Monday, 23 July 2012

Planning for power (1)

I've been reading recently on LinkedIn.com forums questions about how much power should be allowed per cabinet/rack in the Data Centre. Making a mistake with this figure can be expensive. If you calculate too low a power supply it can be a very expensive process to remedy the mistake. Retro fitting power upgrades on a live data centre is an expensive and complex process. It usually comes at the time when a business is not expecting a major capital outlay, not a move popular with the Board. I've seen one case where "glowing bus bars" unable to handle any more power led to an international bank having to hurriedly construct a new data centre in a new location. There was still free space in the old data centre just not enough power for any new equipment for business expansion.

Over-provision for power can also prove expensive in terms of wasted initial capital expenditure, increased maintenance costs and wasted energy during the lifetime of the equipment. 

It would take a book to fully describe the steps needed to calculate and install the right power system. The actual design and installation must be undertaken by qualified electrical engineers, but the management or planners of the data centre have information which is vital to the power system design process. Key to the process is an energy budget and forecast process. The team must be able to measure the usage of existing equipment, predict the actual power usage of  equipment which destined to be installed and also facilitate the calculation of power needed for supporting systems such as air conditioning.

The power budget should be an actual document/spreadsheet/database recording the planned/actual usage of equipment. It should also document the physical location of the equipment so it is possible to calculate how much power should be delivered to a specific location within the data centre. It is normal for power to be delivered over three phases (circuits) of electric power. The engineers will need to know how the power demand will be distributed over the three electrical phases. Cable routes and switching arrangements can be affected by the decision to provide dual power routes to cabinets/rack and their physical location.

Power planning and the physical location of usage will affect the design of the earth bonding reference grid. The sensitive equipment in the technology racks will need electrically clean bonding connections. Those connections will need to be sized suitably for the power draw in that area. The electrical engineers should be asking questions about these matters as part of their design process. Data Centre management should ensure the engineers take bonding requirement into account.

The power budget must also have a change control process to ensure any variations in planned/actual usage are accurately investigated, where necessary authorised and scheduled for implementation. In some cases DCIM software (Data Centre Information Management) can assist the power usage/commitment documentation process.

Power demand planning should anticipate how power requirements will develop over the lime time of the data centre. It may be appropriate to increase power provision as data centre space utilisation increases. For example power equipment which can be expanded by adding further modules may be chosen.

From an early stage, the power budget should incorporate cost recovery/allocation mechanisms. The cost recovery process should be transparent and acceptable to the client base. The process of cost recovery, may require accurate power metering down to the level of individual servers. This will affect the design and investment of the infrastructure. The budgetary responsibility may be in different organisational silos such as ICT and facilities management. Part of the provision may affect the charges levied by the building owner/landlord. Carbon Credit recovery may also affect the budgetary process.

The power budget needs to be subdivided into:
  • Critical power
  • Protected power
  • Unprotected power
Critical power is that which should be continuously available without interruption. Facilities should be provided to provide back-up power sources without interruption when the primary source fails.

Protected power is that which should be continuously available, but can sustain short duration outages  while a standby power generator starts up if the primary source fails.

Unprotected power is routinely supplied by the utility company and may be subject to supply outages of a local or regional basis. Even with unprotected power it is sensible for an organisation to increase protection by attempting to provide dual sources from different sub-stations.

Each of these separate power categories will have different capital and ongoing operational costs associated with them. It is no longer really acceptable to lump all power costs into one financial pot.

The power budget should also separately document in power in-rush when equipment is switched on. This figure can be substantially larger than the steady state power draw. While is does not normally affect the power supply to the building it will affect the design of the circuit breakers in the power circuitry and may also affect the specification of electrical back up systems such as generators and UPS battery systems.

Those people calculating the power requirements should not use the "plated" rating on the outside of the unit, but should undertake careful research with the manufacturer's specification as to steady state usage. This should be backed up with actual measurement of power usage on equipment in use in a production environment.

continues...

No comments:

Post a Comment