Why Can The Network Be a Major Bottleneck in VMware Environments?
Network utilization in VMware environments is typically below 10%. Therefore the network impact on overall performance is frequently underestimated.
Misconception:"My network utilization is under 10% so Networking cannot be an issue."
Reality: Average utilization is misleading since network traffic is extremely spiky. Detailed statistics show that in most environments some network resources become a bottleneck sometimes. In all likelihood you have network issues that you do not see in vCenter but that are very real for your users.
Misconception:Let's look at the VM that creates the most network traffic.
Reality: The more VMs run on a shared v Sphere infrastructure the more you have to look at resource competition. Even best practice VMware users have no clear understanding of the network usage patterns of each VM. Therefore a VMs with similar patterns end up using the same network components with statistical significance. It is key to identify bottlenecks as they appear on physical interfaces and then trace the problem back to the root cause: competing VMs.
Misconception: "Nothing we can do. The networking guys have to look at this".
Analyzing usage patterns allows you to distribute network-intensive VMs across different ESX hosts and network interfaces. The result is better performance and more available capacity without as much as talking to the networking team. This is also true for FibreChannel HBAs and even more so for iSCSI or NFS datastores.



