Gandalf Corvotempesta
2016-Sep-06 12:29 UTC
[Gluster-users] Tiering and sharding for VM workload
Anybody? Il 05 set 2016 22:19, "Gandalf Corvotempesta" < gandalf.corvotempesta at gmail.com> ha scritto:> Is tiering with sharding usefull with VM workload? > Let's assume a storage with tiering and sharding enabled, used for > hosting VM images. > Each shard is subject to tiering, thus the most frequent part of the > VM would be cached on the SSD, allowing better performance. > > Is this correct? > > To put it simple, very simple, let's assume a webserver VM, with the > following directory structure: > > /home/user1/public_html > /home/user2/public_html > > both are stored on 2 different shards (i'm semplyfing). > /home/user1/public_html has much more visits than user2. > > Would that shard cached on hot tier allowing faster access by the > webserver? >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160906/23991d07/attachment.html>
----- Original Message -----> From: "Gandalf Corvotempesta" <gandalf.corvotempesta at gmail.com> > To: "gluster-users" <Gluster-users at gluster.org> > Sent: Tuesday, September 6, 2016 8:29:06 AM > Subject: Re: [Gluster-users] Tiering and sharding for VM workload > > > > Anybody?Paul Cruzner did some tests with sharding+tiering, I think the intent was to investigate the VM workload case. In general, at the moment, the larger the transfer size, the better chance tiering will be able to help you. Shards of VM images would (I suppose) be "large", so your idea may see benefit. The set of webpages/VM shards in your example would have to stay relatively stable over time and fit on the hot tier.> > Il 05 set 2016 22:19, "Gandalf Corvotempesta" < > gandalf.corvotempesta at gmail.com > ha scritto: > > > Is tiering with sharding usefull with VM workload? > Let's assume a storage with tiering and sharding enabled, used for > hosting VM images. > Each shard is subject to tiering, thus the most frequent part of the > VM would be cached on the SSD, allowing better performance. > > Is this correct? > > To put it simple, very simple, let's assume a webserver VM, with the > following directory structure: > > /home/user1/public_html > /home/user2/public_html > > both are stored on 2 different shards (i'm semplyfing). > /home/user1/public_html has much more visits than user2. > > Would that shard cached on hot tier allowing faster access by the webserver? > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users
Krutika Dhananjay
2016-Sep-06 13:08 UTC
[Gluster-users] Tiering and sharding for VM workload
Theoretically whatever you said is correct (at least from shard's perspective). Adding Rafi who's worked on tiering to know if he thinks otherwise. It must be mentioned that sharding + tiering hasn't been tested as such till now by us at least. Did you try it? If so, what was your experience? -Krutika On Tue, Sep 6, 2016 at 5:59 PM, Gandalf Corvotempesta < gandalf.corvotempesta at gmail.com> wrote:> Anybody? > > Il 05 set 2016 22:19, "Gandalf Corvotempesta" < > gandalf.corvotempesta at gmail.com> ha scritto: > >> Is tiering with sharding usefull with VM workload? >> Let's assume a storage with tiering and sharding enabled, used for >> hosting VM images. >> Each shard is subject to tiering, thus the most frequent part of the >> VM would be cached on the SSD, allowing better performance. >> >> Is this correct? >> >> To put it simple, very simple, let's assume a webserver VM, with the >> following directory structure: >> >> /home/user1/public_html >> /home/user2/public_html >> >> both are stored on 2 different shards (i'm semplyfing). >> /home/user1/public_html has much more visits than user2. >> >> Would that shard cached on hot tier allowing faster access by the >> webserver? >> > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160906/c421c076/attachment.html>
On Tue, Sep 6, 2016 at 7:29 AM, Gandalf Corvotempesta < gandalf.corvotempesta at gmail.com> wrote:> Anybody? > >While I have not tested it yet the 2 email chains I have seen from users trying it is that the performance has been worse rather than any increased benefit. Perhaps those using it successfully are just quiet and haven't responded when others had issues. Il 05 set 2016 22:19, "Gandalf Corvotempesta" <gandalf.corvotempesta at gmail.> com> ha scritto: > >> Is tiering with sharding usefull with VM workload? >> Let's assume a storage with tiering and sharding enabled, used for >> hosting VM images. >> Each shard is subject to tiering, thus the most frequent part of the >> VM would be cached on the SSD, allowing better performance. >> >> Is this correct? >> >> To put it simple, very simple, let's assume a webserver VM, with the >> following directory structure: >> >> /home/user1/public_html >> /home/user2/public_html >> >> both are stored on 2 different shards (i'm semplyfing). >> /home/user1/public_html has much more visits than user2. >> >> Would that shard cached on hot tier allowing faster access by the >> webserver? >> > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://www.gluster.org/mailman/listinfo/gluster-users >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160906/18ea1b4d/attachment.html>