powered by Jive Software

Two servers with one database ? How it works?

To have a scalable architecture, I would like to test this:

One database on server3.domain.com

One JM server: server1.domain.com configured with server3

One JM server: server2.domain.com configured with server3

However, it doesn’‘t work correctly, because these two servers can’'t shared the same DB… (configuration problems)

How can I do ? to use this two servers ?

Hi mlainne,

Currently Messenger doesn’‘t support what you’'re trying to do. The best option right now is to setup two (or more) Messenger installations and use s2s to distribute the load. How many users are you looking to support?

There is an effort under way called url=http://www.jivesoftware.org/community/thread.jspa?threadID=15824&tstart=0Pampero Project[/url] to allow Jive Messenger handle at least 100K connections, but this solution is a ways off.

Thanks,

Ryan

Thanks for this answer but what is s2s ?

We have planned to deploy his solution for 60 k users and Jive seems great.

Thanks for this answer but what is s2s ?

Sorry, I should have elaboraded on that more. s2s (short for server-to-server) is what allows servers to communicate to each other and in turn, allows users on different servers to communicate with each other. For instance, say your organization has two offices, you could setup each office with their own instance of Messenger and then link the two via s2s. If the number of users is evenly split between the two offices each Messenger instance would only have to handle half the load.

We have planned to deploy his solution for 60 k users

and Jive seems great.

Do you mean 60k concurrent users or registered users? If it’'s registered users, how many do you anticipate being online at any given time?

Thanks,

Ryan

Ok for s2s, however, we can imagine that a user uses always the same server however I see some problems:

  • All my users (+60k) are in the same LDAP.

  • My first server is server1 (second is server2)

  • Me I’'m connected on server1, my adress is me@server1

  • Bob is connected on server2, bob@server2

When I search bob in the search interface of pandion (for example), it finds bob, but bob@server1, so I have to correct this point because bob is on the server2.

When I connect my client on the server, I put me or bob and I have to fill in the client bob@server2 or me@server1.

And this configuration doesn’'t solve the scalability of the server if server1 craches for example.

We have planned to deploy the IM solution for 60 k, so 60 k concurrent- acces

When I connect my client on the server, I put me or

bob and I have to fill in the client bob@server2 or

me@server1.

Yes, searching for users across servers is an issue if users don’‘t know what server the user they’'re searching for resides on. It maybe possible to make some changes to search plugin to allow it to search across mulitple servers.

And this configuration doesn’'t solve the scalability

of the server if server1 craches for example.

Yes, you’‘re right it doesn’'t. If one of your Messenger instances we to go down there is no way to have another instance automatically pick up the load. The best (only?) thing to do would be to have a spare server ready to go should one go down.

We have planned to deploy the IM solution for 60 k,

so 60 k concurrent- acces

With the right hardware Messenger can handle approximately 5000 current users, so you’‘d be looking at needing at least half dozen servers to handle 60K concurrent users, plus a spare or two. Most likely you’‘d want to spread the load across 16 or more servers so they aren’'t running at 100% capacity all the time.

The two main Messenger developers, Matt and Gato, have more experience with large installations than I do. Maybe they can jump in here and offer up some ideas?

Thanks,

Ryan

Have you some benchs… What is typically the hardware for 5000 k, I mean if I use for example a sunfire 880 - 8 CPUs 2Ghz, with 16 Go RAM, how many users can I handle ?