Re: [lustre-discuss] [EXTERNAL] Re: Fwd: RPCs in Flight are more than the max_rpcs_in_flight value

2021-10-17 Thread Md Hasanur Rashid via lustre-discuss
All my OSS's are mounted on SSD. So, I think that wouldn't be the reason behind such improvements. Thanks, Hasan On Sun, Oct 17, 2021 at 3:54 AM Andreas Dilger wrote: > It seems likely if you are using spinning disks (HDD) that using > max_rpcs_in_flight=1 will reduce the seeking on the OSTs,

Re: [lustre-discuss] [EXTERNAL] Re: Fwd: RPCs in Flight are more than the max_rpcs_in_flight value

2021-10-17 Thread Andreas Dilger via lustre-discuss
It seems likely if you are using spinning disks (HDD) that using max_rpcs_in_flight=1 will reduce the seeking on the OSTs, especially with large RPC sizes (16MB). Cheers, Andreas On Oct 16, 2021, at 16:11, Md Hasanur Rashid wrote:  Hi Everyone, Thanks for getting back to me, Andreas and

Re: [lustre-discuss] [EXTERNAL] Re: Fwd: RPCs in Flight are more than the max_rpcs_in_flight value

2021-10-16 Thread Md Hasanur Rashid via lustre-discuss
Hi Everyone, Thanks for getting back to me, Andreas and Patrick. The reason for me to experiment with max_rpcs_in_flight being 1 is that I have observed significant improvement over the default value when the workload is the following one: define fileset