Feb 02 2011

Load Balancing Health Monitoring Traffic

For the purpose of hardware sizing one of my customers asked me about the amount of bandwidth traffic that is required by Load Balancers in order to maintain a constant View health monitoring.

The most common types of health monitoring are ICMP and HTTP.

HTTP health monitors are optional in a VMware View deployment, however strongly recommended. The Load Balancer appliances issue a GET URL command to retrieve the index.html page from each connection broker in the pool.

When the Load Balancers probe a Security Server, the Security Server immediately performs a check against the Connection Server for which it is configured. This check between the Security Server and Connection Server  happens over the AJP13 (TCP 8009) connection.

The picture below (if you can read) demonstrate a WGET command issued against a connection broker – for VMware View 4.5 the index.html is 1.2K.



As an example, F5 recommend at least a 1:3 +1 ratio between the interval and the timeout (for example, the default setting has an interval of 5 and an timeout of 16).

For every hour there are:

720 iteration x 1.2K = 864K per View server to be pooled.

If your environment has 2 Connection Server, 1.7MB per hour, and 40.5Mbytes per day.


It’s common to run ICMP at the same time, therefore you should add another 32 bytes for each iteration if the interval is also 5.

As you can see, that’s minimal traffic even for a dial-up network; but I thought would be interesting to share. Please note that these numbers do not include any type of application publishing or protocol tunnelling required to allow external access to VMware View environment.

Check out these solutions certified by VMware:

F5 has one here http://www.f5.com/pdf/white-papers/dell-f5-vmware-view-wp.pdf

Cisco has one here http://www9.cisco.com/en/US/docs/solutions/Enterprise/Data_Center/vmware/cisco_VMwareView.pdf

Leave a Reply