how is the outgoing bw? just wondering if they limit connections that leave the OVH network.
root@sys-arm-srv1:~# speedtest-cli
Retrieving speedtest.net configuration...
Retrieving speedtest.net server list...
Testing from OVH Hosting (54.39.62.160)...
Selecting best server based on latency...
Hosted by Fibrenoire Internet (Montreal, QC) [0.63 km]: 4.668 ms
Testing download speed........................................
Download: 507.13 Mbit/s
Testing upload speed..................................................
Upload: 126.55 Mbit/s
I think that's BS, no mention of source/destination and such. just some iperf numbers, not even the command that has been run.
bench from @sin above also suggests otherwise. if the connection is not 250Mbps as advertised I definitely drop it. should be able to at least burst to it within europe, won't wait days if I ever need to recover some GB ... but I'll checl and see in a few anyway.
I think that's BS, no mention of source/destination and such. just some iperf numbers, not even the command that has been run.
bench from @sin above also suggests otherwise. if the connection is not 250Mbps as advertised I definitely drop it. should be able to at least burst to it within europe, won't wait days if I ever need to recover some GB ... but I'll checl and see in a few anyway.
I hope that You are right, because those look like nice backup rsync servers.
Hrrmmm this is weird, one moment I will get really fast upload speeds (above the 250Mbps SYS limit) and then I'll wget something else (over tcp from my ARM server to another server) and it'll be low. I'll have to play around with it some.
@sin said:
Hrrmmm this is weird, one moment I will get really fast upload speeds (above the 250Mbps SYS limit) and then I'll wget something else (over tcp from my ARM server to another server) and it'll be low. I'll have to play around with it some.
Could you test the CPU too?
curl -LsO git.io/bench.sh; chmod +x bench.sh && ./bench.sh -a share
@mtsbatalha said:
Seems that BHS ones is not capped at 250Mbps.
Download speed is not capped on SYS, the upload speed is.
other iperf servers show something similar. seems like there is some traffic shaping happening on outgoing connections?
I am going to do some real world tests later with ssh/wget.
geekbench sadly does not work on ARM but the cpu seems a bit sluggish while doing tasks like decompressing or compiling, but hey - it's only about storage ;-)
Falzo said: seems like there is some traffic shaping happening on outgoing connections?
Mine was going really fast for uploads and then dropped hard, then for a couple of uploads it'll be back to normal speed, and then goes right back to low speeds. I'm not sure what's going on...going to try it again a little later.
Got a drive with only 1621 poweron hours though
What I really want is a E3-SSD-1-16 or E3-SAT-1-16 though just waiting on those new prices...
Yup, I'm gonna try the upload speeds again later on and see if it's fixed. Every now and then my upload speeds will shoot up to normal speed. If the upload speeds would stay normal I would be loving this little box because 2TB for cheap is sweet.
Just out of curiosity, can one of you post /proc/cpuinfo for these ARM CPU's?
Also, it'll be nice to get one of those standard benches that lists out the usual suspects of sha256 hashing (500MB), bzip2 compression (500MB) and openssl aes encryption (500MB) and what is the BogoMIPS on these.
AFAIK there's no "native" AES support on these (but the N2800's worked fine for backups and they too don't have AES support) and considering that these machines are 1Gbps download/250Mbps upload it's definitely better for backups than the KS Atoms (barring the fact that these are non-x86 + 32 bit only cpus).
@nullnothere said:
Just out of curiosity, can one of you post /proc/cpuinfo for these ARM CPU's?
Also, it'll be nice to get one of those standard benches that lists out the usual suspects of sha256 hashing (500MB), bzip2 compression (500MB) and openssl aes encryption (500MB) and what is the BogoMIPS on these.
AFAIK there's no "native" AES support on these (but the N2800's worked fine for backups and they too don't have AES support) and considering that these machines are 1Gbps download/250Mbps upload it's definitely better for backups than the KS Atoms (barring the fact that these are non-x86 + 32 bit only cpus).
I'm very tempted to try them out.
cat /proc/cpuinfo
processor : 0
model name : ARMv7 Processor rev 1 (v7l)
BogoMIPS : 50.00
Features : half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpd32
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x4
CPU part : 0xc09
CPU revision : 1
processor : 1
model name : ARMv7 Processor rev 1 (v7l)
BogoMIPS : 50.00
Features : half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpd32
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x4
CPU part : 0xc09
CPU revision : 1
Hardware : Marvell Armada 375 (Device Tree)
Revision : 0000
Serial : 0000000000000000
obviously a weird strategy on limiting per connection, but somehow not every connection - can't tell how their traffic shaping works or determines when to limit or not.
for once I'll keep the box. that storage is really cheap to drop off things I probably never gonna need again. I am going to only store bigger backup archives (multiple GB) , which I can retrieve via axel or similar like shown above quite speedy if I ever need something again ;-)
The source actually was their API, I found new product codes on their API a few days ago.
I was looking for old prices of Opteron, but since it's not available on their website anymore I went to check.ovh and checkservers.ovh if they still have old servers listed. Apparently checkservers.ovh has the new prices listed too, so I just picked their website as source.
There are another 3 new product codes on their API, 1804sysgame04, 1804sysgame05 and 1804sysgame06. They could be the new game servers.
Falzo said: their pre-config on debian 9 works like a charm
Well that's beside the point, get it working then try speedtests. What is interesting here is if the weird speed shaping applies over IPv6 too. My guess is that it likely doesn't.
Comments
how is the outgoing bw? just wondering if they limit connections that leave the OVH network.
Gonna try more locations
I think that's BS, no mention of source/destination and such. just some iperf numbers, not even the command that has been run.
bench from @sin above also suggests otherwise. if the connection is not 250Mbps as advertised I definitely drop it. should be able to at least burst to it within europe, won't wait days if I ever need to recover some GB ... but I'll checl and see in a few anyway.
@sin (and @Falzo) try
use any value between 5200 and 5209 if one says it's busy.
From the speedtest you ran, it seems yours is in Canada. Maybe only FR ones are affected?
I hope that You are right, because those look like nice backup rsync servers.
From the speedtest you ran, it seems yours is in Canada. Maybe only FR ones are affected?
That's weird, I'm getting around 100-150Mbits uploading to various servers but when I tried your online.net link it shows this:
Just got 61.8MB/s downloading a 512MB file from my SYS ARM to my Ramnode NY VPS
could you test aria2/rtorrent too and download the most recent ubuntu/debian iso over bitttorrent. nothing dodgy just want the numbers.
Hrrmmm this is weird, one moment I will get really fast upload speeds (above the 250Mbps SYS limit) and then I'll wget something else (over tcp from my ARM server to another server) and it'll be low. I'll have to play around with it some.
Seems that BHS ones is not capped at 250Mbps.
I believe SYS isn't capped for download, just upload I think?
only outgoing is capped at 250Mbps at SyS, incoming is no cap 1Gbps
Could you test the CPU too?
curl -LsO git.io/bench.sh; chmod +x bench.sh && ./bench.sh -a share
Download speed is not capped on SYS, the upload speed is.
Additional IPs not available for ARM servers, despite being on SYS.
here we go:
nice fresh disk.
and then there is this:
so indeed the results that have been posted on that polish board.
BUT, let's add multiple connections:
so wtf? ...
other iperf servers show something similar. seems like there is some traffic shaping happening on outgoing connections?
I am going to do some real world tests later with ssh/wget.
geekbench sadly does not work on ARM but the cpu seems a bit sluggish while doing tasks like decompressing or compiling, but hey - it's only about storage ;-)
PS: french location
Mine was going really fast for uploads and then dropped hard, then for a couple of uploads it'll be back to normal speed, and then goes right back to low speeds. I'm not sure what's going on...going to try it again a little later.
Got a drive with only 1621 poweron hours though
What I really want is a E3-SSD-1-16 or E3-SAT-1-16 though
just waiting on those new prices...
added SYS to metadedi, well SYS delivered, storage prices from Hetzner undercut by 1EUR.
download speeds look good to me:
at least for europe that's easily 250Mbps and more
Yup, I'm gonna try the upload speeds again later on and see if it's fixed. Every now and then my upload speeds will shoot up to normal speed. If the upload speeds would stay normal I would be loving this little box because 2TB for cheap is sweet.
Just out of curiosity, can one of you post /proc/cpuinfo for these ARM CPU's?
Also, it'll be nice to get one of those standard benches that lists out the usual suspects of sha256 hashing (500MB), bzip2 compression (500MB) and openssl aes encryption (500MB) and what is the BogoMIPS on these.
AFAIK there's no "native" AES support on these (but the N2800's worked fine for backups and they too don't have AES support) and considering that these machines are 1Gbps download/250Mbps upload it's definitely better for backups than the KS Atoms (barring the fact that these are non-x86 + 32 bit only cpus).
I'm very tempted to try them out.
outgoing they obviously have some weird per connection-limiting in place. I put up a test-file and tried pulling it via wget to a hetzner box:
indeed limited to 5Mbit/s !!
now let's do the same with axel and 10 connections:
ahem.
obviously a weird strategy on limiting per connection, but somehow not every connection - can't tell how their traffic shaping works or determines when to limit or not.
for once I'll keep the box. that storage is really cheap to drop off things I probably never gonna need again. I am going to only store bigger backup archives (multiple GB) , which I can retrieve via axel or similar like shown above quite speedy if I ever need something again ;-)
ooh nice. Ovh's Rent-a-Hitachi drive.
@Falzo @sin try over IPv6
their pre-config on debian 9 works like a charm:
Ovh SoYouStart #SYS Coming next days:
- same as today: first 16 IPv4 0e/mo - 16+ IPv4: 1e/mo (was 2e/mo)
- minimum HDD: 4x 2TB (was 2x 2TB)
- minimum SSD: 2x 480GB (was 2x 240GB)
So the offers posted above are invalid?
Whats even the source ?
That could be new configs.
The source actually was their API, I found new product codes on their API a few days ago.
I was looking for old prices of Opteron, but since it's not available on their website anymore I went to check.ovh and checkservers.ovh if they still have old servers listed. Apparently checkservers.ovh has the new prices listed too, so I just picked their website as source.
There are another 3 new product codes on their API, 1804sysgame04, 1804sysgame05 and 1804sysgame06. They could be the new game servers.
Well that's beside the point, get it working then try speedtests. What is interesting here is if the weird speed shaping applies over IPv6 too. My guess is that it likely doesn't.
These new "storage" offers make the KS1/2/3 look quite dull...