SSD’s put rotational drives in a spin
By Brian Cowan
Corporate IT has been taking advantage of hybrid storage arrays for several years now, using SSD for performance and rotational drives for high capacity. The time may be coming for a rethink on this strategy.
SSD limitations and benefits
The demand to store and retrieve data continues to grow. An estimated 80% of stored information remains static and so it made sense to keep it on high capacity, relatively low cost traditional spinning disks. SSD drives have been limited on capacity and for the budget conscious customer, relatively expensive to deploy. In the early days there was also the issue of reliability compared to rotational media.
The introduction of SSD technology has taken I/O performance off the design table as a concern for many storage projects. Having spent too many hours working through performance analytics and drive metrics to build solutions that met the current performance demand and the xx% growth due to systems as yet unidentified…this was a blessing! Storage tiering has allowed the SSD’s to remain optimised for I/O whilst older, less active data heads south the the slower disks.
The SSD pipeline
Among the SSD headlines announced this week was the Toshiba development of 12Gbps, SAS connected, 64-layer, 3D Flash memory with a capacity scaling up to 30.72 TB.
Also announced this week, the Samsung 128TB SSD as described in article in yesterdays The Register.
Meanwhile, Smart Modular Technologies have their Osmium range of MLC SSD drives that offer 25TB to 50TB capacity.
The cost arguement
The SSD/Rotational drive price differential was narrowing but recent raw material shortages have slowed or in some cases reversed this trend. However, the IT industry will make SSD happen over the coming years. There are other factors to consider also:
Whilst the Capex for a large rotational storage solution may appear attractive today, as the price differential reduces, the energy cost savings that SSD offers will become more of a factor. If an SSD is likened to pushing a car along a flat road, starting up a spinning disk and moving the disk heads into position equates to having to push the car up a hill to get to the flat road for every new disk operation! There are numerous comparisons of the two on the web and a factor of 5 is common for the increase in power usage with rotational disks.
Anybody who has worked in a datacentre where there are large traditional storage arrays will vouch for the noise that these and their cooling fans make (if they can hear the question). Reducing the power demand also reduces the need for array cooling and air conditioning, all of which contributes to a quieter workplace.
Building a storage strategy for SSD
For datacentre requirements that have to be addressed in the short term ( 0-6 months), the hybrid approach will probably be the way to go. However, it is a good idea to develop a strategy that defines the longer term demand and delivery of storage. As mentioned earlier, capacities of SSD drives are set to expand dramatically and the benefits that using this can bring are worth considering, subject of course to the cost of acquiring the arrays. There are other factors that should also be considered including:
Software defined storage is gaining market share with major players such as VMware (vSAN), Microsoft (Storage Spaces) and Nutanix achieving success. Whilst not a pre-requisite, having a layer of SSD storage in each server provides the same performance guarantees as in a traditional storage array. Sizing of the SSD layer is very important and guidelines are available for each product.
Hyper-convergence will remove the need for storage to be provisioned as a separate entity within the datacentre. Rather than wheeling in the fork lift truck as can be the case when a traditional storage array runs low on capacity or performance, more of both can be added with an additional server in the virtualised cluster.
Backup and Business Continuity
Most modern backup solutions offer a “rapid recovery” of a server, should the primary become unusable. However, what has often been missed is the performance needed from the backup server to support the vm’s whilst recovery is in progress. During this time, the backup server can be supporting the full user population, running a database application and also copying the failed vm back to the primary system. For a mission critical application is may be prudent to have dedicated backup and recovery resource, capable of providing the user performance required.
For more information on how Keytech can assist with your strategy please click here, or call us on 01942-311150.