We have 10 to 20 Tomcat web applications with hundreds of concurrent users per application on a server.
Each Tomcat application connects using its own user to a database on Microsoft SQL Server 2008 R2 server. This database contains loads of Stored Procedures that are executed by the Tomcat application. The Stored Procedures usually execute large joins between multiple tables (usually joins between 4-5 tables with each 100k+ records)
Currently this runs on 32-bit Windows 2008 R2 (Enterprise) with 32-bit Tomcat and 32-bit SQL Server 2008 R2. The Servers have 32 - 64GB of RAM on a multi-core processor.
My gut feeling tells me we should update to x64. This will future-proof and prevent some memory issues now that might be better handled in an x64 environment. I can't find benchmarks or proof for this though. Using flags like /AWE and /PAE feels like SQL Server is artificially trying to perform better than it would out of the box, whereas this artificial layer is not necessary in the x64 environment. I can't find any "proof" for this though.
Is it worth updating this configuration to x64? Will we see a performance gain? How much? Are there benchmarks supporting this?