Saturday, March 21, 2009

Dynamips proccessor utilisation

As I mentioned before, I have started working on IE work books. However, as soon as I ran the complete topology (no IGP, BGPetc) , with minimal configuration on non backbone routers, CPU utilization on an average was 55% and 30% for the two dynamips processes.

Yesterday I decided to give a test once again with different idlepc values, and voila it has reduced CPU usages to 14% and 12% for the two dynamips processes respectively.

1) Rather than using only idlepc values, which dynamips states is a potentially better idlepc value (* entries), test with all the idlepc values (including un* values), one after the other to see what works best for you.
2) Remember if you have started a router, then simply telnet to it and leave at that #prompt, even if you're not configuring it.

How to find the idlepc value?
Start one router (eg:R1), get a list of possible idlepc values. Apply them one by one and observe the cpu utilisation with top or htop command in linux. With one router and a good idlepc value, on an average my cpu utilisation is approx 2% or less.

I tried with one of the idlepc values, which was not * and it happened to be the value which I found is
the best rather than dynamips * ones. I had been using only * values assuming it was the best, ignoring un * entries.

Note: I run Dynamips on Ubuntu 64bit 8.10, with Quad Core 9400 with 4GB Ram.
Tip: For IE Vol1 labs, only start devices that you require in the lab exercise.

No comments: