July 24, 2009, 9:32 AM — Q) What is an SSD and how does it work?
A) A solid-state drive (SSD) functions like a hard disk drive (HDD), except that it uses flash memory instead of spinning disk to store data. With no moving parts, SSDs are inherently more durable and consume less power than traditional HDDs. As there are no mechanical delays, they enjoy significantly lower access time and very little latency. In fact, SSDs offer performance improvements of 280X over SATA drives and 100X over Fibre Channel (FC) drives. Current-generation SSDs have no read cycle limits and a write life cycle of at least 140 years with manufacturers suggesting a 10 year life expectancy. Thus, when deployed properly SSDs will outlast the useful life of a typical storage system.
Q) How do SSDs improve existing storage solutions?
A) Many types of application software (transaction processing, batch processing, query or decision support analysis) can benefit from the super-fast access times SSDs offer. Adding a small amount of SSDs to an environment can result in a significant performance improvement and eliminate a major performance bottleneck for an application, while dropping power requirements significantly.
When deployed in a storage system, SSD drive shelves can deliver performance gains of more than 16X that of SATA drive shelves and 5X of FC shelves.
On the power savings front, if a single brick of SSD is deployed instead of four bricks of 15K FC drives, users can save up to 1045 Watts per hour. The four FC bricks, under load, use 1232 Watts per hour and the single SSD brick uses only 186.6 Watts per hour, resulting in a difference of 1045.4 Watts.
The performance and power savings advantages are undeniable.
Q) Is it possible for storage vendors to guarantee the performance of an SSD?
A) For some vendors, absolutely. By placing only high-priority data on SSDs, it guarantees that only those mission-critical applications will use the SSD capacity. Layering Quality of Service (QoS) on top of the shared disk pool will ensure that low-priority data will not impact the performance of I/O’s targeted for SSDs. QoS provides guaranteed resources for applications and does not allow lower priority applications to steal resources and performance from higher priority ones. Vendors that already have utilization and performance guarantees in place are better able to extend current guarantees to include SSDs.
Q) What are the disadvantages of SSDs?
A) Traditional storage architectures still suffer from their first-in first-out models. This means that SSD I/O’s will become trapped behind slower SATA and FC I/O’s. Pillar’s architecture, which was designed around QoS to support technologies like SSDs, limits the bottlenecks whether it’s SSD, FC, or SATA. All other storage systems on the market today treat these very different drive types equally, thus negating in large part the benefits of SSDs.
By delivering multiple QoS levels within a single system, a vendor can ensure that only the most mission-critical data is placed on SSDs, while lower tiers of data are stored on traditional drive types. This is the most cost-effective and energy-efficient approach to SSD implementation.
Q) What’s the best implementation route for SSDs?
A) There are two implementations available on the market today: adding SSDs as additional cache or using the SSD as actual storage capacity. Both implementations make sense, depending on the storage architecture they’re being deployed into. The real key is ensuring that only the most critical applications and data are accessing SSDs, while all other data is being placed on traditional “spinning disks”. There are very few storage architectures available today that offer this level of functionality, or storage QoS.
Q) How do you block and prioritize applications in the capacity model?
A) Today there are few tools that easily enable, much less identify and automate, the process of ensuring that the most performance-sensitive or mission-critical data is placed in the new Tier 0.
This problem can be solved with Application-Aware technology. It allows an administrator to easily define the application type and its associated value to the business through a simple drop-down menu. The system then automatically tunes itself to deliver differentiated storage service levels based on the business priority of each application. Those storage service levels include determining which drive type (SSD, FC or SATA) to write the data.
Q) Aren’t SSDs cost-prohibitive for most customers?
A) Not if they’re deployed and accessed properly. SSDs can drive down overall power consumption by up to 85 percent.
Additionally, customers tend to buy disk drives to meet their I/O needs. Thus, they end up with far more capacity than they need. The industry average of this kind of “wasted storage” is around 75 percent. This means that while a customer pays the full price for storage capacity, they can use only ¼ of what they purchased – a hugely inefficient practice. With SSDs they can buy far fewer drives to meet their high-I/O requirements. For example, the data in a database doesn’t need the performance of SSDs but the redo logs and rollback segments and other key, high I/O pieces of the database do need that performance. SSDs can dramatically improve the performance of the entire database. In fact, Pillar will drive down the $/IOP of SSDs to half that of traditional FC drives. Again, it’s all about the storage architecture SSDs are placed into, and how well that system can differentiate storage services based on business priority.
Q) What’s next for Pillar in this area?
A) With our initial implementation we support SSDs with SATA. Users will see drastic performance increases and cost savings with this offering. Moving forward, Pillar will support FC, SATA and SSD in the same system along with a new QoS level and Application-Aware profiles to further exploit the performance of SSDs. Ongoing, Pillar will be evaluating new SSD technologies such as faster performance and larger capacities, as they come to market. Pillar will also upgrade Slammer and Brick components as they come to market to increase IOPS and bandwidth to the SSD bricks. The company is also looking at placing SSDs into the cache pool, and eventually providing different QoS levels of SSDs – Single Level Cell (SLC) and Multi Level Cell (MLC) technologies. Pillar has a unique advantage with this technology in that it is the only company today that can deliver true storage QoS, thereby taking full advantage of the performance gains and power savings of SSDs.
Bob Maness is Vice President of Worldwide Marketing and Channel Sales for Pillar Data Systems