Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Mute
- Printer Friendly Page
sequential read/write, and random read/write performance on SSD?
2019-08-28
08:06 AM
5,184 Views
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
In SSD world, does that make any differences in distinguashing the performance?
sequential read; sequential write; random read; random wirte
I would think they should be all the same.
Thanks for your inputs
3 REPLIES 3
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
The blocks of an SSD presented to the attached system are really virtualised cells of the underlaying memory modules. SSDs contain something called a flash translation layer (FTL), which does wear leveling, device layer encryption and garbage collection and moves data from the virtual blocks to the physical cells.
By their very nature they adapt best to sequential write/read workloads.
FTLs can have different optimisations depending on workload, which have tradeoffs against device lifespan and usable capacity vs native capacity, this was more common with smaller capacity devices, but most are general purpose now. Inside NetApp systems, we perform additional data stream analysis to best match the streams of data to and from clients to the way our SSDs work.
Hope this helps!
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thanks for infor. They are helpful.
So, can I say the performane of all these 4 different type of I/O's are all about the similar?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
They are not similar, no
have a look at this review for a SSD and the difference between different profiles - https://www.storagereview.com/samsung_860_evo_ssd_review
With regards to NetApp systems, spreading the workload over more containers does reduce the impact of random IO somewhat, but it is still a consideration to work through with your partner SE in designing a system
