The new generation of data centres might just be like bacteria, splitting and doubling in number at frequent intervals, to create a host of smaller individual data centres instead of just one massive one. At least one data centre provider sees cloud computing as limited in terms of latency, making the “big data centre in the sky” with its 100-millisecond or more latency less suitable for certain applications, like real time control of machines. Other advantages include local data storage without the need to send high volumes of data over the net. But what about disaster recovery for this new plethora of data centres?
Suitable disaster recovery will depend on several factors:
- The importance of the data being gathered, processed, and stored. In theory, if a local operation is important enough to warrant having its own micro data centre it is important enough to require good DR planning.
- The operational or functional department looking after the micro data centre. Just as cloud services are available to departments other than the IT department, micro datacentres with their “out of the box” availability may appeal to operations, production, field service or other non-IT-specific groups. But will these groups recognise the need and the importance of DR?
- The DR objectives. Data backups, for example, will have to be stored elsewhere than in the corresponding micro data centre, to avoid the risk of a disaster wiping out all resources stored locally. However, this means data is likely to be transferred for safekeeping over the Web, whereas web latency of communications and transfers was a reason for choosing micro data centres in the first place.
Latency issues aside, an efficient disaster recovery strategy may involve micro datacentres storing each other’s data, and “parachuting in” replacement data centres (at about $50,000 per data centre, say) if a catastrophe causes significant hardware damage. The bigger challenge may still be to convince all concerned of the need to plan and manage DR in the first place.