Network performance are terrible and worsened over the last time...
I have a linode in newark and I'm connecting to it using a gigabit ethernet from Milan.
My gigabit has no limitations and gives me the full bandwidth in Europe,
when I surf the America the average download speed is lowered at half a gigabit.
So, my connection in the worst case is able to download a filefrom the America at half a gigabit.
If I download a file from my Linode in newark I download it at 100KByte/sec…
What's happening to linode?
20 Replies
caker
Linode Staff
Nothing. We've only been increasing our network capacity, adding more peering and transit and working on building out our own backbone across the planet.
You are going through the Internet to get to your Linodes in Newark. Sadly, we do not control the entire Internet, so you are going to be at the mercy of the lines between you and us.
I suggest debugging this with some MTRs in both directions.
Hope that helps,
-Chris
@caker:
What's happening to linode?
Nothing. We've only been increasing our network capacity, adding more peering and transit and working on building out our own backbone across the planet.
You are going through the Internet to get to your Linodes in Newark. Sadly, we do not control the entire Internet, so you are going to be at the mercy of the lines between you and us.
I suggest debugging this with some MTRs in both directions.
Hope that helps,
-Chris
Hi, thanks for the fast answer.
I see nothing strange in my MTR.
Are you sure that the problems is "the internet" and not your infrastructure?
I don't know if this is an info you want to share but what is your average network load on your network equipments?
@sblantipodi:
@caker:What's happening to linode?
Nothing. We've only been increasing our network capacity, adding more peering and transit and working on building out our own backbone across the planet.
You are going through the Internet to get to your Linodes in Newark. Sadly, we do not control the entire Internet, so you are going to be at the mercy of the lines between you and us.
I suggest debugging this with some MTRs in both directions.
Hope that helps,
-Chris
Hi, thanks for the fast answer.
I see nothing strange in my MTR.
Are you sure that the problems is "the internet" and not your infrastructure?
I don't know if this is an info you want to share but what is your average network load on your network equipments?
it's not a complaint, my money is well spent here
it's just curiosity.
The download speed from my node is about 100KByte/sec.
I have a gigabit and I can download at about 50MByte/sec from other servers in newark.
I remember some months ago, probably a year or more ago,
that I was able to download files from my linode at various megabytes per seconds.
What's changed, since than? The internet?
My MTR show that the worst results comes from 173.255.239.7 to my linode IP.
173.255.239.7 is a linode device, so the worst results comes from the linode internals.
how this can be possible?
is linode exploding?
hey cake, please add some network device
Disclaimer: cake don't be defensive with me, I am a Linode fan and I spread the verb all over the world.
I am able to send and receive 200mbps from a Linode in our Newark DC to a server hosted in Milan by Seeweb.
linode03:~$ iperf -c 85.94.202.10 -d
–----------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
Client connecting to 85.94.202.10, TCP port 5001
TCP window size: 45.0 KByte (default)
[ 5] local 45.79.137.222 port 48146 connected with 85.94.202.10 port 5001
[ 4] local 45.79.137.222 port 5001 connected with 85.94.202.10 port 40646
[ ID] Interval Transfer Bandwidth
[ 5] 0.0-10.0 sec 211 MBytes 177 Mbits/sec
[ 4] 0.0-10.0 sec 243 MBytes 204 Mbits/sec
Please provide me with a source IP you are coming into our Newark datacenter from so i can test. Providing a file i can download would be optimal.
Thanks,
Dan
@dspataro:
Hello Dan from Linode here,
I am able to send and receive 200mbps from a Linode in our Newark DC to a server hosted in Milan by Seeweb.
linode03:~$ iperf -c 85.94.202.10 -d
–----------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
Client connecting to 85.94.202.10, TCP port 5001
TCP window size: 45.0 KByte (default)
[ 5] local 45.79.137.222 port 48146 connected with 85.94.202.10 port 5001
[ 4] local 45.79.137.222 port 5001 connected with 85.94.202.10 port 40646
[ ID] Interval Transfer Bandwidth
[ 5] 0.0-10.0 sec 211 MBytes 177 Mbits/sec
[ 4] 0.0-10.0 sec 243 MBytes 204 Mbits/sec
Please provide me with a source IP you are coming into our Newark datacenter from so i can test. Providing a file i can download would be optimal.
Thanks,
Dan
sent the ip via PVT
"The internet" is a very busy place…
mmm… no it can't be the internet…
I had many problems with some servers and I had to move them 3 times, before they became stable. I also run a database cluster and I had a lot of disconnects between the cluster nodes (even though they where communicating via the private 192.168 interface). Maybe you are having a similar issue?
@IfThenElse:
Do you need your server to be hosted in newark? Maybe performance would improve if it was moved to the frankfurt data centre?
I had many problems with some servers and I had to move them 3 times, before they became stable. I also run a database cluster and I had a lot of disconnects between the cluster nodes (even though they where communicating via the private 192.168 interface). Maybe you are having a similar issue?
the problem is why a server hosted in newark should give me a bandwidth of 50KB/sec
With my Galera cluster, I had daily disconnects and linode support had no clue what the problem was. Eventually they moved the servers 3-4 times (to different hardware nodes) until the disconnects went away (probably someone in the data centre corrected the problem eventually).
@IfThenElse:
Do you really care why? Just ask support to move your server to another hardware node.
With my Galera cluster, I had daily disconnects and linode support had no clue what the problem was. Eventually they moved the servers 3-4 times (to different hardware nodes) until the disconnects went away (probably someone in the data centre corrected the problem eventually).
what does it means to move to another node?
is it possible to ask this? will I have downtime? how this thing works?
thanks.
Just open a support ticket via the Linode Manager and ask for a transfer to another hardware node. There will be some downtime, but you can schedule the transfer to happen when you want, like during the night when you have little traffic. In my experience they take 10 or 30 minutes, but it depends on the size of your server.
I can downlod from London (to Italy) at 10 megabytes per seconds,
from Frankfurt (that is a lot nearer than london) at 4 megabytes per seconds,
from Newark at 60kbytes per seconds.
How can be possible that London is faster than Frankfurt?
How it is possible that I download from Newark datacenter at only 60 kilobytes/sec?
This is a speed test from a good working server in New York at 600 megabits/sec.
~~http://beta.speedtest.net/result/6575098002.png
I can download from a good server in New York at 600 megabytes/sec and from a Linode in Newark at 60kbytes/sec?
what's this crap?~~
@T-A:
Use traceroute to see what's between the end nodes.
I have used traceroute and MTR and I don't know why Linodes in newark are so slow but 60kB/sec is pretty untolerable.
As I repeat, I download from Linode at 60kB/sec and I download from a good server in NY at 75mB/sec, this is 150 times faster,
why Linode is so slow?
We were sending traffic to the ASN you are behind (Vodafone IT) though Cogent, I have changed the outbound route via GTT now.
I am using the NLNOG Ring to test speeds by doing a wget to a large file hosted on your Linode in Newark. I have seen some speed increases since i made the change when I wget from Milan.
Before 2017-09-05 15:55:12 (8.22 MB/s) - ‘Presskit-MortgageCalculatorPRO.zip.1’ saved [168014806/168014806]
After 2017-09-05 16:18:46 (12.7 MB/s) - ‘Presskit-MortgageCalculatorPRO.zip.4’ saved [168014806/168014806]
For reference here is a wget from our Frankfurt DC to your Linode in Newark
2017-09-05 15:59:01 (391 MB/s) - ‘Presskit-MortgageCalculatorPRO.zip’ saved [168014806/168014806]
Please see if the routing changes has made a difference for you.
@dspataro:
Dan from Linode back again,
We were sending traffic to the ASN you are behind (Vodafone IT) though Cogent, I have changed the outbound route via GTT now.
I am using the NLNOG Ring to test speeds by doing a wget to a large file hosted on your Linode in Newark. I have seen some speed increases since i made the change when I wget from Milan.
Before 2017-09-05 15:55:12 (8.22 MB/s) - ‘Presskit-MortgageCalculatorPRO.zip.1’ saved [168014806/168014806]
After 2017-09-05 16:18:46 (12.7 MB/s) - ‘Presskit-MortgageCalculatorPRO.zip.4’ saved [168014806/168014806]
For reference here is a wget from our Frankfurt DC to your Linode in Newark
2017-09-05 15:59:01 (391 MB/s) - ‘Presskit-MortgageCalculatorPRO.zip’ saved [168014806/168014806]
Please see if the routing changes has made a difference for you.
no, I download from 40kB/sec to 80kB/sec…
it's not a problem of my home connection, I work for a big telco operator, tomorrow I will test the speed from one of my servers.
in any case the problem is not from my linode to my pc, the problem is from newark linodes to my PC.
this files stored in newark:
has the same speed of my linode.
this file in Frankfurt does not have this problems:
it downloads at 5MByte per seconds that is way way way better, in any case, no of your server is brilliant when talking about network performance.
100MB-newark.bin 100%[======================================================================>] 100.00M 71.0MB/s in 1.4s
2017-09-05 20:49:22 (71.0 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
100MB-newark.bin 100%[======================================================================>] 100.00M 155MB/s in 0.6s
2017-09-05 20:50:44 (155 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
100MB-newark.bin 100%[======================================================================>] 100.00M 48.8MB/s in 2.0s
2017-09-05 21:07:03 (48.8 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
100MB-newark.bin 100%[======================================================================>] 100.00M 73.8MB/s in 1.4s
2017-09-05 21:07:42 (73.8 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
@dspataro:
I am seeing much different results when I wget the test file from different networks. Please provide me any SRC IP addresses you are getting slow speeds downloading the test file from. If you can provide me a shell that would be even better.
100MB-newark.bin 100%[======================================================================>] 100.00M 71.0MB/s in 1.4s
2017-09-05 20:49:22 (71.0 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
100MB-newark.bin 100%[======================================================================>] 100.00M 155MB/s in 0.6s
2017-09-05 20:50:44 (155 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
100MB-newark.bin 100%[======================================================================>] 100.00M 48.8MB/s in 2.0s
2017-09-05 21:07:03 (48.8 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
100MB-newark.bin 100%[======================================================================>] 100.00M 73.8MB/s in 1.4s
2017-09-05 21:07:42 (73.8 MB/s) - ‘100MB-newark.bin’ saved [104857600/104857600]
sent info via PVT, please answer on the thread with no private data.
What have you done?