How to purchase and manage your data storage capacity

If effective data storage capacity management is a difficulty in your organisation, you're not alone. In this SearchStorage.co.UK interview, GlassHouse Technologies (UK) principal consultant Steve Pinder discusses why users buy more data storage than they need, as well as how they can manage data growth and cut data storage costs with effective capacity planning processes.

You can listen to Steve's answers as an MP3 or read a transcript of them below.

What are the key reasons customers buy more data storage capacity than they should?

There are many reasons people buy more data storage capacity than they should. One of the main reasons is what I call the 'incremental buffer zone,' where everybody adds a buffer to their requirements just in case they need it.

For example, a department wants to build a new application and think they'll need 50 GB for the data. They'll speak to a DBA [database administrator] about their capacity requirements and actually say they need 100 GB. The DBA agrees with them but just to be on the safe side, he makes it 200 GB. The DBA asks the storage pro for 200 GB but he allocates 300 GB just in case it's needed for growth in the next year. And so an allocation of 300 GB is made for what is only a 50GB requirement.

Another example is not allocating storage from hosts that don't need it anymore. This causes so-called 'orphaned storage.' The details of the storage owner become lost over time and admin staff becomes scared to remove the allocation in case it's still required. After all, no one has been fired for adding storage, but plenty have been fired for taking it away.

Also, sometimes storage is available but is not in the correct place. For example, a server on a storage-area network (SAN) 5 requires some additional capacity. Unfortunately, all of the arrays on SAN 5 are full, but an array on SAN 6 has plenty of capacity left. In this case, additional storage will have to be purchased for SAN 5 as the server cannot access the available storage on SAN 6.

Are there best practices for efficiently managing storage capacity and growth?

There are many best practices in managing storage growth. Storage pros should ensure that the correct tier of storage is allocated to a host. In many firms there's no effective means of identifying classes of service that hosts should expect from their storage infrastructure. All application owners therefore gravitate to the most resilient storage available as they don't understand what service level they can expect from IT. If levels of service were clearly defined, business units could allocate storage according to the application requirement.

A realistic chargeback model must be implemented for storage services. Unless business units are charged a hefty uplift for premium services such as snapshots and replication, they will want these for all their applications. If the cost truly represents the service provided, application owners will only put critical applications on tier 1 storage because they're being charged extra for it.

Finally, a major reason why application owners ask for more storage than they need is because they are continually being rebuffed by IT when they ask for more. If they knew they would be given additional capacity a few days after they requested it and they were also charged for it when they used it, they would be more likely to only ask for what they actually need.

What data storage technologies can help customers cut down on storage capacity spend?

Let's take a look four storage technologies that can help users cut down on capacity growth.

Firstly, thin provisioning can reduce wastage in overallocation. This technology allows storage to be presented to more than one host with hosts only using capacity when it is written to physical disks. With thin provisioning, care should be taken to always ensure enough buffer space is available so that capacity does not run out. If capacity does run out, many applications will hang or stop immediately.

Secondly, storage pros should use RAID 5 or RAID 6 instead of RAID 10. These levels have lower performance than RAID 10 but will give much more capacity per physical drive.

Thirdly, we can use cheaper storage technologies than Fibre Channel (FC). If applications do not need Fibre Channel performance and seek times, why not use SATA instead?

Lastly – and this is not a storage technology, but more of a process – customers should ensure they have a robust capacity planning model. If high demand is forecast for a particular time period, it may make sense to purchase shared storage.

This can also allow negotiation with a vendor to increase the discount that's available or possibly get vendors into a bidding war for your business, lowering the cost of your storage spend.

This was first published in December 2009


COMMENTS powered by Disqus  //  Commenting policy