Super User is a question and answer site for computer enthusiasts and power users. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm running Debian Squeeze on a machine based around a Jetway J7F5M1G2E-VDE-LF moherboard. This particular install was transferred from a different machine, but I've tried with a fresh install and it exhibits the same behaviour.

Using the onboard Via Velocity gigabit ethernet (VT6120/VT6121/VT6122 according to lspci), even SSH was sluggish and doing something as simple as running 'dmesg' in an SSH session would cause the session to hang. Given this, it was completely unusable for file-serving purposes (its main purpose) as even just running 'ls' in a directory on an NFS share would fail. Given the behaviour (i.e. that it seemed to be somewhat related to the volume of data) I suspected it to be MTU-related, but was unsure how to investigate further.

As a stopgap, I installed the network card from the machine this install had previously been running on (a Realtek RTL8169-based card) but as this board, being mini-ITX, only has one PCI slot which I'd intended to use for other purposes, this isn't ideal. SSH is now usable, as is NFS, although it exhibits a huge disparity between read and write speeds: although read speeds are reasonable (around 35MB/s), write speeds are a fraction of that (initially maxing out at around 2MB/s). Although limited by the previous machine's IO subsystem, this card was still able to achieve a steady 30MB/s in both directions via NFS.

Investigating this suggested that it could be related to use of the 'sync' mount option, so I set it to 'async' and the write speed roughly doubled (to around 5MB/s at best), but was still far short of both the read speed and even what the previous setup could achieve. Other reports suggested it could be NFS-related in general and that performance was as expected with Samba, so I experimented with Samba only to find that (using Debian's default config) write performance was roughly half of that initially achieved with NFS (~500KB/s)

However, just doing some copying on the machine itself, I've realised that it isn't necessarily network-related. The drives in use are a Toshiba MK2006GAL on the IDE interface and (currently) a single Western Digital WD2000EARX on one of the SATA interfaces and the only disk controller listed by lspci is 'VIA CX700/VX700 RAID Controller'. Although raw read speed for the SATA drive is about 80MB/s (as reported by hdparm as well as observed generally), I am currently attempting to copy a 30GB file (with the SATA drive as both source and destination); almost 12 hours after starting it is now only at 28GB.

So there are two issues I'm looking to resolve: the write speed issue and being able to revert to usng the onboard LAN interface. I'm currently at a loss and would appreciate any insight.

share|improve this question
+1, very well phrased question, very detailed and explained efforts and previous steps. If every first question was like that... – That Brazilian Guy Jan 15 '13 at 14:58

You must log in to answer this question.

Browse other questions tagged .