I/O speed for a budget VPS

birdie25

30MB/s + is perfectly fine for a budget VPS.

Is this true?

Ash_Hawkridge

30MB/s is probably enough sequential write to perform most tasks, yes. Disk I/O by the way of DD testing is usually just a competition for higher numbers but in reality nobody needs 180MB/s to perform every day VPS tasks.

djvdorp

IMHO 30mb/s is a little bit low

Aldryic

BuyVM

-09:09:07- Nurvus:~ :: root # dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 записей считано
16384+0 записей написано
 скопировано 1073741824 байта (1,1 GB), 4,67377 c, 230 MB/c

Hostigation

-19:15:14- Jehuty:~ :: root # dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 записей считано
16384+0 записей написано
 скопировано 1073741824 байта (1,1 GB), 6,60342 c, 163 MB/c

SecureDragon

-18:11:42- Status3:~ :: root # dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 записей считано
16384+0 записей написано
 скопировано 1073741824 байта (1,1 GB), 8,43695 c, 127 MB/c

I would say no... just because a provider is "budget" doesn't mean the quality has to be poor.

AsuraHosting

@Aldryic said: I would say no... just because a provider is "budget" doesn't mean the quality has to be poor.

True, but what's considered okay for a budget VPS?

I think anything above 50mb/s would be alright, anything below... I'd be a bit wary.

Derek

[root@mich sbin]# 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, 7.69921 s, 139 MB/s

From ErrantWeb.

Aldryic

@AsuraHosting said: True, but what's considered okay for a budget VPS?

An unmanaged provider that strives to bring their clients the best quality they can. I'm not saying we should always be on par with the F500 companies... but I firmly believe that the "oh well, I get what I pay for" mentality is doing nothing but hurting the market for us.

LES

@Aldryic said: скопировано 1073741824 байта (1,1 GB), 4,67377 c, 230 MB/c

My BuyVM is not that fast:

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
1073741824 bytes (1.1 GB) copied, 14.7777 s, 72.7 MB/s

but very generous with ram :-p

free
total used free shared buffers cached
Mem: 4294967295 0 4294967295 0 0 880800
-/+ buffers/cache: 4294086496 880799
Swap: 0 0 0

yomero

wtf?¨That is... weird

gsrdgrdghd

@LES said: My BuyVM is not that fast:

Mine neither, its 65 mb/s
But anyway thats fast enough i'd say

Aldryic

@LES @gsrdgrdghd - Mind popping in a ticket for us? IO shouldn't be that low.

And yeah, the Ram issue is some funky bug Fran's currently working on :P

Edit: For the record, that dd took place on node kvm02. Here's another quick one for consistency:

-09:53:51- Nurvus:~ :: root # dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 записей считано
16384+0 записей написано
 скопировано 1073741824 байта (1,1 GB), 5,00609 c, 214 MB/c

Legendlink

Speaking of awful disk I/O, 123Systems:

root@din:~# 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, 95.3984 s, 11.3 MB/s

As compared to my other VPS's such as @Aldryic posted.

LES

@Aldryic said: Mind popping in a ticket for us?

I/O is fine for a 15$/Y but if you like to look into it (ram bug also here) it is on 128 (node24)

Francisco

You guys are probably on older nodes that are still not redone

The KVM upgrades have been quite awesome for sure!

Francisco

u4ia

I think the real question is, at what point is it detrimental to everyday VPS performance?

FWIW on my (at work) desktop with a single 7200rpm sata drive, i get an average of about 65 MB/s.

dnom

@Aldryic said: IO shouldn't be that low.

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, 19.9644 s, 53.8 MB/s 

that's on node25. It's fast enough for my needs and faster than lots of budget vps providers so I have no complain. Also my other vps on another node is having ~175 MB/s ave. w/c is excellent i think

Francisco

@u4ia said: I think the real question is, at what point is it detrimental to everyday VPS performance?

FWIW on my (at work) desktop with a single 7200rpm sata drive, i get an average of about 65 MB/s.

Very true, but that's dedicated to pretty much you

Things like nightly/hourly crons all fire at once and will really drag on a nodes IO if it's anything big. Distros like CentOS have 'rpmq' updating hourly for some retarded reason and that'll try to eat a full core no matter what.

@dnom - You're on a pony6-1 setup, should improve greatly whenever you get pulled.

Francisco

KuJoe

Our "limit" is 50MB/s, if it goes any lower (consistently) we consider it to slow for production.

Consistency is better than performance though and should be the true test of disk I/O. If your VPS is consistently 50MB/s then it's much much better than a VPS that constantly fluctuates between 100MB/s and 10MB/s.

KuJoe

@u4ia said: I think the real question is, at what point is it detrimental to everyday VPS performance?

FWIW on my (at work) desktop with a single 7200rpm sata drive, i get an average of about 65 MB/s.

And to add more reference. My desktop I built was capped at 38MB/s which was still usable with Windows 7 although sluggish, if I wasn't running a GUI I'm sure 38MB/s would have worked great.

dnom

@Francisco said: You're on a pony6-1 setup, should improve greatly whenever you get pulled.

Although I don't really need high disk i/o speed It'll definitely feel better being on a node pulled by stronger ponies!

dmmcintyre3

$30/year Hostigation KVM:

[root@tiger ~]# 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, 5.31558 s, 202 MB/s

miTgiB

@Aldryic said: some funky bug Fran's currently working on

or created, errr Franned into existence

Aldryic

That would be more accurate, yes

fly

[herp @ de1]$ dd if=/dev/zero of=urt.tar.gz bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 7.54315 s, 142 MB/s

errant web
was like 175 when i first got it (probably cuz new), but now its settled around 150

Aldryic

Use the <pre> tags.

Asim

@VMPort said: 30MB/s is probably enough sequential write to perform most tasks, yes. Disk I/O by the way of DD testing is usually just a competition for higher numbers but in reality nobody needs 180MB/s to perform every day VPS tasks.

Yes, its perfectly fine if the host is good. If its a new provider then chances are that they will oversell and you will see 2.3MB/sec etc

Kairus

My hostigation KVM:

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, 10.2865 s, 104 MB/s

My Alienlayer OVZ:

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, 13.643 s, 78.7 MB/s

The alienlayer VPS feels sluggish sometimes, if I ran ioping I bet the latency would be all over the place, so raw I/O speed is just a part of it really.

Clinton

**AllGamer **VPS

"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, 3.6868 seconds, 291 MB/s"

Clinton

Double Post -Sorry bad connection-

Mon5t3r

raid10

root@test:~# dd if=/dev/zero of=test.file bs=64k count=16k conv=fdatasync;rm -f test.file
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1.32328 s, 811 MB/s
root@test:~# pwd
/root

raid1

root@lux:~# dd if=/dev/zero of=test.file bs=64k count=16k conv=fdatasync;rm -f test.file
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1.29378 s, 830 MB/s
root@lux:~# pwd
/root

yomero

@Clinton said: **AllGamer **VPS

Your company isn't Budget at all.

@Mon5t3r said: raid10

@Mon5t3r said: raid1

That isn't budget too ¬_¬
Also, how a Raid1 can be the same as Raid10