Planning for SAS use on solid state drives

4

SAS software benefits from the high performance of solid state drives (SSD), whether it is in the form of flash memory cards or hard-drive form-factor SSDs. The following general points should be considered when planning SSD storage for SAS usage:

• Whether in flash card or hard-drive form-factor, solid state storage is faster than traditional disk devices. Even the poorest performance we tested for SSDs (large sequential writes), solid state storage is still slightly faster than traditional storage. On the other hand, SSD is much more costly.

• Solid state storage typically performs best for random reads. The portions of SAS applications that have a lot of random components (OLAP, heavy index operations, POINT= traversals, etc.) will benefit the most.

• Random writes and sequential reads also perform far better on SSD but generally to a lesser degree than random reads. They are still much faster for typical SAS operations.

• Sequential writes on SSD storage typically exhibit a lesser degree of performance gain over traditional disk devices (in some cases, only very slightly). Garbage collection activities associated with cell-level write operations on populated disks can significantly dampen performance. The manufacturers of newer drive models may promise certain levels of improvement in speed. However, because SAS invokes so many large sequential writes, you may not realize the promised level of improvement when using SAS on these drives.

• Striping across flash cards or solid-state disks with a RAID5 stripe aggregates device performance well.

• The co-location of flash cards in system board slots and the physical distance between them affects performance.

Not all brands or device models of solid state storage are created equally, nor perform equally. Do your homework and research--sequential performance can significantly vary across models.

Many thanks to Tony Brown from SAS Performance Validation for today's content.

Margaret

Share

About Author

Margaret Crevar

Manager, SAS R&D Performance Lab

Margaret Crevar has worked at SAS since May 1982. She has held a variety of positions since then, working in sales, marketing and now research and development. In her current role, Crevar manages the SAS Performance Lab in R&D. This lab has two roles: testing future SAS releases while they're still in development to make sure they're performing as expected; and helping SAS customers who are experiencing performance issues overcome their challenges.

4 Comments

  1. What are the dangers of long term SAS processing on SSD drives? We have lots of staff running SAS on laptops with 4Gig of RAM and 128-256gig of SSD disk space. The amount of file system writes has me concerned that we are going to kill SSD's en mass. Do you folks have any thoughts on that?

    regards

    Frank

    • With the original SSD drives on the market, this was a very important concern. We told SAS users to monitor the IO throughput on their new SSD drives and when they started seeing a significant performance degradation to remove all data from the SSD drive and reformat it and the copy all the data back. Because of this, SAS took awhile to fully endorse the SSD drives.

      However, with today's SSD drives, we do not see that performance degradation with writes like we did 1-2 years ago.

      Hope this helps. Please let me know if you have additional questions or concerns.

  2. Margaret Crevar on

    We are so glad you enjoyed the article. Please note we have lots of other articles available regarding how to configure hardware (both computers and storage arrays) and operating systems to work best with SAS. You can find these articles on the SAS support web site at http://support.sas.com/kb/42/197.html

    With regard to suggestions on RAID SSD vs single SSD vs RAID HDD, that choice is best left to your storage array vendor. From a SAS software perspective, just be sure to choose a configuration that gives you the I/O throughput required to meet user SLAs for SAS applications.

    Margaret

  3. This is a great article. I was initially surprised that no one has commented on it, but I suppose most people just use the computer they are given and have no say in hardware. I think the discussions on hardware performance under SAS is an under-served topic, especially for client workstations (it's a typically a topic discussed when running servers).

    I'd like to see more articles like this one. For example, some suggestions on using a single SSD v.s. RAID SSD v.s. RAID HDD.

Leave A Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Back to Top