space
As data centres move towards cluster based computing, requiring a large number of rack mount servers in close proximity, the demand for large blocks of contiguous U space is growing. But how often is the need to create or preserve contiguous U space factored into data centre deployment plans?
This example shows a rack where slot 28 (highlighted in green) has been earmarked for a new 4U server. What the installer doesn’t know is that another 4U server, in slot 10 (shown in red) is scheduled for decommissioning, so would be available by the time of the installation. Without this knowledge the installation goes ahead in slot 28 as planned, resulting in the fragmentation of available space. Such occurrences are all too common when data centre managers can’t see the ‘bigger picture’ of future decommissioning and deployment of servers.
Power
There’s a close relationship between all four resource components, but it is particularly apparent between server space and power supply. These are both finite resources within a data centre design; too often, though, a simplistic approach is taken to the allocation of PDUs to cabinets, with total power from a distribution unit divided equally between cabinets regardless of the components they house.
However, the relationship between the amount of space required and the power draw is different for different technologies – for example, rack mounted storage systems use large numbers of low powered disk shelves that can fill a cabinet without coming close to its power limit, whereas servers and blades can use up a cabinet’s full power allocation while occupying just half its available U space.
Future Facilities’ King explains: “Failure to manage these imbalances properly will result in usable capacity being lost. Even if the uncommitted power can be migrated to adjoining racks this will increase the risk of equipment failure.”
Network
Data centre designers must also account for the different ports to power ratios of various server technologies. Rack mount servers can use as many as 10 copper ports for around 500W of processing power, while a blade chassis can provide 5kW of processing power for that same number of ports.
Thus providing a ‘one-size-fits-all’ 48 network ports and 5kW of power
The problem is that while many capacity planning and forecasting tools assume that unused capacity is readily available, in reality a significant proportion won’t be. We’ve looked into each of the four critical resource components – space, power, network and cooling – to determine why this happens and how it can be prevented
June 2012 I
www.dcseurope.info 49
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56