IDP node stickiness without an SLB?

Jim Fox fox at washington.edu
Thu Jan 23 14:15:28 EST 2014


[ Caveat:  I am in no way an expert at tuning either apache or tomcat. ]

I think that horizontal scaling is the most dynamic and efficacious way 
for any application to respond to increased demand.  I try to put that
capability into anything I build.  Maybe that's just me.

I want to be able to handle a thousaand or so simultaneous login
requests.  We get hit this way at times.  I could set Apache's
maxclients to a thousand, and tomcat's threads to a couple of
thousand.  Maybe that would work, but it seems like there's got to
be a limit somewhere.

On the other hand, I can cluster a bunch of idp hosts, each accepting
a few hundred simultaneous clients, and handle the same load.

I think I can go farther with the latter approach.  But that's
admittedly thinking and not knowing.

Jim

On Thu, 23 Jan 2014, Wessel, Keith wrote:

> Date: Thu, 23 Jan 2014 07:19:35 -0800
> From: "Wessel, Keith" <kwessel at illinois.edu>
> To: Shib Users <users at shibboleth.net>
> Reply-To: Shib Users <users at shibboleth.net>
> Subject: RE: IDP node stickiness without an SLB?
> 
> Interesting, Jim. Could you give a few more details on this to help us avoid this if our situation turns out to be similar? We're running in a VMWard environment and can easily throw more processors/memory at the system as well as bump up Apache max clients and Tomcat max threads.
>
> Keith
>
>
> -----Original Message-----
> From: users-bounces at shibboleth.net [mailto:users-bounces at shibboleth.net] On Behalf Of Jim Fox
> Sent: Wednesday, January 22, 2014 10:25 PM
> To: Shib Users
> Subject: Re: IDP node stickiness without an SLB?
>
>
> Our experience was that active/passive didn't scale very well.
>
> Jim
>
>
> On Jan 22, 2014, at 4:04 PM, "Cantor, Scott" <cantor.2 at osu.edu>
> wrote:
>
>> On 1/22/14, 5:51 PM, "Wessel, Keith" <kwessel at illinois.edu> wrote:
>>
>>> But all of this sure is making active/passive tempting.
>>
>> My next upgrade will be, no question about it.
>>
>> -- Scott
>>
>>
>> --
>> To unsubscribe from this list send an email to users-unsubscribe at shibboleth.net
>
> --
> To unsubscribe from this list send an email to users-unsubscribe at shibboleth.net
> --
> To unsubscribe from this list send an email to users-unsubscribe at shibboleth.net
>


More information about the users mailing list