Scaling unified communications systems with high performance requires a load balancing solution. Whether you are a new load balancing customer, or have a lot of experience with load balancers, it’s important to always check the basics. In a recent deployment we discovered that even when using access vlans and no tagging, there are still some simple guidelines to follow to ensure peak performance. Before we made these changes, the customer had 2 million+ concurrent TCP connections and saw dropped packets were losing TCP retransmissions.
First of all, the latest Citrix NetScalers have two management interfaces, 0/1 and 0/2, and you should not plug these into the same VLAN broadcast domain otherwise loops and broadcast storms are inevitable – as with any networking device. Disable 0/2 and only use management interface 0/1. If you really want to use both management interfaces, then Link Aggregate them together.
If you are using multiple interfaces and multiple VLANs, be sure to create a Subnet IP Address (SNIP) for each VLAN and bind it to the VLAN, along with a VLAN ID, and this will limit the broadcast domains on that subnet/VLAN to it’s appropriate interface.
If you are still in doubt, run a diagnostic trace to view if the packets are coming and going out of the appropriate interfaces. Make sure no data is being sent through the management interface, because it is not going to be able to handle a large amount of connections.