<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">2017-04-25 9:03 GMT+02:00 Xavier Hernandez <span dir="ltr"><<a href="mailto:xhernandez@datalab.es" target="_blank">xhernandez@datalab.es</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi Ingard,<span class=""><br>
<br>
On 24/04/17 14:43, Ingard Mevåg wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I've done some more testing with tc and introduced latency on one of my<br>
testservers. With 9ms latency artificially introduced using tc ( sudo tc<br>
qdisc add dev bond0 root netem delay 9ms ) to a testserver in the same<br>
DC as the disperse volume servers I get more or less the same throughput<br>
as I do when testing DC1 <-> DC2 (which has ~9ms ping).<br>
<br>
I know distribute volumes were more sensitive to latency in the past. At<br>
least I can max out a 1gig link with 9-10ms latency when using<br>
distribute. Disperse seems to max at 12-14MB/s with 8-10ms latency.<br>
</blockquote>
<br></span>
A pure distributed volume is a simple configuration that simply forwards a request to one of the bricks. No additional overhead is needed.<br></blockquote><div> </div><div>Well, we've still got gluster 3.0 running on an old cluster and this cluster is also dead slow when mounted at the other DC - About the same perf as get with disperse on 3.10. So some work has been done to make distribute volumes work better with increased latency.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
However a dispersed 4+2 volume needs to talk simultaneously to 6 bricks, meaning 6 network round-trips for every request. Additionally it needs to keep integrity so one or more additional requests are needed.<br></blockquote><div> </div><div>The number of bricks doesnt appear to affect the throughput. I've tried different variations of data and redundancy bricks, but the throughput seems to stay the same. For instance 8+4 compared to 4+2 has double amount of connections, but half the throughput per connection.</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
If network latency is high, all these requests contribute to increase the overall request latency, limiting the throughput.<br>
<br>
Have you tried a replica 2 or 3 ? it uses very similar integrity mechanisms so it'll also add some latency. Maybe not so much as a dispersed 4+2, but it should be perceptible.<br></blockquote><div><br></div><div>We're after capacity with this setup.</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Another test to confirm that the limitation is caused by latency is to do multiple writes in parallel. Each write will be limited by the latency, but the aggregated throughput should saturate the bandwidth, specially on a 1Gb ethernet.<br></blockquote><div><br></div><div>That has been confirmed.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Even better performance can be achieved if you distribute the writes to multiple clients or mount points (assuming they are not writing to the same file).<br>
<br>
Xavi<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">
<br>
ingard<br>
<br>
2017-04-24 14:03 GMT+02:00 Ingard Mevåg <<a href="mailto:ingard@jotta.no" target="_blank">ingard@jotta.no</a><br></span>
<mailto:<a href="mailto:ingard@jotta.no" target="_blank">ingard@jotta.no</a>>>:<span class=""><br>
<br>
I can confirm mounting the disperse volume locally on one of the<br>
three servers i got 211 MB/s with dd if=/dev/zero of=./local.dd.test<br>
bs=1M count=10000.<br>
<br>
Its not very good concidering 10gig network, but at least 20x better<br>
than 10-12MB/s<br>
<br>
2017-04-24 13:53 GMT+02:00 Pranith Kumar Karampuri<br></span>
<<a href="mailto:pkarampu@redhat.com" target="_blank">pkarampu@redhat.com</a> <mailto:<a href="mailto:pkarampu@redhat.com" target="_blank">pkarampu@redhat.com</a>>>:<span class=""><br>
<br>
+Ashish<br>
<br>
Ashish,<br>
Could you help Ingard? Do let me know what you find.<br>
<br>
On Mon, Apr 24, 2017 at 4:50 PM, Ingard Mevåg <<a href="mailto:ingard@jotta.no" target="_blank">ingard@jotta.no</a><br></span><span class="">
<mailto:<a href="mailto:ingard@jotta.no" target="_blank">ingard@jotta.no</a>>> wrote:<br>
<br>
Hi. I can't see a fuse thread at all. Please see attached<br>
screenshot of top process with threads. Keep in mind this is<br>
from inside the container.<br>
<br>
2017-04-24 12:17 GMT+02:00 Pranith Kumar Karampuri<br></span>
<<a href="mailto:pkarampu@redhat.com" target="_blank">pkarampu@redhat.com</a> <mailto:<a href="mailto:pkarampu@redhat.com" target="_blank">pkarampu@redhat.com</a>>>:<span class=""><br>
<br>
We were able to saturate hardware with EC as well. Could<br>
you check 'top' in threaded mode to see if fuse thread<br>
is saturated when you run dd?<br>
<br>
On Mon, Apr 24, 2017 at 3:27 PM, Ingard Mevåg<br></span><div><div class="h5">
<<a href="mailto:ingard@jotta.no" target="_blank">ingard@jotta.no</a> <mailto:<a href="mailto:ingard@jotta.no" target="_blank">ingard@jotta.no</a>>> wrote:<br>
<br>
Hi<br>
I've been playing with disperse volumes the past<br>
week, and so far i can not get more than 12MB/s when<br>
i do a write test. I've tried a distributed volume<br>
on the same bricks and gotten close to gigabit<br>
speeds. iperf confirms gigabit speeds to all three<br>
servers in the storage pool.<br>
<br>
The three storage servers have 10gig nics (connected<br>
to the same switch). The client is for a now a<br>
docker container in a 2nd DC (latency roughly 8-9 ms).<br>
<br>
dpkg -l|grep -i gluster<br>
ii glusterfs-client<br>
3.10.1-ubuntu1~xenial1 amd64<br>
clustered file-system (client package)<br>
ii glusterfs-common<br>
3.10.1-ubuntu1~xenial1 amd64<br>
GlusterFS common libraries and translator modules<br>
ii glusterfs-server<br>
3.10.1-ubuntu1~xenial1 amd64<br>
clustered file-system (server package)<br>
<br>
$ gluster volume info<br>
<br>
Volume Name: DFS-ARCHIVE-001<br>
Type: Disperse<br>
Volume ID: 1497bc85-cb47-4123-8f91-a07f55<wbr>c11dcc<br>
Status: Started<br>
Snapshot Count: 0<br>
Number of Bricks: 1 x (4 + 2) = 6<br>
Transport-type: tcp<br>
Bricks:<br>
Brick1: dna-001:/mnt/data01/brick<br>
Brick2: dna-001:/mnt/data02/brick<br>
Brick3: dna-002:/mnt/data01/brick<br>
Brick4: dna-002:/mnt/data02/brick<br>
Brick5: dna-003:/mnt/data01/brick<br>
Brick6: dna-003:/mnt/data02/brick<br>
Options Reconfigured:<br>
transport.address-family: inet<br>
nfs.disable: on<br>
<br>
Anyone know the reason for the slow speeds on<br>
disperse vs distribute?<br>
<br>
kind regards<br>
ingard<br>
<br>
______________________________<wbr>_________________<br>
Gluster-users mailing list<br>
<a href="mailto:Gluster-users@gluster.org" target="_blank">Gluster-users@gluster.org</a><br></div></div>
<mailto:<a href="mailto:Gluster-users@gluster.org" target="_blank">Gluster-users@gluster.<wbr>org</a>><br>
<a href="http://lists.gluster.org/mailman/listinfo/gluster-users" rel="noreferrer" target="_blank">http://lists.gluster.org/mailm<wbr>an/listinfo/gluster-users</a><span class=""><br>
<<a href="http://lists.gluster.org/mailman/listinfo/gluster-users" rel="noreferrer" target="_blank">http://lists.gluster.org/mail<wbr>man/listinfo/gluster-users</a>><br>
<br>
<br>
<br>
<br>
--<br>
Pranith<br>
<br>
<br>
<br>
<br>
--<br>
Ingard Mevåg<br>
Driftssjef<br>
Jottacloud<br>
<br></span>
Mobil: <a href="tel:%2B47%20450%2022%20834" value="+4745022834" target="_blank">+47 450 22 834</a> <tel:+47%20450%2022%20834><br>
E-post: <a href="mailto:ingard@jottacloud.com" target="_blank">ingard@jottacloud.com</a> <mailto:<a href="mailto:ingard@jottacloud.com" target="_blank">ingard@jottacloud.com</a>><br>
Webside: <a href="http://www.jottacloud.com" rel="noreferrer" target="_blank">www.jottacloud.com</a> <<a href="http://www.jottacloud.com" rel="noreferrer" target="_blank">http://www.jottacloud.com</a>><span class=""><br>
<br>
<br>
<br>
<br>
--<br>
Pranith<br>
<br>
<br>
<br>
<br>
--<br>
Ingard Mevåg<br>
Driftssjef<br>
Jottacloud<br>
<br></span>
Mobil: <a href="tel:%2B47%20450%2022%20834" value="+4745022834" target="_blank">+47 450 22 834</a> <tel:+47%20450%2022%20834><br>
E-post: <a href="mailto:ingard@jottacloud.com" target="_blank">ingard@jottacloud.com</a> <mailto:<a href="mailto:ingard@jottacloud.com" target="_blank">ingard@jottacloud.com</a>><br>
Webside: <a href="http://www.jottacloud.com" rel="noreferrer" target="_blank">www.jottacloud.com</a> <<a href="http://www.jottacloud.com" rel="noreferrer" target="_blank">http://www.jottacloud.com</a>><span class=""><br>
<br>
<br>
<br>
<br>
--<br>
Ingard Mevåg<br>
Driftssjef<br>
Jottacloud<br>
<br>
Mobil: <a href="tel:%2B47%20450%2022%20834" value="+4745022834" target="_blank">+47 450 22 834</a><br></span>
E-post: <a href="mailto:ingard@jottacloud.com" target="_blank">ingard@jottacloud.com</a> <mailto:<a href="mailto:ingard@jottacloud.com" target="_blank">ingard@jottacloud.com</a>><br>
Webside: <a href="http://www.jottacloud.com" rel="noreferrer" target="_blank">www.jottacloud.com</a> <<a href="http://www.jottacloud.com" rel="noreferrer" target="_blank">http://www.jottacloud.com</a>><span class=""><br>
<br>
<br>
______________________________<wbr>_________________<br>
Gluster-users mailing list<br>
<a href="mailto:Gluster-users@gluster.org" target="_blank">Gluster-users@gluster.org</a><br>
<a href="http://lists.gluster.org/mailman/listinfo/gluster-users" rel="noreferrer" target="_blank">http://lists.gluster.org/mailm<wbr>an/listinfo/gluster-users</a><br>
<br>
</span></blockquote>
<br>
</blockquote></div><br><br clear="all"><div><br></div><div class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div><div dir="ltr"></div></div></div></div>
</div></div>