Solid state disk (SSD) has a significant upfront cost that makes most IT professionals hesitate on considering the technology. Depending on the SSD type, it can be 10 to 15 times more expensive than the equivalent hard drive technology when considered on a cost per GB basis. Those in the solid state community are quick to point to SSD's I/O-per-second cost advantage, but evaluating the true cost of adding SSD to your environment is more complex than that.
Recently, as solid state prices have continued to decrease and storage system builders have become more sophisticated in their memory-based storage implementation, the technology's potential has become significantly wider. Also, as we will discuss on an upcoming InformationWeek webinar, flash-controller manufacturers are adding more intelligence to their controller technology to make less-expensive flash technologies like multi-level cell more practical.
Even with all this advancement, there are still costs associated with solid-state memory beyond just the cost of the system. Understanding these costs may impact how you implement flash in your environment or which vendor you select when looking for storage systems that include flash.
The Cost Of Placement
There is a cost associated to knowing which data should be placed on the SSD tier. Essentially there are three ways to get data to the SSD tier: static, automated, or flash-only placement.
Static placement means understanding which data needs to be accelerated, and manually placing that data on the solid state device. There is a cost associated with analyzing the data in your environment to know which data qualifies, and updating that analysis so that only the right data is on that static solid state tier at all times. There is also a cost associated with integrating this separate tier and potentially separate system (in the case of a solid state appliance) into your data protection and disaster recovery scheme. We discuss this in our article Integrating SSD and Maintaining Disaster Recovery.
Automated data placement, via either caching or automated tiers, removes much of the manual cost of data placement, but replaces it with a cost of resources, as well as the risk of inaccuracies--meaning that the data could be on hard disk instead of solid state when you need it. Resources are needed to automate the analysis of data and to provide sufficient bandwidth to move data back and forth between hard disk and solid state. This high back-and-forth copy also requires many vendors to use a higher endurance flash media like single-level cell.
Flash-only systems avoid all analysis, both manual and automated. Everything is placed on solid state. These systems have the potential downside of no hard drive tier to reduce costs. To offset the cost of a solid state-only environment, most of these systems can use lower cost flash media because they don't have the copy-back-and-forth problem that the automated solutions do.
These systems also leverage deduplication and compression to further wring out the costs associated with being flash only. The space optimization technologies work surprisingly well in these environments, as we explained in a recent Chalk Talk, but they do add a burden to the system that needs to be compensated for.
Cost of placement is just one consideration in calculating the true cost of SSD. There is also the cost savings that SSD technologies can bring to the enterprise. That will be the subject of our next entry.
Follow Storage Switzerland on Twitter