New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
some interesting ioping stats for diff providers
All are OpenVZ
a. Ramnode SSD Cached in New York @Nick_A
ioping -c 10 ./
10 requests completed in 9011.0 ms, 992 iops, 3.9 mb/s
min/avg/max/mdev = 0.1/1.0/2.6/0.9 ms
ioping -RD ./
12926 requests completed in 3000.1 ms, 5665 iops, 22.1 mb/s
min/avg/max/mdev = 0.0/0.2/81.1/1.2 ms
b. Ramnode SSD in Seattle @Nick_A
ioping -c 10 ./
10 requests completed in 9001.5 ms, 13889 iops, 54.3 mb/s
min/avg/max/mdev = 0.1/0.1/0.1/0.0 ms
ioping -RD ./
32605 requests completed in 3000.1 ms, 26803 iops, 104.7 mb/s
min/avg/max/mdev = 0.0/0.0/4.7/0.0 ms
c. Hostus 768MB Plan (Node Server 2, $10/yr) @Alexander_M
ioping -c 10 ./
10 requests completed in 9.0 s, 382 iops, 1.5 MiB/s
min/avg/max/mdev = 90 us / 2.6 ms / 22.4 ms / 6.6 ms
ioping -RD ./
4.9 k requests completed in 3.0 s, 1.7 k iops, 6.5 MiB/s
min/avg/max/mdev = 41 us / 603 us / 47.2 ms / 1.6 ms
d. VMBox 1GB Plan (Node: LAOVZ1, $20/yr) @vmbox
ioping -c 10 ./
10 requests completed in 9022.9 ms, 467 iops, 1.8 mb/s
min/avg/max/mdev = 0.2/2.1/18.0/5.3 ms
ioping -RD ./
9947 requests completed in 3000.1 ms, 4176 iops, 16.3 mb/s
min/avg/max/mdev = 0.1/0.2/27.2/1.1 ms
e. Rijx ($5/yr plan) @Rijx
ioping -c 10 ./
10 requests completed in 9003.0 ms, 5718 iops, 22.3 mb/s
min/avg/max/mdev = 0.1/0.2/0.3/0.0 ms
ioping -RD ./
22252 requests completed in 3000.4 ms, 14666 iops, 57.3 mb/s
min/avg/max/mdev = 0.0/0.1/3.0/0.1 ms
f. XVM Labs ($2.87/yr Plan) @dcc
ioping -c 10 ./
10 requests completed in 9004.5 ms, 188679 iops, 737.0 mb/s
min/avg/max/mdev = 0.0/0.0/0.0/0.0 ms
ioping -RD ./
10245 requests completed in 3000.1 ms, 28998 iops, 113.3 mb/s
min/avg/max/mdev = 0.0/0.0/13.7/0.3 ms
g. VPSdime ($15/yr plan in Dallas) @Serverian
ioping -c 10 ./
10 requests completed in 9010.6 ms, 1062 iops, 4.1 mb/s
min/avg/max/mdev = 0.1/0.9/7.3/2.1 ms
ioping -RD ./
18120 requests completed in 3000.1 ms, 10319 iops, 40.3 mb/s
min/avg/max/mdev = 0.1/0.1/13.4/0.2 ms
h. VPSDime Storage Plan ($7/mo for 500GB in Dallas) @Serverian
ioping -c 10 ./
10 requests completed in 9008.7 ms, 1341 iops, 5.2 mb/s
min/avg/max/mdev = 0.1/0.7/6.5/1.9 ms
ioping -RD ./
1063 requests completed in 3004.0 ms, 362 iops, 1.4 mb/s
min/avg/max/mdev = 0.0/2.8/6.9/2.4 ms
i. Iniz ($7/yr in NY) @Iniz
ioping -c 10 ./
10 requests completed in 9007.0 ms, 2051 iops, 8.0 mb/s
min/avg/max/mdev = 0.1/0.5/3.4/1.0 ms
ioping -RD ./
326 requests completed in 3009.3 ms, 109 iops, 0.4 mb/s
min/avg/max/mdev = 0.1/9.2/698.4/51.8 ms
j. crissic ($15/yr for 512MB Plan in Jacksonville) @Skylar_M
ioping -c 10 ./
10 requests completed in 9.0 s, 11.8 k iops, 46.3 MiB/s
min/avg/max/mdev = 62 us / 84 us / 111 us / 15 us
ioping -RD ./
62.4 k requests completed in 3.0 s, 22.0 k iops, 86.0 MiB/s
min/avg/max/mdev = 28 us / 45 us / 23.1 ms / 191 us
k. GVH SSD Cached (!!) ($9/yr plan) @GreenvalueHost
ioping -c 10 ./
10 requests completed in 9.5 s, 18 iops, 74.0 KiB/s
min/avg/max/mdev = 449 us / 54.0 ms / 150.1 ms / 48.1 ms
ioping -RD ./
89 requests completed in 3.0 s, 29 iops, 118.3 KiB/s
min/avg/max/mdev = 230 us / 33.8 ms / 283.7 ms / 53.2 ms
l. LES in Dallas ($4/yr plan) @anthonysmith
ioping -c 10 ./
10 requests completed in 9003.9 ms, 4211 iops, 16.4 mb/s
min/avg/max/mdev = 0.2/0.2/0.4/0.1 ms
ioping -RD ./
13925 requests completed in 3000.2 ms, 6578 iops, 25.7 mb/s
min/avg/max/mdev = 0.1/0.2/3.9/0.1 ms
m. FTPIT ($20/yr SSD Plan) @ftpit_radi
ioping -c 10 ./
10 requests completed in 9001.7 ms, 13210 iops, 51.6 mb/s
min/avg/max/mdev = 0.1/0.1/0.1/0.0 ms
ioping -RD ./
30501 requests completed in 3000.1 ms, 23536 iops, 91.9 mb/s
min/avg/max/mdev = 0.0/0.0/188.1/1.1 ms
n. i-83 in France ($6/yr ipv6 plan) @i83
ioping -c 10 ./
10 requests completed in 9063.3 ms, 161 iops, 0.6 mb/s
min/avg/max/mdev = 0.1/6.2/39.4/12.8 ms
ioping -RD ./
547 requests completed in 3002.9 ms, 184 iops, 0.7 mb/s
min/avg/max/mdev = 0.1/5.4/103.2/7.0 ms
o. i-83 in Bulgaria ($6/yr IPv6 plan) @i83
ioping -c 10 ./
10 requests completed in 9003.0 ms, 4838 iops, 18.9 mb/s
min/avg/max/mdev = 0.2/0.2/0.2/0.0 ms
ioping -RD ./
554 requests completed in 3003.6 ms, 187 iops, 0.7 mb/s
min/avg/max/mdev = 0.1/5.4/42.0/3.6 ms
p. WeLoveServers ($19/yr in Buffalo) @Oxide
ioping -c 10 ./
10 requests completed in 9.1 s, 173 iops, 694.1 KiB/s
min/avg/max/mdev = 69 us / 5.8 ms / 29.7 ms / 11.4 ms
ioping -RD ./
47.6 k requests completed in 3.0 s, 16.3 k iops, 63.8 MiB/s
min/avg/max/mdev = 30 us / 61 us / 218.5 ms / 1.2 ms
Thanked by 1i83
Comments
Here's ours:
These assume you have a
noopscheduler andvdaattachment, with the latest virtio drivers installed [ie: best case scenario.] These tests were run on a full node for fair's sake.I'm not very surprised about the low disk seek test, as we do not use local storage; there is "some" latency attached to your requests. We balance out the latency issue with lots and lots of I/O read caching on the nodes, however; the "RD" switches bypass this caching completely.
EDIT
I ran one more with the -RD flag just in case:
Seems this one is really spotty, and depends on your luck more than anything else.
Sorry, complete waste of time those results, 90% of them are just hitting the disk cache
ioping + VPS = pointless results, OpenVZ will always beat Xen+KVM as there is no separation on the storage.
Someone explained it really well a while ago maybe @rds100 ? but essentially unless your using native storage dont pay to much attention to basic ioping results they are skewed due to so many variables that give inaccurate real word results.
For example apparently an LES node with 400+ containers on a single disk is better than some others? I think not
your just hitting the cache is all which will always give better results than an actual disk i.e. where your data actually lives 
This is a very good point. One more point is that there really is no good way to benchmark the I/O "performance" of a virtual storage. It's very important to note that a benchmark, by design, is artificial; and not indicative of real world performance.
The best way to test how a VM will hold up to load, is giving it real load.
so ioping -RD is hitting the cache too?
The "RD" can hit the cache if the host machine forces caching, but the guest [ KVM/XEN ] writes directly to the virtual disk. I have found a few ways to trick these benchmarks in my own testing, so I would not even trust these.
It hits the cache of the RAID card, if there is one. It also hits the cache of the HDDs. You can't flush those caches.
Well, on the LSI MegaRAID cards [ some of them at least ] you can control the caching, including disabling it completely; flushing it; changing the caching policy; etcetera. You can even make these changes live on an active VD in some cases [ updated firmware + latest utility + newer cards. ]
@GoodHosting even if "you" can, ioping can't and don't do anything with the cache of the RAID card, neither it has any knowledge of the RAID cards, etc.
Of course, of course
.
Ouch.