Stripe width in the contexct of write operation

[baber]

1- I read about stripe width and it mentioned that we can guarantee that stripe can increase the performance when there is a large amount of data to destage from cache tier to capacity tier on the other means cache tier capacity has used more than 70%. now would you please explain how stripe can help to increase performance here ? Would you please give an example ?

2- Is that means as we want to write and write data in the first step (Hybriad and AFA) will place on cache tier /write buffer since the write buffer is full so write block will be spread on the other cache ?

[baber]

Can anyone help me about it ?

[depping]

I don't understand the question. but when you increase the stripe size it is very likely that your data will be spread across more diskgroups, which means that you will have access to more caching and capacity devices, which means that potentially you could see a better performance.

[baber]

According to your book "vSAN deep dive" page 71  from write operaton perspective  "The only
occasion where an increased stripe width could definitely add value is when there is a large amount of data to destage from the cache tier to the capacity
tier. In this case, having a stripe could improve destage performance as multiple capacity devices would be available to destage data to."

I just wanted the reason for it . why increase stripe width will improve the vm performance from write perspective while the size of cache tier/write buffer is about 30% ?

[TheBobkin]

@baber "I just wanted the reason for it . why increase stripe width will improve the vm performance from write perspective"

Because writing to 2 or more Cache-tier devices at the same time is faster than just writing to 1.

[baber]

That is correct but according to article , it is saying from write operation perspective, we can guarantee that increase stripe width while most of the space on cache tier /write buffer is used can help to improve vm performance but imagine we have 7 hosts that there is one disk group on each of them . My means is that why it is saying stripe width can increase performance while most space of cache tier / write buffer used ?

[TheBobkin]

So, let's say all stripes are on the same Disk-Group (but on different Capacity-tier devices as this is what SW=2/3/4 does from a placement perspective), if the Cache-tier write-buffer is filling to the point that it is actively destaging data to the Capacity-tier devices then destaging to 2 or more Capacity-tier devices is always going to be faster than a single Capacity-tier device.

A simple analogy: I have wide pipe that can drain 10L of water a minute, but it is flowing into a thinner pipe that can only take 5L of water a minute - this is going to be a bottleneck - but if I have 2x5L pipes accepting the flow then nothing is left waiting.

[baber]

So thanks. I understand your means but cannot clearly understand this part "when there is a large amount of data to destage from the cache tier to the capacity tier. In this case, having a stripe could improve destage performance as multiple capacity devices would be available to destage data to."  now imagine we can see write buffer free percentage is about 78%

1- How can understand when a large amount of data want to destage to capacity tier ?

[IRIX201110141]

Are you referring to "Number of disk stripes per object" within a vsan storage policy?

[baber]

َAs I understood when there is a large amount of data to destage from cache tier to the capacity tier. In this case, having a stripe could improve destage performance as multiple capacity devices would be available to destage data to.

how can understand there is a large amount of data to destage ? Can I understand from write buffer free percentage ? what percentage shows there are a large amount of data to destage ? 20% , 30 ? 50 ? 70 ? 80 ? 

[baber]

@TheBobkin  "I just want to know how can understand there are a lot of block data for destage ? I just know we have to see write buffer free percentage chart but what is percentage ? 10 , 20, 30, 40, 50 , 60 , 70 , ...... ?

[baber]

I just want to know what number in write buffer free percentage chart shows there are a lot of data for destage ?

[IRIX201110141]

https://blogs.vmware.com/virtualblocks/2019/10/01/write-buffer-sizing-vsan/