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.
Test the disk I/O of your VPS
This discussion has been closed.
Comments
(RAID5) x 300gb 10k SAS Disks.
I think some epic abuse running on the server
Offtopic: that comment made my day.
Great guys over at @vps6net fixed it . Getting 60's and 80's now :P
wtf i was getting like ~100mB/s when i first joined @vps6net now its like 60
BuyVM 128MB:
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 6.4627 s, 166 MB/s
ChicagoVPS $7 2GB plan, Node 31:
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 13.4049 s, 80.1 MB/s
ChicagoVPS $7 2GB plan, Node 32:
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 33.154 s, 32.4 MB/s
Virpus 1.5GB:
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 39.359 s, 27.3 MB/s
Shouldn't surprise anyone, this.
ChicagoVPS $7 2GB
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 9.58574 s, 112 MB/s
BuyVM $5.95 512/1024
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 3.73633 s, 287 MB/s
Not sure if it's just me, but I've noticed that my BuyVM I/O has been pretty damn quick recently.. thinking I've been moved somewhere slightly faster? :P
Nope. The major reason why we've not had stock is we've been rebuilding all of our nodes. Whenever .32 is stable I/O will improve quite a bit more as a lot of our own monitoring/etc scripts can be put back in place.
Francisco
SecuredSpeed 256 OpenVZ LA
Running the freevps benchmark script and the dd test.
BuyVM (512MB OVZ — I had to strip out the NL/Singapore downloads since for some reason it wouldn't download them)
And, just for kicks, here's my dedi (i3 (1.6GHz is because it's under low load, so it downclocked itself), 16GB)
What's uname -a report?
I'm thinking pony6-1?
If not you should log a ticket for us to check that box, the network should be a lot more like 60 - 80MB/sec
Francisco
hostigation 256
[root@prod ~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 8.68977 seconds, 124 MB/s
buyvm 128
[root@srvdebbac ~]# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 5.80553 seconds, 185 MB/s
ubservers 512
root@serv2:~# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 118.017 s, 9.1 MB/s
@NickP @kbar Let us know if you're seeing speeds like that, even 60-80 MB/s is no good -- you should be getting 125 or above with those write tests.
I run a VPS on a RamDisk when I need todo something quick.
ou new nodes current have vps
16384+0 records in
16384+0 records out
268435456 bytes (268 MB) copied, 0.91274 seconds, 294 MB/s
[root@x32 ~]#
@prometeus Node pm14 has too poor performance.
Prometeus VZ5 on pm14 node
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 482.979 seconds, 2.2 MB/s
when did you take this shot? Just now?
what is the id of your vps?
today 7:50 GMT
mmmm.... can run the test again? At moment the node seems ok:
right now:
open a ticket, I'll give a look at this, there should be something wrong..
thanks
I think I found the problem.
When you can please retry the test.
@netguy @prometeus
I got same problem since yesterday, untill a moment ago.
Iperweb 192
This is from IPXcore 192
VPS6 128
XenVZ 256
And SecureDragon 128
@andri Right, as I said I found the problem: forgot to lower swappiness on the node...
@prometeus No problem. I'm about to open a ticket, but then it's back to normal again.
Really from a general point of view almost all applications should have worked fine, dd was hitting swap and what you saw was vswap in action
@prometeus Pardon me if I'm wrong, when I/O speed is low the download speed also decreasing yes?
you all are using conv=fdatasync, so this mean you are asking dd to pass all your data to kernel and then at the end of the write make a sync of the disk. If you want to bypass the kernel buffering you should use oflag=direct instead of conv=fdatasync...
ie:
dd if=/dev/zero of=testfilex bs=1M count=1k oflag=direct
@prometeus Fine now. You are a mage! Thank you!
when you have cpu busy on i/o disk queues and you are writing a download on disk, then yes. If you only are using sockets activity without hitting the disk then your network speed should not be affected so much.