Exchange 2010 Hardware Load Balancing recommendations

[Pomegranate]

I am not so expert in this additional than also I am planning for an Exchange 2010 deployment. I desire to be familiar with if there are several recommendations for hardware load balancing apparatus to build a CAS Array (co-located by means of a DAG). I need some help here. If anyone is familiar with it then you can post your reply here.

[Sabastiaan]

As far as I be familiar with there is not several recognized recommendations for the substitute supporting hardware load balancing. Theoretically, there are two factors I determine to consider at the same time as selecting a hardware load balancer. First and perceptibly extremely important that the apparatus be supposed to be accomplished of handling the protocols used by Exchange CAS and HT boxes. And the additional one would the competence of the apparatus which be supposed to be generally checkered by means of the vendor or the apparatus manufacturer. For Exchange server 2007 CAS deployment or consumption, I have comprise and seen Big IP used as a hardware load balancer. Yet, is there several unambiguous reasons why you would similar to utilize hardware load balancer as an alternative of windows NLB or DNS round and robin.

[Visala28]

The Windows NLB cannot be used for the reason that it’s not sustained or maintained on the similar box as the mailbox server by means of dag. DNS Round Robin is not a success conscious, i.e. if the CAS server is not responding the DNS server determine to tranquil direct clients to this server. ISA by means of Web Publishing Load balancing is an alternative for they additional than has 2 caveats. To have comprise a responsibility tolerant surroundings I would necessitate to deploy a ISA (or TMG) Array. The entire the internal clients must in addition utilize the ISA Server for CAS have right of entry. Something I do not like extremely much.

[teenQ]

Thank you for the reply. I originate a Load Balancer from Baracude which seems to be reasonable: Modell 340. This Modell sustain and maintain Cookie Persistence which is a prerequisite. I determine to try to get a demo apparatus and try if it’s running by means of Exchange 2010 RTM. If there is a Cisco equipment ready, I determine to try this in addition. Might you share the parameters you tested the apparatus against. I am essentially seeking for the characteristics and supportability of the apparatus and how you tested them.

[Candace]

For your information, they determine to be documenting this almost immediately and releasing a white paper on parameters at a high level. The 3rd party vendors determine to then acquire our doc and modify it for their hardware. I do not encompass an ETA on when this determine to be ready additional than determine to post it almost immediately. Presently finished testing by means of a Barracuda Modell 340 and two CAS Servers in a DAG. Runs great! Might you probable send me a screenshot of your settings that you used for the Barracuda to get this to run. I am not capable to get the NLB running by means of our 2 node DAG. I have by no means used a NLB previous to and am actually struggling by means of it.

[Chandranath]

You be supposed to not identify a tcp port in keep a lived arrangement file in virtual_server segment as load balanced port. This determines to permit you to do LB on each tcp port, so RPC and additional protocols determine to be load balanced by defaulting. The Virtual server in maintain lived arrangement be supposed to be configured to utilize "lb_kind DR" (this determine to facilitate direct return) and "lb_algo RR". Direct return is extremely fast, that is why 1 active virtual machine is able to serve additional than 100 outlook clients. When the packet goes to VIP direct return algorithm immediately reinstates MAC in packet form virtual to real server's MAC and frontwards this packet to load balanced exchange server, all the communication back from exchange goes to client IP's directly bypassing keep a lived. On every exchange server you be supposed to facilitate weak send and weak receive functionality on networks cards that determine to be used for NLB (loop back adapter, and network adapter used for NLB). Docs on netsh command determine to assist you.

[Aloke]

My communications is extremely easy. Exchange servers holding the entire roles (CAS,HUB, Mailbox) and 2 virtual apparatus by means of keepalived which make available RPC and HTTP/HTTPS load balancing for substitute. keepalived serves virtual ip (VIP) which is organizing as client right of entry array in Exchange 2010. Each exchange 2010 has a loopback adapter by means of the similar VIP. The entire these sever are in comparable ip subnet. keepalived configured to run in active/passive scenario. It uses vrrp protocol to hold failover if energetic node goes down or became stalled. The second node goes up and handles the entire the requests for VIP. Yes they had tested the entire the scenarios when one of keepalived or exchange server node goes down: clear shutdown/reboot, replicated network errors, replicated power loss, out of free disk space and so on. The whole thing runs extremely well w/o any tribulations.