[load balancing] nodes causing lag

From: Chuck <vegan-lbIZZATwancom.com>
Date: Mon May 02 2005 - 12:04:03 EDT

Doing some testing this weekend on a new 6400 (9.05) we saw some
application delays (about 400-500 ms). Have a plain vanilla VIP with
about 45 nodes (least connections member, no persistence, not
mirroring, etc). Doing a pretty low BW/low user test I saw the delay
(600 users 10Mbits). I dont think this is unique to the 6400 - we
previously ran on 5100 - we have basically always seen this lag in our
application - now we think we may know where it came from.

Changed setup to use just one node and the delay went away (under same
user/BW load). Through trial and error saw that as the # nodes
increased, my delay increased. Jumped through numerous hoops - plugged
nodes into BIGIP switch, used primary then secondary cat, added virutal
interfaces, added ports, removed ports, had all nodes on one box,
spread all nodes to 12 different boxes, went through fw, removed fw,
turned off ssl, etc, etc ... In each permutation always the same - the
more nodes - the more delay.

Anyone have any thoughts on this? In my wildest imagination (and let
me tell you it gets pretty wild) I cant imagine that simple LB to 45
nodes introduces ANY latency - and especially not the 400 ms I am
seeing. However - every test I have done points to that being the
culprit.

Open to suggestions for further tests, alternative conclusions, etc -
would like to hear about anyone elses experience with pool-size, etc

Not sure this matters - but our setup is RMI, with WebLogic
____________________
The Load Balancing Mailing List
Unsubscribe: mailto:majordomo@vegan.net?body=unsubscribe%20lb-l
Archive: http://vegan.net/lb/archive
LBDigest: http://lbdigest.com
MRTG with SLB: http://vegan.net/MRTG
Hosted by: http://www.tokkisystems.com
Received on Mon May 2 13:19:34 2005

This archive was generated by hypermail 2.1.8 : Mon May 02 2005 - 13:44:13 EDT