WAAS Scaling Questions

[jheckart]

Hello,

I've been testing with 2 512-1's across a T1 for about 3 weeks with wccp and version 4.0.11b34. I haven't had a single problem yet, and have seen excellent results with MAPI, CIFS, FTP, HTTP, and SQL.

For the final design, I am looking at 2 512-2's in the core, and 512-1's throughout 7 remote sites. I have the following questions based on what I am unable to test without full deployment.

1 - How do the data stores function? On the cores, is there a data store per site, meaning that what was cached for one site would need to be duplicated in another data store?

2 - How will the clustering of the 2 512-2's work out? Will I have duplication of data between the two? The bigger question is, when one dies, how does the single take over assuming both are at about 75% capacity?

3 - What happens when my TCP session limit is hit? Do connections simply pass through without optimization, or are connections dropped. Is the answer the same when the limit is hit for CIFS?

Thanks a lot.

[tblancha]

1.) Yes, the DRE contexts are per peer. So, the data that is sent out to peer1 and peer2 are separate and maintained twice in the core device.

2.) Based on the IP address, the router maintains buckets and will send traffic from a source to whichever WAE 'owns/registers' that bucket. If each is at 75%, then when one fails, the buckets failover to the other device(s). So, then that one device would be oversubscribed at 150% which takes us to the next item

3.) If the TCP session limit is hit or there are not resources available to optimize, then the WAE will make new sessions as passthru.

[jheckart]

tblancha,

Thanks for the response.

If I have a cluster of 2 in the core, does each of the 2 have a duplicate copy, or do they each have their own data?

When the CIFS traffic exceeds a session limit, does the WAE passthru, or drop. Competition is telling me that WAE's will drop this traffic.

Thanks.

[tblancha · Accepted solution]

Each device at the core would have their own data.

We would go passthru if resources are not available or hit hard limit. No dropping sessions in WAAS.