This post is more than 5 years old
148 Posts
0
3181
Difference between Striped and Concatenated Thin Meta
Hi All,
Good morning.
I have question here related to thin meta devices.As per my understanding ,write to the thin devices will striped across the data device.So what is the basic difference between Striped and concatenated thin meta devices ?
Quincy561
1.3K Posts
0
May 3rd, 2012 05:00
Yes, both are striped across the backend.
However when performing random IO to a striped meta volume, you are very likely to have more than one volume with active IO, and with a concat meta only one volume is likely to be active. The difference in performance comes from the volume being able to do IO to and from cache, and has nothing to do with the backend at all.
anoopcr
148 Posts
0
May 3rd, 2012 05:00
Thanks Quincy.
So considering the performance factor, Striped thin meta will give more performance than concatenated one. Because each metamember gets a small percentage of cache and striped meta is likely to offer more writeable cache to the metavolume
Is that correct ?
Quincy561
1.3K Posts
0
May 3rd, 2012 05:00
Again, this really has nothing to do with the backend at all. We are not recommending striped metas on VP because of the backend like thick volumes.
Quincy561
1.3K Posts
0
May 3rd, 2012 05:00
For example, say you have a 2TB volume with 4 members of 240GB each.
If it is concat, the LBA ranges will be sequential, meaning if IO is occurring in the first 240GB of space, they will all land on the same volume.
If striped, every 960k the volume will switch, making it much more likely to have multiple volumes active.
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
May 3rd, 2012 05:00
i get this question from time to time as well, do you have a pictorial representation of how one works vs the other ?
Thanks
Quincy561
1.3K Posts
1
May 3rd, 2012 06:00
It really has nothing to do with the amount of cache, one volume can have 100% of cache in use for reads.
Yes, each volume has a WP limit, but in VMAX that is already quite large, 5% of total writable cache.
It has to do with the concurrency a single volume has. More volumes, more concurrency.