Forum Discussion
NimbostratusBIG-IP LTM for Exchange 2010 HUB/CAS load balancing (Single vs Dual NIC)
We plan to use BIG-IP LTM device for load balancing our 2 Exchange HUB/CAS (both roles on same server) servers. If we use single NIC on HUB/CAS servers, its DG will be BIG-IP LTM device IP. this would mean all communication From/To HUC/CAS would be via BIG-IP device.
Our concern is even traffic between HUC/CAS <--> Mailbox servers would be via LTM device. All servers are virtual and sitting on same hardware, behind single virtual switch
Our solution is use 2 NIC on HUB/CAS - 1) IP address from BIG-IP LTM segment for all user communication with DG as NIG-IP LTM device 2) IP address from normal production server farm segment (same as mailbox servers) for HB/CAS <--> Mailbox communication. No DG (Of course static routes on HUB/CAS to manage traffic for second NIC)
Question - Any downside for this setup OR any better solution?
3 Replies
Juerg_WiesmannHi Sach I do not see a reason, Traffic between Hub/CAS Servers would go accross BIG-IP, when they reside on the same Layer 2 Network. 2nd NIC would work, but might Require modifications on the Routing. Some Customers are using a 2nd NIC for Maintenance / Backup Traffic.
What_Lies_Bene1Cirrostratus
There's no requirement to have that second NIC but people obviously like having it separated. Anyway, with a single NIC it's just about using static routes for traffic you don't want to go through the default gateway and F5.
Hello Sach2910-
Presumably your CAS/Hub and mailbox servers are on different subnets; if they're on the same one, obviously, the default gateway (or static routes) doesn't matter for inter-server/inter-role communication, since all communication is local. That probably should go without saying but I wanted to put it out there just in case.
Once that question is out of the way, here's one important caveat: by default, a BIG-IP does NOT route traffic. So just defining one of a BIG-IP's self IP addresses as a default gateway isn't enough to enable inter-subnet communication. There are a couple of ways to do this; by far the most-preferred is to create two 'Forwarding (IP)' virtual servers (one per subnet), of type 'Network', and 'All' selected for Service Ports, then specify the destination subnet and mask in each, with no Source Address Translation (SNAT) enabled. BIG-IP will operate like a standard L3 router for those subnets, assuming it has an interface on each. [You can restrict traffic sources in the virtual server configuration if you want to limit the overall routing to just those two source networks, or even specific hosts.]
Please let us know if you have further questions.
