Re: Persitence Identifier

Absolutely true.

That's why I think, when using cookie learning, using a specific value per server is a far better option than per client, in the case of the creation of a specific cookie dedicated to persistance.

regards,

Surya


router# show problem | include solution
 

--- En date de : Jeu 16.7.09, Ed Toro <[email protected]> a écrit :

De: Ed Toro <[email protected]>
Objet: Re: [load balancing] Persitence Identifier
À: "Load Balancing Mailing List" <[email protected]>
Date: Jeudi 16 Juillet 2009, 19h37

Keep in mind that using these server-initiated per-client cookie persistence methods means that the ADC needs to keep state information on each and every client until the cookie is allowed to expire. That can create a lot of overhead.

When the ADC issues a persistence cookie itself, it's usually encoded in a way that tells it which server to send the request to but it doesn't need to track every single cookie for every single client.


Surya ARBY wrote:
> Hello Tony.
> 
> I guess the cookies generated on his apache services all have the same name but all different values in the payload, so by using cookie-learning the LB can recognize the right back end server to forward the request, in fact this technique is perfectly similar to a cookie insert where the payload of the cookie identifies the back end service to use except that it's purely passive from the LB (if the payload of the cookie generated by the apache is unique and stays local). So to have this method working, the payload of cookies generated must never be shared between two apache servers, in this case persistance would be broken.
> 
> Anand, using cookie insertion on the LB is a better method than cookie learning because
> 
> 1 - cookie learning consumes memory ressources (the LB has to build a table where learnt values are stored), with cookie insertion some vendors put in the payload the IP and the port of the back end service used so the result of the persistance process is predictable and deterministic (this is not the case with ALL vendors ok :) some also use different cookie values and also build a table, Cisco ACE works like that as I far as I know)
> 
> 2 - the persistance cookie inserted is usually stripped off automatically from the requests, it will not be the case with passive cookie learning and some apps will break when receiving unexpected cookies.
> 
> regards,
> 
> Surya
> *
> router# show problem | include solution*
> 
> 
> --- En date de : *Jeu 16.7.09, Tony Bourke /<[email protected]>/* a écrit :
> 
> 
>     De: Tony Bourke <[email protected]>
>     Objet: Re: [load balancing] Persitence Identifier
>     À: "Load Balancing Mailing List" <[email protected]>
>     Date: Jeudi 16 Juillet 2009, 18h24
> 
>     Hello Anand,
> 
>     Is there a particular reason you cannot use insert cookie mode?
> 
>     Apache generally doesn't issue cookies, rather an application
>     platform like PHP or Tomacat would (running on top of Apache).     The reason is that these application platforms will need to
>     generate a unique value for each user.  If you can get Apache to
>     generate a cookie, I'm not sure it would be unique.
> 
>     If you do want to use passive cookie mode (where the web server
>     generates the cookie), the cookie value you generate will need to
>     be unique to each user.
> 
>     The WSD 3000 should be capable (it's been a while since I've
>     logged into a Radware) of inserting its own cookie.  In your case,
>     that seems like it may be the best solution (I believe others have
>     already mentioned this).
> 
>     Regards,
> 
>     Tony
> 
> 
> 
> 
> 
>     On Jul 16, 2009, at 7:37 AM, anand kumar wrote:
> 
>     > Dear All,
>     >
>     > I have generated a cookie named "xyz " through my Apache Server
>     by giving Directives
>     > , but when i have connected the WSD 3000, i couldnt get the
>     cookie back,
>     > what i have to give in the cookie Key(Persistency identifier) on WSD
>     > --Anand.M
>     > HCL Infosystems Ltd
>     >
>     > _______________________________________________
>     > lb-l mailing list
>     > [email protected] </mc/[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
> 
>     _______________________________________________
>     lb-l mailing list
>     [email protected] </mc/[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
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> 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
>   

_______________________________________________
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

_______________________________________________
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