Reaching out about Jackrabbit performance and scalability...

[Robert Botzer <Ro...@axispointhealth.com>]

Hi...

I am looking for any thoughts and info about how much volume a single instance of a Jackrabbit server can handle...
Especially if anyone has experience with using it as a component of a large web based
(and if there is a way and how to measure / predict that..)

Here is some background on what I'm looking for...

We are currently using a  Jackrabbit server and content repository as a component  of our web application solution.  ( I believe it is  2.2.13 )
We are storing simple text documents like 1 page notes,  or 1- 5 page letters.
We are using simple file system storage, and we store a reference link to the document in our relational database.
We DO NOT do any searching against these documents.  (So we have turned that indexing off..)
So basically we upload a small document,  keep and store the reference to it, and retrieve the doc based on that.
We also use a hashing algorithm to distribute documents to separate folders so we do not run into file system performance issues
(like 20,000 files in one directory).

We can scale our web application server 'out' by having multiple instances behind a load balancer...
But all of those instances currently use the same (singular) Jackrabbit server / repository.
( btw: The file system repository is on the same server as the Tomcat/Jackrabbit instance - so no network access for Jackrabbit  to access the docs.)

We are having good results with 3  of our application servers - for   ~ 600 users,  using a single Jackrabbit instance...

However, we have a new prospect/client that is asking for  ~ 3000 users...  So we could be deploying  something like 30+ servers  each connecting to Jackrabbit

I think that ultimately the actual Jackrabbit load will likely  depend on how many documents per minute  ( or some unit of time).
and also the capacity would be limited by the amount of storage available.
The  Prospect/Client ( of course) has no idea how to answer the documents / minute ( up or down question)...

Rather than start changing our product to use multiple JCR's  ( WEBDAV  Stuff ?  - we know little about)...
We would be looking to 'Scale UP' our Jackrabbit server..

So I'm trying to get a handle on exactly what that means, and how far we can push it..

Here are my thoughts...

1) We'll need a lot of disk (for document storage) and we can actually grow that capacity as time goes on.
2) If the server is dedicated only running Tomcat and Jackrabbit an abundance of memory would NOT help significantly  ( Assuming we deploy with 8Gb today)
3) Network bandwidth could be a limiting factor.


Any advice, guidance, experience, metrics etc.   in this direction would be appreciated...

(Would rather not take the 'well let's just try it and see what happens' approach..)

Thanks.

-Bob