Telnet output speed

[grossmj]

If I remember correctly, before the Dynamips community releases the Console/Telnet output was in chunk of text and not line by line like it is now. Most likely a buffer flush was added somewhere in the code. Apparently some users miss this because it looks faster for them?! I think you don't really have this issue on fast computers.

Related post: http://forum.gns3.net/topic10338.html

[flaviojs]

This could be a completely unrelated issue. Reminds me of the 'router console freezing' issue, where the actual problem was that the router was not running at all. (there was no feedback in GNS3 unless it ordered it to stop)

The 'slow' consoles I've seen so far are one of these two scenarios:

1. A bad idlepc value, landing on a bad spot in the console handling method. Which made the router slow down a lot whenever there was console activity. (idlemax was changed from 1500 to 500, so it's more visible)
2. The real cpu near 100%, leaving the router(s) starved.

I need more information. What router model, image and idlepc value was being used? How busy was the cpu? How was the router configured? If this doesn't happen on fast computers, then it's probably one of the two scenarios.

[flaviojs]

On another note, vtty_flush uses fsync, which means that slowdowns in the vtty socket are also incurred by the calling (router?) thread.

[grossmj]

The real cpu near 100%, leaving the router(s) starved.

I do not have this problem but I think this is exactly what happened to others.

[flaviojs]

Hmm, I wonder if using sched_yield in the cpu execution cycle could improve this. =/ It can degrade performance so it needs testing, which means I need to be able to reproduce it first.