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.
Does Disk IO Speed Affect Network Speeds
agoldenberg
Member, Host Rep
in Help
I have a server colocated in Chicago and I find the network speeds to be low considering it's on a gigabit port.
Any thoughts?
Comments
what site did you use to test?
It depends how you're measuring the disk IO. It can affect it but obviously depends on the numbers we are talking about, it is one of the less likely reasons though.
if you want to test just the network speeds, pipe wget output to /dev/null:
wget -O /dev/null http://cachefly.cachefly.net/100mb.test
MTR to verify packet loss / latency
Close but not quite there. They got me to do a test from an IP in their network and the speed was even worse. 45MB/s they're looking into it.
100%[==================================================================>] 104,857,600 75.0M/s in 1.3s
2013-06-08 19:01:03 (75.0 MB/s) - `/dev/null' saved [104857600/104857600]
Yes. Think about it - if your disks can only write at speeds of say, 50MB/s, then you'll only be able to download (note: writing the file) at up to 50MB/s (400Mbps).
That's why a lot of the time you'll only be able to pull 2-3Gbit on a 10Gbit link (FDC's dedicated servers for example).
3Gbit = 375MB/s = SATA2 (theoretical) speeds
6Gbit = 750MB/s = SATA3 (theoretical) speeds
Also do not forget that wget is using only single stream for downloading.You may need to optimize tcp parameters and or use multiple stream downloaders to be able to measure more accurately.You can always pipe to /dev/null to exclude I/O bottleneck.
http://fasterdata.es.net/host-tuning/linux/
http://axel.alioth.debian.org/
WHat would be the proper IO test to perform on the drive to figure out if it is affecting the network speed?
You can use dd to test your sequential write speed of your I/O.
dd if=/dev/zero of=test bs=64k count=16k conv=fdatasyncIf you want to output to /dev/null, then simply use the O option
There are a number of dd and ioping tests that are commonly accepted as a disk benchmarking around here. Disk shouldn't be much of a factor until you're fully saturating a 1 Gbit port and beyond.
Yum it's the drive speed.
16384+0 records in dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records out
1073741824 bytes (1.1 GB) copied, 12.545 s, 85.6 MB/s
You are on a shared 1Gb port. What is wrong with 75MB/s
And downloads are not necessarily going to be sequential writes, you have ram buffers too etc etc window sequence, source saturation.
So yes... disk IO 'can' affect network speed to some degree but this thread is full of wrong.
1000 mbit port on a software network bridge with unoptimised stacks, no information about the path or latency or source or destination port and switch saturation etc it is impossible to say, 75MB p/s is fine though.