WAAS - Any large deployments?

clausonna · ‎02-19-2008

Hi folks.

We're looking at potentially doing an 80+ site WAAS deployment and I was wondering if anyone out there had already gone down this path. We'd have about 70 branch office sites with 512-2's, 8 big regional offices with 612's, and two 7341's at the corporate DC's.

Branch offices connect to the WAN with 2821's and single (or sometimes dual) T1's. Regional branches have 2821's and two or three data T1's, and the DC's have 3845's with 45Mbps DS3's.

There are file servers everywhere, with no expectation of replacing/removing any (yet), and -plenty- of cross-WAN any-to-any file server access with daily multi-megabyte/gigabyte) data transfers.

My primary concern is scalability, especially with regards to the per-peer data caches and how quickly data gets flushed out. Overall manageability and ability to support the solution are also a concern.

Any and all comments would be greatly appreciated.

Many thanks!

Zach Seils · ‎02-20-2008

Neil,

We have done many deployments this size and larger. A few things to take under consideration:

- For any-to-any file services acceleration, you'll need to run both the Core and Edge services on each WAE. This configuration requires a minimum of 2GB of memory in each WAE.

- Spend the time up front to understand the traffic flows, especially between branch offices (if it exists). This will assist you with the WAE cache sizing.

- For manageability, all of your device configuration can be managed from the Central Manager. We have also improved the CLI setup script in 4.0.15 to help ease deployments.

Regards,

Zach

billmo · ‎02-22-2008

Neil,

One other thing to keep in mind as you look to scale to that size will be the fan-out numbers for the smaller devices.

There were some marked improvments in 4.0.15 for DRE fan-out but the CIFS fan-out limitations have not changed much. You may find that you'll need to upgrade to larger devices to accomodate a fully meshed CIFS caching network.

Also try to take some time to understand the connectivity directives you'll have to configure for each site. As we look toward a large deployment we have made some hard decisions not to create a fully meshed network mostly based on the operational overhead of setting up these connectity directives.

- Bill