Doug,
When you configure your WAE-674 to speak to your routers, you specify a router list. The routers Highest Loopback IP Addresses (if doing GRE Negotiated Return) or the directly connected Layer-2 adjacent interfaces (If doing IP Forwarded Return) are used. This will negotiate a MASK of address space to be distributed between both WAE boxes from your routers. Each routers WCCP setup will map a specific group of addresses to each WAE.
Setup on WAE
wccp router-list 1 10.0.0.254 10.0.0.253 (These are the IP address on your router)
wccp tcp-promiscuous router-list-num 1
wccp version 2
From here you have three real options:
#1 - Distribute load to both WAE boxes based on Source Address (WCCP Service 61 on WAN Interface of Data Center)
#2 - Distribute load to both WAE boxes based on Destination Address (WCCP Service 62 on WAN Interface of Data Center)
#3 - Create Custom WCCP Services and unique Access lists that send replication traffic to one WAE, all other traffic to the other WAE
Method#1
With a WAE 674-8GB, if you push less than 100-150GB per day, you would normally employ method#1, WCCP 61 on the WAN interface and have duplicate traffic types on both WAE devices based on client source address. This is great if one WAE dies the other WAE will generally have the same data cached. Watch your data loads and service actions (TFO/LZ/DRE) for different applications and make sure you are not poisoning your DRE space by sending uncachable content to DRE.
Method#2
With a WAE 674-8GB, if you push more than 150-200GB per day, you would lean toward Option#2. I am not sure how WAAS version 4.4.1's DRE caching architecture will change the game. If you determine you are over running the 300GB DRE partition every day or so, then you could swap gears and distribute unique traffic types to each WAE by using WCCP service group 62 on the Data Center WAN Interface. This will send traffic bound for Server-A to WAE#1 and traffic bound for Server-B to WAE#2. This would effectivly double your DRE Cache space, if required. You would also modify your WCCP MASK value to either alternate WAE boxes for every other destination address, or by every other subnet if you have duplicate servers in alternate subnets. I believe your destination mask would be 0x1 to alternate WAE devices based on 32-bit host addresses. Effectivly giving you 2 buckets. A WAE fault in this scenario would force all traffic to a single WAE and require the data to be cached as new content.
Method#3
Finally if you had FTP/Data Domain/ComVault/SnapMirror etc... replication traffic that was 300GB+ per day on a 674. You could use one of your WAE boxes as Replication Mode. You would create one access list for replication traffic types and bind tjhose to a custom WCCP service group. All other traffic would go on your regular WCCP access list and get bound to your regular WCCP 61 / 62 services pointed to the other WAE.
There use to be a WCCP Mask calculator on this link, but it appears to be broken. If I can find a local copy of it I will attach it to this thread.
https://communities.cisco.com/docs/DOC-9521
