Hello All,
After an initial test on a CentOS_7 VM with 4 CPUs and 16GB of RAM I'm
testing s3ql on a physical box, a Dell r910 with 24 cores and 768GB of
RAM. Two 10GB NICs in an ALB bond0 pair.
I'm running s2ql with its cache on a 300GB RAMDISK and compression set
at zlib=1 , cachesize at 320GB, against a SwiftStack cluster.
I'm loading it with multiple rsync streams coming via NFS (v3, auth_sys)
from multiple cluster nodes.
I'm watching metrics with htop, iftop, and a running "watch" of
s3qlstat, df and free.
When the streams start, I'm getting incoming throughput to the machine
with bursts up to 190MB/s and an average of 120MB. CPU cores are all
working but none pegged or stuck, overall load average is 6-8. Memory
doesn't seem to be a factor, without the RAMdisk we're only using ~5G.
s3ql stat shows the cache at ~35GB. "Dirty" cache starts at zero and
climbs steadily to the full size of the cache, and when they match
(cache is all dirty), input slows to half (60MB). Cache size and dirty
cache slowly drop together along with throughput, currently at 26GB and
38MB/s.
What kind of performance can I hope for with s3ql, and what do I need to
know to tune for it?
What other information should I be providing?
Let me know,
Randy
--
You received this message because you are subscribed to the Google Groups
"s3ql" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
