Cisco Cisco TelePresence Video Communication Server Expressway
Appendix 13 – Cisco VCS and hardware load balancers in front of a bank of FEPs
Cisco VCS Deployment Guide: Microsoft Lync 2010 and Cisco VCS X6.1
Page 74 of 81
Appendix 13 – Cisco VCS and hardware load
balancers in front of a bank of FEPs
balancers in front of a bank of FEPs
Background
For Lync Server to scale to support large numbers of users, a pool of Front End Processors (FEPs)
can be created for the Server system. Each FEP is then run on a separate piece of physical hardware
so that the hardware resources of a single platform are no longer the limitation on call and IM
processing.
can be created for the Server system. Each FEP is then run on a separate piece of physical hardware
so that the hardware resources of a single platform are no longer the limitation on call and IM
processing.
So that endpoints (Lync clients, peer proxies etc) do not have to be individually configured to route
their traffic to specific Front End Processors, a Hardware Load Balancer (HLB) is used to share out
the traffic amongst the FEPs. The HLB provides a single virtual IP address for all of the FEPs.
their traffic to specific Front End Processors, a Hardware Load Balancer (HLB) is used to share out
the traffic amongst the FEPs. The HLB provides a single virtual IP address for all of the FEPs.
When a HLB is sent data (like a SIP message), it uses an algorithm to decide where to route that
message.
message.
Frequently source address routing is initially used, so that if a device has already communicated
with an FEP (within the recent past) any further traffic from that device will also be routed to that
same FEP.
with an FEP (within the recent past) any further traffic from that device will also be routed to that
same FEP.
If source address routing does not define which FEP to send the message to, then either round-
robin, or another more sophisticated algorithm that can tell the loading of the FEP will be used to
find an appropriate FEP to route this new communication to.
robin, or another more sophisticated algorithm that can tell the loading of the FEP will be used to
find an appropriate FEP to route this new communication to.
Some hardware load balancers have the ability to receive SIP traffic, and rather than routing it
based on source address can route it based on the SIP device it relates to (allowing load
balancing of SIP traffic from a proxy, like Cisco VCS).
based on source address can route it based on the SIP device it relates to (allowing load
balancing of SIP traffic from a proxy, like Cisco VCS).
The HLB will perform destination address NATing, meaning that messages addressed to the hardware
load balancer’s (virtual) IP address will be re-addressed to the required FEP.
load balancer’s (virtual) IP address will be re-addressed to the required FEP.
Many load balancers also perform source address NATing by default in order to get replies for outside
devices routed back via the load balancer, so that the outside device (Cisco VCS, for example) sees
responses coming from the single IP address that it knows to be “Lync Server”.
devices routed back via the load balancer, so that the outside device (Cisco VCS, for example) sees
responses coming from the single IP address that it knows to be “Lync Server”.
The content of any messaging is left unchanged.
Example infrastructure is shown below: