Re: [Gluster-users] GlusterFS 3.0.2 small file read performance benchmark
Ed, oplocks are implemented by SAMBA and it would not be a part of GlusterFS per se till we implement a native SAMBA translator ( something that would replace the SAMBA server itself with a thin SAMBA kind of a layer on top of GlusterFS itself ). We are doing that for NFS by building an NFS translator. At some point, it would be interesting to explore, clustered SAMBA using ctdb, where two GlusterFS clients can export the same volume. ctdb itself seems to be coming up well now. Regards, Tejas. - Original Message - From: Ed W li...@wildgooses.com To: Gluster Users gluster-users@gluster.org Sent: Wednesday, March 3, 2010 12:10:47 AM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi Subject: Re: [Gluster-users] GlusterFS 3.0.2 small file readperformance benchmark On 01/03/2010 20:44, Ed W wrote: I believe samba (and probably others) use a two way lock escalation facility to mitigate a similar problem. So you can read-lock or phrased differently, express your interest in caching some files/metadata and then if someone changes what you are watching the lock break is pushed to you to invalidate your cache. Seems NFS v4 implements something similar via delegations (not believed implemented in linux NFSv4 though...) In samba the equivalent are called op locks I guess this would be a great project for someone interested to work on - op-lock translator for gluster Ed W ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
Re: [Gluster-users] GlusterFS 3.0.2 small file read performance benchmark
Well, oplocks are an SMB definition, but the basic concept of opportunistic locking is independent of the filesystem. For example it appears that oplocks now appear in the NFS v4 standard under the name delegations (I would assume some variation of oplocks also exists in GFS and OCFS, but I'm not familiar with them) The basic concept would potentially provide a huge performance boost for glusterfs because it allows cache coherent writeback caching. In fact lets cut to the chase - what we desire is cache coherent writeback caching, ie reads to one server can be served from local client cache, but if the file is changed elsewhere then instantly our cache here is invalidated, and likewise we can write at will to a local copy of the file and allow it to get out of sync with the other servers, but as soon as some other server tries to read/write to our file then we must be notified and flush our cache (and request alternative locks or fall back to sync reads/writes) How do we do this? Well in NFS v3 and before and I believe in Glusterfs there is implemented only a cache and hope option, which caches data for a second or so and hopes the file doesn't change under us. The improved algorithm is opportunistic locking where the client indicates to the server the desire to work with some data locally and get it out of sync with the server - the server then tracks that reservation and if some other client wants to access the data it pushes a lock break to the original client and informs it that it needs to fsync and run without the oplock I believe that an oplock service this could be implemented via a new translator which works in conjunction with the read and writeback caching. Effectively it would be a two way lock manager, but it's job is somewhat simpler in that all it needs do is vary the existing caches on a per file basis. So for example if we read some attributes for some files then at present they are blindly cached for X ms and then dropped, but our oplock translator will instead allow the attributes to be cached indefinitely until we get a push notification from the server side that our cache must be invalidated. Same also with writes - we can use writeback cache as long as no one else has tried to read or write to our file, but as soon as someone else touches it we need to fsync and run without cache I have had a very quick glance at the current locks module and it's quite a bit more complex than I might have guessed... I had wondered if it might not be possible to make the locks module talk to the cache module and add server side lock breaking through that module? Essentially it's the addition of the push lock breaking which helps, so if we are reading away and some other client modifies a file then we need a feedback loop to invalide our read cache Perhaps this is all implemented in glusterfs already though and I'm just missing the point... Cheers Ed W On 02/03/2010 18:52, Tejas N. Bhise wrote: Ed, oplocks are implemented by SAMBA and it would not be a part of GlusterFS per se till we implement a native SAMBA translator ( something that would replace the SAMBA server itself with a thin SAMBA kind of a layer on top of GlusterFS itself ). We are doing that for NFS by building an NFS translator. At some point, it would be interesting to explore, clustered SAMBA using ctdb, where two GlusterFS clients can export the same volume. ctdb itself seems to be coming up well now. Regards, Tejas. - Original Message - From: Ed Wli...@wildgooses.com To: Gluster Usersgluster-users@gluster.org Sent: Wednesday, March 3, 2010 12:10:47 AM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi Subject: Re: [Gluster-users] GlusterFS 3.0.2 small file readperformance benchmark On 01/03/2010 20:44, Ed W wrote: I believe samba (and probably others) use a two way lock escalation facility to mitigate a similar problem. So you can read-lock or phrased differently, express your interest in caching some files/metadata and then if someone changes what you are watching the lock break is pushed to you to invalidate your cache. Seems NFS v4 implements something similar via delegations (not believed implemented in linux NFSv4 though...) In samba the equivalent are called op locks I guess this would be a great project for someone interested to work on - op-lock translator for gluster Ed W ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
Re: [Gluster-users] GlusterFS 3.0.2 small file read performance benchmark
On 27/02/2010 18:56, John Feuerstein wrote: It would be really great if all of this could be cached within io-cache, only falling back to a namespace query (and probably locking) if something wants to write to the file, or if the result is longer than cache-timeout seconds in the cache. So even if the file has been renamed, is unlinked, has changed permissions / metadata - simply take the version of the io-cache until it's invalidated. At least that is what I would expect the io-cache to do. This will introduce a discrepancy between the cached file version and the real version in the global namespace, but isn't that what one would expect from caching...? I believe samba (and probably others) use a two way lock escalation facility to mitigate a similar problem. So you can read-lock or phrased differently, express your interest in caching some files/metadata and then if someone changes what you are watching the lock break is pushed to you to invalidate your cache. It seems like something similar would be a candidate for implementation with the gluster native clients? You still have performance issues with random reads because when you try to open some file and you still need to check it's not open/locked/needs replicating from some other brick. However, what you can do is have proactive caching with an active notification of any cache invalidation and this benefits the situation where you re-read stuff you already read, and/or you have an effective read-ahead which is grabbing stuff for you Interesting problem Ed W ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
[Gluster-users] GlusterFS 3.0.2 small file read performance benchmark
Greetings, in contrast to some performance tips regarding small file *read* performance, I want to share these results. The test is rather simple but yields some very remarkable results: 400% improved read performance by simply dropping some of the so called performance translators! Please note that this test resembles a simplified version of our workload, which is more or less sequential, read-only small file serving with an average of 100 concurrent clients. (We use GlusterFS as a flat-file backend to a cluster of webservers, which is hit only after missing some caches in a more sophisticated caching infrastructure on top of it) The test-setup is a 3 node AFR cluster, with server+client on each one, single process model (one volfile, the local volume is attached to within the same process to save overhead), connected via 1 Gbit Ethernet. This way each node can continue to operate on it's own, even if the whole internal network for GlusterFS is down. We used commodity hardware for the test. Each node is identical: - Intel Core i7 - 12G RAM - 500GB filesystem - 1 Gbit NIC dedicated for GlusterFS Software: - Linux 2.6.32.8 - GlusterFS 3.0.2 - FUSE inited with protocol versions: glusterfs 7.13 kernel 7.13 - Filesystem / Storage Backend: - LVM2 on top of software RAID 1 - ext4 with noatime I will paste the configurations inline, so people can comment on them. /etc/fstab: - /dev/data/test /mnt/brick/test ext4noatime 0 2 /etc/glusterfs/test.vol /mnt/glusterfs/test glusterfs noauto,noatime,log-level=NORMAL,log-file=/var/log/glusterfs/test.log 0 0 - *** Please note: this is the final configuration with the best results. All translators are numbered to make the explanation easier later on. Unused translators are commented out... The volume spec is identical on all nodes, except that the bind-address option in the server volume [*4*] is adjusted. *** /etc/glusterfs/test.vol - # Sat Feb 27 16:53:00 CET 2010 John Feuerstein j...@feurix.com # # Single Process Model with AFR (Automatic File Replication). ## ## Storage backend ## # # POSIX STORAGE [*1*] # volume posix type storage/posix option directory /mnt/brick/test/glusterfs end-volume # # POSIX LOCKS [*2*] # #volume locks volume brick type features/locks subvolumes posix end-volume ## ## Performance translators (server side) ## # # IO-Threads [*3*] # #volume brick # type performance/io-threads # subvolumes locks # option thread-count 8 #end-volume ### End of performance translators # # TCP/IP server [*4*] # volume server type protocol/server subvolumes brick option transport-type tcp option transport.socket.bind-address 10.1.0.1 # FIXME option transport.socket.listen-port 820 option transport.socket.nodelay on option auth.addr.brick.allow 127.0.0.1,10.1.0.1,10.1.0.2,10.1.0.3 end-volume # # TCP/IP clients [*5*] # volume node1 type protocol/client option remote-subvolume brick option transport-type tcp/client option remote-host 10.1.0.1 option remote-port 820 option transport.socket.nodelay on end-volume volume node2 type protocol/client option remote-subvolume brick option transport-type tcp/client option remote-host 10.1.0.2 option remote-port 820 option transport.socket.nodelay on end-volume volume node3 type protocol/client option remote-subvolume brick option transport-type tcp/client option remote-host 10.1.0.3 option remote-port 820 option transport.socket.nodelay on end-volume # # Automatic File Replication Translator (AFR) [*6*] # # NOTE: node3 is the primary metadata node, so this one *must* # be listed first in all volume specs! Also, node3 is the global # favorite-child with the definite file version if any conflict # arises while self-healing... # volume afr type cluster/replicate subvolumes node3 node1 node2 option read-subvolume node2 option favorite-child node3 end-volume ## ## Performance translators (client side) ## # # IO-Threads [*7*] # #volume client-threads-1 # type performance/io-threads # subvolumes afr # option thread-count 8 #end-volume # # Write-Behind [*8*] # volume wb type performance/write-behind subvolumes afr option cache-size 4MB end-volume # # Read-Ahead [*9*] # #volume ra # type performance/read-ahead # subvolumes wb # option page-count 2 #end-volume # # IO-Cache [*10*] # volume cache type performance/io-cache subvolumes wb option cache-size 1024MB option cache-timeout 60 end-volume # # Quick-Read for small files [*11*] # #volume qr # type performance/quick-read # subvolumes cache # option cache-timeout 60 #end-volume # # Metadata prefetch [*12*] # #volume sp # type performance/stat-prefetch # subvolumes qr #end-volume # # IO-Threads [*13*] # #volume
Re: [Gluster-users] GlusterFS 3.0.2 small file read performance benchmark
After reading the mail again I'm under the impression that I didn't make it clear enough: We don't have a pure read-only, but mostly read-only workload. This is the reason why we've tried GlusterFS with AFR, so we can have a multi-master read/write filesystem with a persitent copy on each node. If we wouldn't need write access every here and then, we could have gone with plain copies of the data. Now another idea is the following, based on the fact that the local ext4 filesystem + VFS cache is *much* faster: GlusterFS with populated IO-Cache: real0m38.576s user0m3.356s sys 0m6.076s # Work directly on the back-end (this is read-only...) $ cd /mnt/brick/test/glusterfs/data # Ext4 without VFS Cache: $ echo 3 /proc/sys/vm/drop_caches $ for ((i=0;i100;i++)); do tar cf - . /dev/null done; time wait real0m1.598s user0m2.136s sys 0m3.696s # Ext4 with VFS Cache: $ for ((i=0;i100;i++)); do tar cf - . /dev/null done; time wait real0m1.312s user0m2.264s sys 0m3.256s So the idea now is to bind-mount the backend filesystem *read-only* and use it for all read operations. For all write operations, use the GlusterFS mountpoint which provides locking etc. (This implies some sort of Read/Write splitting, but we can do that...) The downside is that the backend read operations won't make use of the GlusterFS on-demand self-healing. But since 99% of our read-only files are write once, read a lot of times... -- this could work out. After a node failure, a simple ls -lR should self-heal everything and the backend is fine too. The chance to read a broken file is very low? Any comments on this idea? Is there something else that could go wrong by using the backend in a pure read-only fashion that I've missed? Any ideas why the GlusterFS performance/io-cache translator with a cache-timeout of 60 is still so slow? Is there any way to *really* cache meta and filedata on GlusterFS _without_ hitting the network and thus getting very poor small file performance introduced by network latency? Are there any plans to implement support for FS-Cache [1] (CacheFS, Cachefiles), shipped with recent Linux kernels? Or to improve io-cache likewise? [1] http://people.redhat.com/steved/fscache/docs/FS-Cache.pdf Lots of questions... :) Best regards, John ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
Re: [Gluster-users] GlusterFS 3.0.2 small file read performance benchmark
Another thing that makes me wonder is the read-subvolume setting: volume afr type cluster/replicate ... option read-subvolume node2 ... end-volume So even if we play around and set this to the local node or some remote node respectively, it won't gain any performance for small files. Looks like the whole bottleneck for small files is meta-data and the global namespace lookup. It would be really great if all of this could be cached within io-cache, only falling back to a namespace query (and probably locking) if something wants to write to the file, or if the result is longer than cache-timeout seconds in the cache. So even if the file has been renamed, is unlinked, has changed permissions / metadata - simply take the version of the io-cache until it's invalidated. At least that is what I would expect the io-cache to do. This will introduce a discrepancy between the cached file version and the real version in the global namespace, but isn't that what one would expect from caching...? Note that the cache-size was in all tests on all nodes 1024MB, and the whole set of test-data was ~240MB. Add some meta-data and it's probably at 250MB. In addition, cache-timeout was 60 seconds, while the whole test took around 40 seconds. So *all* of the read-only test could have been served completely by the io-cache... or am I mistaken here? I'm trying to understand the poor performance, because network latency should be eliminated by the cache. Could some Gluster-Dev please elaborate a bit on that one? Best Regards, John ___ Gluster-users mailing list Gluster-users@gluster.org http://gluster.org/cgi-bin/mailman/listinfo/gluster-users