Reply
Fry
cdhesc
Posts: 3
Registered: 07-09-2008
0

Negative Optimization?

Hi Bokad,

I had similar issues and these were resolved by upgrading RiOS, you did not mention what
version you are using but I believe that my problems were under Version 3.07.

Currently I am using version 5.01a but if you are running a version below Version 4.x
you must upgrade to V 4.x before upgrading to V5.
Administrator
Posts: 385
Registered: 02-28-2007
0

Negative Optimization?

Hi Bokad,

Looking at the screenshot, this looks like an HTTP application. I have seen scenarios where the read-ahead mechanism employed by the RiOS application streamlining module for HTTP is aggressive and can pull much more data than was was requested off the server and over the WAN. If the HTTP client never explicitly asks for the data that was predictively prefetched, it's counted against the WAN totals, but never registered on the LAN total. I have seen this scenario with some applications when you have newly installed Steelheads and are performing alot of cold operations.

Although you sometimes incur the increased WAN traffic for some of these traffic flows, there are really two positives that come from this:

1. Improved response times during the operations involved in this read-ahead traffic
2. The datastore gets pre-warmed with data form the server so future requests for the data get warm hits

At the end of the day, the result should be improved response time for the application in question. That is the fantastic benefit you get from the aggressive nature of the predictive, read-ahead nature of the application streamlining module.

Bob
Fry
bokad
Posts: 7
Registered: 09-19-2007
0

Negative Optimization?

[ Edited ]
I have some large connections going through my Riverbed that are experiencing a high degree of "negative optimization". ie. The connection is bigger on the WAN side than it was on the LAN! I can understand a small degree of inefficiency for packet overhead on connections where there is no redundant data but the bloat on these big losers is really wiping out my performance gains! For example:

LAN: 2.7MB WAN: 8.9MB Bloat: 229%
LAN: 5.5MB WAN: 11.7MB Bloat: 112%
LAN: 11.1MB WAN: 17.4MB Bloat: 57%
LAN: 13MB WAN: 15.6MB Bloat: 20%

Is anyone else seeing this? Potential Causes? Fixes?
Message Edited by bgilbert on 01-28-2009 05:22 PM
‬‪‬‪‬‪