Specs needed for a new Foglight Install of a LARGE vSphere Environment

Greetings all--

New to Foglight for Virtualization Enterprise and we are trying to bring up a trial to see how well it performs in an environment our size.

We've been working with the Dell/Quest/Foglight reps but so far haven't been able to get very far in properly "tweaking" for usable performance; so, I'm reaching out to everyone here for some thoughts.

We have a fairly large vCenter with currently over 7400 VMs, 20+ DataCenters, 450ish hosts, spread across the country.  It's a 5.1Update1 environment and the vCenter and related services (VC, SQL, VUM, SSO, WebClient) are all Windows 2008R2 VMs on a dedicated 2-node ESXi cluster and all the vSphere VMs have all the horsepower in the world to run it.  vSphere itself is running fine.

For our Foglight install (The latest FfVE v7) consistes of 3 VMs.  An FMS server, a SQL server (2008R2 - Dedicated SQL not shared w/any other DBs), and a vSphere Agent server.  We have thrown all kinds of resources at this as well-- however the performance is soooo bad that it's just not usable and we haven't even present all 7k+ VMs... it's only at arount 5k of them.

All are Windows 2008R2 -

The vSphere Agent VM was configured for 4 vCPUs and 12GB RAM.  It is on the same network segment and same physical host as vCenter so it can run on the same vSwitch and not have excess network traffic between the two.  The disk is SAN attached IBM XIV with 8GB HBAs.

The SQL VM is 6 vCPUs and 12GB RAM.  Different host, different network segmant, and different SAN attached IBM XIV storage.  This SQL Server is ONLY running the FL DB.  The DB is locked at 8GB RAM, initial disk size it 250GB and there's plenty of disk.  The vDisk is THICK provisioned and latency within norms.

The FMS Server is on a dedicated host-- no other guests running on it (for troubleshooting purposes though it hasn't made any difference).  It has all the host CPUs all to itself, has 32GB RAM, the VM's memory is all locked/reserved.

For all three VMs, the vCPUs are not in contention, READY times are low, there is no memory sharing, balooning, or swapping, and disks are thick with low latency.

The vSphere Agent server's JVM is set to 8GB min/max.

The FMS Server JVM is set to 12GB min/max.

So you can see, we've given 'er everthing she's got but it is VERY VERY VERY SLOW.  To the point it's entirely unusable.

As a comparison-- VMware's vCenter Operations manager is a 2-VM vAPP appliance, has access to the same vSphere with all the same objects and it's running perfectly fine and smooth.

My gut tells me there's something not right in the JVMs but I don't know enough. 

PLEASE HELP!