Applications and Workloads for Solid State Disks

1. Clustered File Systems: Example of a clustered file system is SAM-QFS:
One prominent application, which SpeedStor dramatically enhances, is clustered file systems such as Sun’s SAM-QFS file system. We recommend placing the metadata onto the SpeedStor. By placing the metadata onto the SpeedStor your customer will witness dramatic improvement in overall responsiveness. Specifically the customer will benefit from reduced time to perform SAMFS dumps by 7X or more (see Sun benchmarks and success stories).

The SAM-QFS file system normally provides extraordinarily higher data management performance largely due to an ability to separate metadata from the file content onto devices tuned more closely for the types of IO associated with metadata processing. In a traditional SAM-QFS configuration, metadata will be placed on devices optimized for small block random IO protected with a RAID-1 type disk mirroring, while that actual data content will reside on some variant of RAID-5 devices optimized to the data size and access patterns.

While this approach allows scaling a file system to be able to manage 100s of millions of files and 100s of simultaneous users, there still will reach a point where file system performance is limited by the ability to access the metadata. Although smaller faster spinning disks can maximize performance, even this technology for many systems/users is no longer sufficient. To increase the performance of these huge established environments as well in designing new systems today to handle billions of files, we are offering a technical “step up” for metadata processing.

The use of SSD and the zero-latency nature of these devices allows quantum level increases in the levels that metadata can be accessed. Translated for a SAM-QFS file system, performance for creating, updating, deleting, and scanning/accessing files increases to levels otherwise unattainable with spinning disks. For a customer with a SAM-QFS file system that has reached its performance limit, the replacement of disk device(s) with SSD extends significantly the life of that same file system architecture. This allows for additional time and redesign efforts on the existing architecture to move forward. But for the nominal additional costs of a SpeedStor, the system performance may be regained to allow the file system to continue in its current configuration an additional year or more.

Within the SSD realm, there are primarily two types of memory technology used to create these devices; DRAM and NAND. The key characteristics/differences as they apply to solving customer metadata performance limitations comes down to balancing cost-performance. Most customers will dramatically benefit from the performance improvement of the SpeedStor and its NAND technology. While not as high performance values as can be achieved via DRAM based systems, the cost-performance point will be far more attractive to many more existing (and new) SAM-QFS customers. While there will be those customers that have system-level requirements justifying the costs of DRAM-based systems (and they should be addressed as well!), our target is the other 90% of SAM-QFS customers who can be shown the cost-performance benefit of a SpeedStor will be within their budget. The discussions should be particularly around the impact on TCO when not increasing the performance of metadata.

2. VDI-VirtualStorm: has all applications packaged and placed on a centralized disk, VirtualStorm then allows the golden image to be accessed simultaneously by multiple users (essentially created a distributed disk based file system through multiple VMs). This means that thousands of users will by accessing many applications, simultaneously. Concurrent random reads is the “sweet spot” for the SpeedStor drive. The drive is designed to have multiple IO paths accessing and managing the IO in parallel.

3. File Server, Exchange Server, Mail Server, Voice Mail Server, News Server, Web Server These applications are predominantly single block random reads and writes. This data is generally stored on a local disk as it is too risky to place in volatile host based memory. These applications generate random access from many concurrent sources. They include file-orientated servers like mail and web servers where lots of people might attempt to access their data at the same time. The amount of data required is normally quite small but owing to the large number of users the data is scattered all over the storage devices. With conventional disk drives the time involved waiting for the head actuator mechanism to position the head and waiting for the rotational delays results in a tremendous amount of time consuming mechanical delay. This delay is eliminated with the SpeedStor.

Sun compared the response time of the 6140 to a single SpeedStor drive for NFS file serving. The SpeedStor is up to a factor of five times faster and this is only with one file system. Imagine that the cache of the 6140 was still solely available for metadata. If the cache were reduced the difference would be 10X!


4. Trading, Banking, Purchasing, Online transactions In these areas data is typically held in relational databases, which serve data to many distributed users. As well as decreasing access times to the data table portion of the database solid state disks can make a vast difference to the access times to the database index and log files. All too often the best solution to these problems is perceived to be to add another server and some more disk drives, even when the actual amount of data being held is quite small. Adding SpeedStor Solid State device to any spare slot where a 3.5 inch (1inch high) conventional disk drive could be fitted will make the disk bottleneck disappear. No extra space, power or software licenses are required. This is a much more cost-effective solution than buying and maintaining additional servers and storage.

5. Securities Fill the performance gap between CPU and disk and you can:

6. Search engine cache, Swap files Solid State disks can also be deployed to speed up systems, which maintain swap files or search engine caches. In these cases they act as a very large data cache that does not have to be reloaded when the system is restarted.

7.Video Streaming IO. In order to take advantage of the 0- 2 m.s. latency of the SpeedStor, it is important to stream the IO in small (64k or less) multiple paths (not sequential)…if the software is capable of doing so (i.e. SAM-QFS) then the SpeedStor will offer blazing performance.