3

We would like to (continuously) write to a log file on one linux machine, and continuously read from it on another machine. We're looking for low latency updates (this is for real-time plotting) and easy setup. What the best way to achieve this?

I have tinkered with nfs, sshfs, and tail -f, but updates only come in at 1Hz. I assume these filesystems don't support the select syscall, and so top polls at it's internal rate.

Andrew Wagner
  • 22,677
  • 21
  • 86
  • 100

1 Answers1

0

This smells like a bad architecture choice.

Is it for RPC? If so, why don't you just create a webservice?

If it's just for transfering data, try memcached.

Edson Medina
  • 9,862
  • 3
  • 40
  • 51
  • We have a robot using orocos that generates streams of data. Orocos has built-in support for dumping the streams out to files. We also use KST2, a graphing application that can do live plotting of data in a file. Latency is good locally, but if the file is being shared over the network performance is bad, partly because KST has to poll for changes to the file instead of using the select syscall. memcached looks interesting, but I'm looking for something that will just work transparently with the tools we already have, and still be fast. – Andrew Wagner Aug 22 '12 at 10:02