OVH Community, your new community space.

Bandwidth test


glidewave
05-11-2010, 10:36
Quote Originally Posted by Myatu
You just answered your own question
My statement/question was sarcastic I know how transit/peering works...just because it's a 10G, I know not to expect stellar performance outside OVH's own network because of how the Internet is routed, OVH may lead the way as far as offering 10G (at i d i o tic setup fees) (yes I had to do it that way since the forum parsed the word i d i o tic"). Anyone that is in a location that isn't peer'd or transit'd from ovh shouldn't expect 10G speeds... common freakin' sense.

What it boils down to is that OVH does what it says.... they just need to GREATLY work on their support.


J

iUnknown
31-10-2010, 20:47
100%[================>] 125,000,000 4.98M/s in 16s

2010-10-31 20:44:43 (7.40 MB/s) - `1Gb.dat' saved [125000000/125000000]
UK - RapidSwitch

Myatu
31-10-2010, 12:28
Quote Originally Posted by glidewave
like some ISPs do.
You just answered your own question

Andy
31-10-2010, 10:05
A file could be cached on ISP's local cache servers. Virgin Media used to use them to save bandwidth but in the end they realised they were causing more harm than good and so they removed them over the course of a year. Some ISP's may still use them though.

glidewave
31-10-2010, 05:53
Quote Originally Posted by Myatu
I was just checking this again, and noticed the files are static (even includes MD5 and SHA1 checksums).

For the smaller files this is kind of moot really, as they'll end up being cached somewhere along the line and distort the test results. Or are they re-generated every few minutes?
I would love to know where a "file" would be cached along the way, not counting on the destination box in ram. Proper DC's (like ovh) don't do caching/speedup like some ISPs do.

10GBit file:
To another ovh gig-e (with other activity): (96.4 MB/s)

Transport from ovh to usa or from any other eu dc to usa is going to be variable. That is how it goes with trans/a/p transit - just because you have gig-e in the us, doesn't mean you get gig e from outside the usa.... although the closest I've gotten is softlayer or theplanet at over 100mbps with fdc a close 3rd.

J.

Myatu
30-10-2010, 19:00
I was just checking this again, and noticed the files are static (even includes MD5 and SHA1 checksums).

For the smaller files this is kind of moot really, as they'll end up being cached somewhere along the line and distort the test results. Or are they re-generated every few minutes?

BELLonline
28-09-2010, 18:50
Quote Originally Posted by NickW
From my Gbit server in Ohio, US.

Code:
# wget http://proof.ovh.net/1Gio.dat -O /dev/null
--2010-09-27 07:34:49--  http://proof.ovh.net/1Gio.dat
Resolving proof.ovh.net... 188.165.12.106, 2001:41d0:2:876a::1
Connecting to proof.ovh.net|188.165.12.106|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1073741824 (1.0G) [application/octet-stream]
Saving to: `/dev/null'

100%[====================================>] 1,073,741,824 21.9M/s   in 59s

2010-09-27 07:35:49 (17.4 MB/s) - `/dev/null' saved [1073741824/1073741824]
Speed maxed at 25-26MB/s then would drop to 2-3MB/s and speed up again. Why such unstable speed?
Maybe they accidentally capped their own server lol

NickW
27-09-2010, 12:37
From my Gbit server in Ohio, US.

Code:
# wget http://proof.ovh.net/1Gio.dat -O /dev/null
--2010-09-27 07:34:49--  http://proof.ovh.net/1Gio.dat
Resolving proof.ovh.net... 188.165.12.106, 2001:41d0:2:876a::1
Connecting to proof.ovh.net|188.165.12.106|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1073741824 (1.0G) [application/octet-stream]
Saving to: `/dev/null'

100%[====================================>] 1,073,741,824 21.9M/s   in 59s

2010-09-27 07:35:49 (17.4 MB/s) - `/dev/null' saved [1073741824/1073741824]
Speed maxed at 25-26MB/s then would drop to 2-3MB/s and speed up again. Why such unstable speed?

markrandall
27-09-2010, 11:39
A few US tests all servers on gigabit ports.

First of all from gigenet:

Code:
100%[======================================>] 125,000,000 11.7M/s   in 13s

2010-09-27 05:26:04 (9.36 MB/s) - `1Gb.dat.1' saved [125000000/125000000]
Then softlayer:

Code:
100%[======================================>] 125,000,000 3.94M/s   in 40s

2010-09-27 05:27:03 (3.01 MB/s) - `1Gb.dat' saved [125000000/125000000]
Lastly from Hurricane Electric:

Code:
100%[======================================>] 125,000,000 13.7M/s   in 15s

2010-09-27 03:29:24 (7.85 MB/s) - `1Gb.dat' saved [125000000/125000000]
All from proof.ovh.net, then

From gigenet:

Code:
100%[======================================>] 125,000,000 22.8M/s   in 9.3s

2010-09-27 05:33:31 (12.8 MB/s) - `1Gb.dat.2' saved [125000000/125000000]
From softlayer:

Code:
100%[======================================>] 125,000,000 4.61M/s   in 50s

2010-09-27 05:34:27 (2.38 MB/s) - `1Gb.dat.1' saved [125000000/125000000]
From Hurricane Electric:

Code:
100%[======================================>] 125,000,000 14.5M/s   in 11s

2010-09-27 03:31:35 (10.7 MB/s) - `1Gb.dat.1' saved [125000000/125000000]
Those were using http://91.121.215.100/

Interestingly though, a transfer from my own OVH server to softlayer is much much faster:

Code:
100%[======================================>] 104,857,600 11.1M/s   in 14s

2010-09-27 05:36:47 (7.05 MB/s) - `100MBtest.zip.2' saved [104857600/104857600]

yonatan
20-09-2010, 19:50
Quote Originally Posted by kro
yonatan wrote:
> there is a nasty bug with the intel cards + openVZ , which i found a
> dirty hack fix for.


That's a software problem, not hardware problem.
--
Felix
OVH Team
Bad Bad Software problem
I'm so happy we found a solution!

Performance is just SOO GOOD now

Myatu
20-09-2010, 16:50
Quote Originally Posted by kro
That's a software problem, not hardware problem.
He didn't indicate the contrary

kro
20-09-2010, 16:36
yonatan wrote:
> there is a nasty bug with the intel cards + openVZ , which i found a
> dirty hack fix for.


That's a software problem, not hardware problem.
--
Felix
OVH Team

Thelen
20-09-2010, 05:19
What I want to see is jumbo packets enabled for the OVH network, then we will see what speeds (actually in theory no increase for such a big file, but for smaller files... )

curiosity
20-09-2010, 01:44
Quote Originally Posted by Andy
A kitten lives!
not for long , normal service will be resumed a soon as possible...........au revoir Monsieur Pussy Cat !

glidewave
19-09-2010, 19:18
Quote Originally Posted by Winit
People wouldn't have to multi-thread if it was any decent.

What is your connection speed?
Incorrect.

10Mbps/10Mbps and at another location 30Mbps/2Mbps. Additionally, I have a server with The Planet out of Texas and can nearly get 50Mbps from an ovh box to texas (1thread, tx box is 100mbps). Max I've gotten from any ovh box to my tx server at one time is 70Mbps. Which is nearly the same as any other EU host I've used (leaseweb, rootservers, serverloft, uknet, keyweb)

ISP network/toplogy and peering/transit plays a great deal into speeds an eu get as well.

While I dislike their strange, ever changing pricing and erratic support, their network is better than most.

Winit
19-09-2010, 19:06
Quote Originally Posted by glidewave
Indeed ovh to usa means nothing.... I am able to max out my connection from most eu hosts including OVH, you have to multi-thread them.
People wouldn't have to multi-thread if it was any decent.

What is your connection speed?

glidewave
19-09-2010, 17:42
Indeed ovh to usa means nothing.... I am able to max out my connection from most eu hosts including OVH, you have to multi-thread them.

glidewave
19-09-2010, 17:39
from proof.ovh to Leaseweb (evo):
2010-09-19 11:29:50 (32.1 MB/s) (threading this multiple times doesn't help, which is a LW ns bw issue)

from 91.121.215.100 to Leaseweb (evo):
2010-09-19 11:35:34 (11.1 MB/s)

yonatan
19-09-2010, 02:18
Oles,
what switch model do you use for HG 10Gbit servers?

there is a nasty bug with the intel cards + openVZ , which i found a dirty hack fix for.

i think this needs to be checked

see ticket #537241

Andy
18-09-2010, 23:47
A kitten lives!

makno
18-09-2010, 23:41
ouch oles posting in english :P

oles@ovh.net
18-09-2010, 21:38
> OVH to USA is usually poor.

"OVH to USA" means nothing.


_Lemon_
17-09-2010, 13:34
1Gbps unmetered from Leaseweb:

Code:
[llama ~] wget http://proof.ovh.net/10Gb.dat -O /dev/null
--2010-09-17 12:33:01--  http://proof.ovh.net/10Gb.dat
Resolving proof.ovh.net... 188.165.12.106
Connecting to proof.ovh.net|188.165.12.106|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1250000000 (1.2G) [application/octet-stream]
Saving to: `/dev/null'

100%[=================================================================================================================>] 1,250,000,000 32.6M/s   in 42s     

2010-09-17 12:33:44 (28.3 MB/s) - `/dev/null' saved [1250000000/1250000000]
Code:
[llama ~] wget http://91.121.215.100/10Gb.dat -O /dev/null
--2010-09-17 12:34:07--  http://91.121.215.100/10Gb.dat
Connecting to 91.121.215.100:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1250000000 (1.2G) [application/octet-stream]
Saving to: `/dev/null'

100%[=================================================================================================================>] 1,250,000,000 11.9M/s   in 1m 42s  

2010-09-17 12:35:49 (11.7 MB/s) - `/dev/null' saved [1250000000/1250000000]

YouWhat
16-09-2010, 22:24
Good evening,

Thank you!

If I read your feedbacks well (more than 250), in top of
100Mbps (12MB/sec) the results are to 5% similar. It is
to say with the ADSL, the server 100Mbps and everywhere in
Europe.

In continuation, if you have a server 1Gbps/10Gbps in France
or in Europe, that depends. That is to say the direct one is more
rapid, is the proof-a10 is faster. I think that
that depends from where the traffic is drawn and outdistances it in
kilometers between the IP. The fragmentation of the packages
with MTU 1500/9216 fact that the ACK take some
time to arrive. And that must slow down the flow. But
can be also the fact that one does not have to decontaminate yet
the MPLS between Roubaix/Paris plays too. In any case it is
very interesting and there are many things to include/understand
and to improve in the absolute one but also in the setting in
place des load balancor like a10. I did not activate
the mask interns on the router to force the transfers
between the infrastructures but one can still optimize
flows.

Thank you with all.

In a friendly way
Octave
Translated via babelfish

oles@ovh.net
16-09-2010, 22:22
Good evening,

Thanks!

If I've read your feedback correctly (over 250), above 100Mbps (12Mb/sec) the results are similar to 5% i.e.: with ADSL, 100Mbps servers and across Europe.

Next, if you have a server 1Gbps/10Gbps in France or Europe, it depends. The direct-proof is faster, as the a10-proof is faster. I think it depends on where the traffic is drawn and the distance in kilometers between the IPs. Packet fragmentation with MTU 1500/9216 done with the ACKs take some time to arrive. And this will slow the flow. But may also be the fact that we did not disable the MPLS between Roubaix / Paris that plays on too. In any case it's very interesting and there are many things to understand and improve in absolute terms but also in setting up load balancer as a10. I have not activated the internal cache on the router to force the transfer between the infrastructure, but we can further optimise flows.

Thank you all.

Regards

Octave

YouWhat
16-09-2010, 21:25
If you have 12 minutes…

Good evening,
If you have 12 minutes, thank you to test 2 remote loadings
on http://91.121.215.100 then 2 same remote loadings
on http://proof.ovh.net then to send to me on oles@ovh.net
your verdict: it is the same thing or there is a difference?

Version skirts:

We will benefit from this advertisement to test
a new router for distribution of load.
It is about a load balancer of mark a10 connected
in 10G in Paris. While connecting itself on http://91.121.215.100/
on a10 you “are redirected/load balanced” on
http://proof.ovh.net which him is in… Roubaix.
Why make? To test the performances. In
theory one must have the same performances since it
acts of a waiter proof either on line or with a10
in front of. But is it true? Is it
stable?

This new router there one thinks of using for Cloud Computing
on the level of cd. load balancing but not only. It allows
to relocate IPv6 towards IPv4 and thus if you have a site
who walks, without anything to change and in less than 5 minutes
anywhere in the world, you can start your site
in IPv6… the SSL in hardware, compression gzip in hardware
etc etc short, there is hope… but good before it is necessary
to validate the tests…

Thank you for your assistance.

In a friendly way
Octave
Translated via babelfish

oles@ovh.net
16-09-2010, 20:54
If you have 12 minutes ...

Good evening,
If you have 12 minutes could you please test the 2 downloads on http://91.121.215.100 then the same download on http://proof.ovh.net then send them to me on oles@ovh.net
Your verdict: it's the same thing or there's a difference?

Long version:

We will use this test to announce a new router for load balancing. This is a a10 load balancer brand connected in 10G to Paris. it's connecting to http://91.121.215.100/ on the a10 you are "redirected/load balanced" on http://proof.ovh.net which is found in.....Roubaix.

Why do this? To test the performances. In theory it should have the same performance as that of the proof server either directly or before with the a10 . But is this right? Is it stable?

We're thinking of using this new router for Cloud Computing at the dC load balancing level but not only just this. It can translate IPv6 to IPv4 and so if you have a working site, without changing anything, in less than 5 minutes anywhere in the world, you can start your site in IPv6 ... the SSL hardware, hard gzip compression etc. etc.... In short, there is what we're hoping for ... but well before that we must validate the tests ...

Thanks for your help.

All the best

Octave

Winit
16-09-2010, 20:19
Quote Originally Posted by mks
I am downloading at 100kb/s max from USA on 6mbit DSL same speed as the servers I have with OVH.
OVH to USA is usually poor.

LawsHosting
16-09-2010, 15:09
Kimsufi i7-2T with 1gb nic:
31% [==========> ] 3,113,940,062 102M/s eta 65s
Kimsufi old Daul-Core:
0% [ ] 96,499,544 11.28M/s ETA 14:02

mks
16-09-2010, 14:55
I am downloading at 100kb/s max from USA on 6mbit DSL same speed as the servers I have with OVH.

YouWhat
16-09-2010, 14:54
Resolving proof.ovh.net... 188.165.12.106
Connecting to proof.ovh.net|188.165.12.106|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1048576000 (1000M) [application/x-ns-proxy-autoconfig]
Saving to: `1Gio.dat'

100%[====================================>] 1,048,576,000 11.2M/s in 90s

2010-09-16 15:49:57 (11.1 MB/s) - `1Gio.dat' saved [1048576000/1048576000]
Off Kimsufi Q-1T

Resolving proof.ovh.net... 188.165.12.106
Connecting to proof.ovh.net|188.165.12.106|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1048576000 (1000M) [application/x-ns-proxy-autoconfig]
Saving to: `1Gio.dat'

100%[================================================== ============================>] 1,048,576,000 105M/s in 9.9s

2010-09-16 15:53:16 (101 MB/s) - `1Gio.dat' saved [1048576000/1048576000]
Off EG Best Of

yonatan
16-09-2010, 14:29
Code:
:~# wget -O /dev/null http://proof.ovh.net/10Gb.dat
--2010-09-16 15:25:35--  http://proof.ovh.net/10Gb.dat
Resolving proof.ovh.net... 188.165.12.106
Connecting to proof.ovh.net|188.165.12.106|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 10000000000 (9.3G) [application/x-ns-proxy-autoconfig]
Saving to: `/dev/null'

100%[================================================>] 10,000,000,000  422M/s   in 22s

2010-09-16 15:25:57 (425 MB/s) - `/dev/null' saved [10000000000/10000000000]
Impressive,

Andy
16-09-2010, 12:22
Hello
It's been a while since we have been asked to
set up a server specifically
for testing bandwidth. By
example, to download the files and
see what the quality of the tape
bandwidth and throughput.

Done:
http://proof.ovh.net/

The server is connected HG 10G
our network of HG as a standard HG.

Regards
Octave

oles@ovh.net
16-09-2010, 12:22
Hello
It's been a while since we have been asked to set up a server specifically for testing bandwidth. Eg.: to download files and
see the quality of bandwidth together with the broad band.

It's done:
http://proof.ovh.net/

The server is a 10G HG connected to our network from a HG like a HG standard.

All the best

Octave