Re: how to achieve URL hash LB algorithm on ZXTM

Indeed an interesting thread J

 

Just wanted to comment that another approach, which - based on my experience - would nicely fit in some cases, is to use a URL-learning algorithm, which basically means that when a request to a URL arrives, the ADC will select the target server based on the stored persistent data (if exists); otherwise – in case the URL is a new one – the LB algorithm will be applied according to the configured criteria (e.g. least # of users, fastest response time, etc.).  The good side is that it can provide an even traffic distribution with minimal impact in cases servers are removed or becoming unavailable.  The only drawback of this approach is that the persistency table might get pretty large in some scenarios (e.g. if you deploy complex applications…).

 

Btw, Radware AppDirector also supports URL hash mapping (in fact, hash mapping based on any layer 7 parameters) which can be applied out-of-the-box (with no scripting involved), and also the URL-learning algorithm described above.

 

Nir Ilani

Director, Advanced ADC, Radware

 

From: [email protected] [mailto:[email protected]] On Behalf Of Eric Y. Zhang
Sent: Friday, August 07, 2009 5:03 AM
To: Load Balancing Mailing List
Subject: Re: [load balancing] how to achieve URL hash LB algorithm on ZXTM

 

thanks to all of you who discuss this. I learned to many from your threads. this is the most helpful mailing list I ever been.

I will try to figure out the best way to spread load/maintain the availability.

thanks again


Eric

On Thu, Aug 6, 2009 at 11:56 PM, Ben Mansell <[email protected]> wrote:

Yes, there's an irony in having two separate features that are in effect opposites of each other :)

Going back to the original problem... there may be another solution. Squid supports several kinds of inter-cache communication - ICP, HTCP and CARP. It may be helpful to try setting up some of these to spread the workload and direct requests to the server that has the freshest content. This in combination with some kind of intelligent balancing might be the best solution for this use case.



On 06/08/2009 15:27, Kenneth Salchow wrote:

And, in case it wasn't clear, that is in no way a ZXTM issue.  I've always
espoused that persistence is the opposite of load-balancing and as Ben
points out--the two are rather discordant.  Your results are likely to be
fairly similar with any device on the market given your incongruous goals.

_______________________________________________
lb-l mailing list
[email protected]
http://vegan.net/mailman/listinfo/lb-l
Searchable Archive: http://vegan.net/lb/archive
http://lbdigest.com Load Balancing Digest
http://lbwiki.com Load Balancing Wiki




-- 
Life is hard

_______________________________________________
lb-l mailing list
[email protected]
http://vegan.net/mailman/listinfo/lb-l
Searchable Archive: http://vegan.net/lb/archive
http://lbdigest.com Load Balancing Digest
http://lbwiki.com Load Balancing Wiki