Mike, So I do have to ask, where would the extra latency be coming from if all my OSDs are on the same machine that my test VM is running on? I have tried every SSD tweak in the book. The primary concerning issue I see is with Read performance of sequential IOs in the 4-8K range. I would expect those to pull from three SSD disks on a local machine atleast as fast one Native SDD test. But I don't see that, its actually slower.
On Wed, Sep 18, 2013 at 4:02 PM, Jason Villalta <[email protected]> wrote: > Thank Mike, > High hopes right ;) > > I guess we are not doing too bad compared to you numbers then. Just wish > the gap was a little closer between native and ceph per osd. > > C:\Program Files (x86)\SQLIO>sqlio -kW -t8 -s30 -o8 -fsequential -b1024 > -BH -LS > c:\TestFile.dat > sqlio v1.5.SG > using system counter for latency timings, 100000000 counts per second > 8 threads writing for 30 secs to file c:\TestFile.dat > using 1024KB sequential IOs > enabling multiple I/Os per thread with 8 outstanding > buffering set to use hardware disk cache (but not file cache) > using current size: 10240 MB for file: c:\TestFile.dat > initialization done > CUMULATIVE DATA: > throughput metrics: > IOs/sec: 180.20 > MBs/sec: 180.20 > latency metrics: > Min_Latency(ms): 39 > Avg_Latency(ms): 352 > Max_Latency(ms): 692 > histogram: > ms: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 > 24+ > %: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 > 100 > > > > On Wed, Sep 18, 2013 at 3:55 PM, Mike Lowe <[email protected]>wrote: > >> Well, in a word, yes. You really expect a network replicated storage >> system in user space to be comparable to direct attached ssd storage? For >> what it's worth, I've got a pile of regular spinning rust, this is what my >> cluster will do inside a vm with rbd writeback caching on. As you can see, >> latency is everything. >> >> dd if=/dev/zero of=1g bs=1M count=1024 >> 1024+0 records in >> 1024+0 records out >> 1073741824 bytes (1.1 GB) copied, 6.26289 s, 171 MB/s >> dd if=/dev/zero of=1g bs=1M count=1024 oflag=dsync >> 1024+0 records in >> 1024+0 records out >> 1073741824 bytes (1.1 GB) copied, 37.4144 s, 28.7 MB/s >> >> As you can see, latency is a killer. >> >> On Sep 18, 2013, at 3:23 PM, Jason Villalta <[email protected]> wrote: >> >> Any other thoughts on this thread guys. I am just crazy to want near >> native SSD performance on a small SSD cluster? >> >> >> On Wed, Sep 18, 2013 at 8:21 AM, Jason Villalta <[email protected]>wrote: >> >>> That dd give me this. >>> >>> dd if=ddbenchfile of=- bs=8K | dd if=- of=/dev/null bs=8K >>> 8192000000 bytes (8.2 GB) copied, 31.1807 s, 263 MB/s >>> >>> Which makes sense because the SSD is running as SATA 2 which should give >>> 3Gbps or ~300MBps >>> >>> I am still trying to better understand the speed difference between the >>> small block speeds seen with dd vs the same small object size with rados. >>> It is not a difference of a few MB per sec. It seems to nearly be a >>> factor of 10. I just want to know if this is a hard limit in Ceph or a >>> factor of the underlying disk speed. Meaning if I use spindles to read >>> data would the speed be the same or would the read speed be a factor of 10 >>> less than the speed of the underlying disk? >>> >>> >>> On Wed, Sep 18, 2013 at 4:27 AM, Alex Bligh <[email protected]> wrote: >>> >>>> >>>> On 17 Sep 2013, at 21:47, Jason Villalta wrote: >>>> >>>> > dd if=ddbenchfile of=/dev/null bs=8K >>>> > 8192000000 bytes (8.2 GB) copied, 19.7318 s, 415 MB/s >>>> >>>> As a general point, this benchmark may not do what you think it does, >>>> depending on the version of dd, as writes to /dev/null can be heavily >>>> optimised. >>>> >>>> Try: >>>> dd if=ddbenchfile of=- bs=8K | dd if=- of=/dev/null bs=8K >>>> >>>> -- >>>> Alex Bligh >>>> >>>> >>>> >>>> >>>> >>> >>> >>> -- >>> -- >>> *Jason Villalta* >>> Co-founder >>> <EmailLogo.png> >>> 800.799.4407x1230 | www.RubixTechnology.com<http://www.rubixtechnology.com/> >>> >> >> >> >> -- >> -- >> *Jason Villalta* >> Co-founder >> <EmailLogo.png> >> 800.799.4407x1230 | www.RubixTechnology.com<http://www.rubixtechnology.com/> >> _______________________________________________ >> ceph-users mailing list >> [email protected] >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >> >> >> > > > -- > -- > *Jason Villalta* > Co-founder > [image: Inline image 1] > 800.799.4407x1230 | www.RubixTechnology.com<http://www.rubixtechnology.com/> > -- -- *Jason Villalta* Co-founder [image: Inline image 1] 800.799.4407x1230 | www.RubixTechnology.com<http://www.rubixtechnology.com/>
<<EmailLogo.png>>
<<EmailLogo.png>>
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
