Many network managers look to VDI to reduce PC hardware and ease desktop management. But not everything about VDI is so easy, and its complex parts can generate a lot of network traffic.
The network traffic created by virtual desktop infrastructure (VDI) can increase latency and reduce throughput. Before you travel down the VDI path, you need to determine the expected VDI network performance and map out network connections. It all comes down to measuring performance and then calculating the potential loads VDI could add. Network performance management tools can help you monitor network operations, includinglatency, throughput and network capacity.
The most important thing is to look for peaks and valleys in network usage, which will help you determine the maximum load your VDI network can handle. For example, network traffic may peak at 9 a.m. because of morning logon activity, when the network is saturated with authentication requests and application launches. In that case, adding VDI to the mix — where virtual machines (VMs) will be launched over the network — could slow network performance to a crawl. Also, remember that backup operations, quarterly processing and other event-driven occurrences can affect VDI network traffic and throughput.
Network issues determined by type of VDI
Once you determine how VDI could add to network traffic, you must figure out how to improve VDI network performance and address heavy loads. The type of VDI implementation you have comes into play when diagnosing and solving network performance issues.
Connected VDI. In some VDI implementations, all processing activity (the hosted desktop) takes place in the data center, also known as connected VDI. In those cases, VDI network traffic is usually broken down into two types: traffic created by provisioning (creating and launching) the virtual PC, and the traffic to and from the endpoint. The external network (the user to the VDI host) is very chatty, but with small amounts of data — mostly screen updates and key strokes, which are transmitted via a display protocol from the endpoint to the host and vice versa.
In VDI environments where processing occurs within the data center, you can segment the network and apply quality of service (QoS) controls to lessen traffic between the endpoint and the host. High-speed links between storage and servers will help to further minimize the performance impact of VDI.
Disconnected VDI. Another style of VDI is an environment that supports disconnected virtual sessions. In other words, the VM actually runs on the user’s endpoint PC. With this kind of VDI implementation, the VM is delivered to the endpoint and then launched locally. Delivery of the VM’s virtual hard drive file can consume a great deal of networkbandwidth. Luckily, that event only happens during initial provisioning. After that, VDI network traffic is equivalent to what a traditional PC uses (unless you need to replicate the virtual hard drive at any point).
Simply put, disconnected VDI is bandwidth-intensive only during provisioning and replication, which are usually infrequent events. That means server-to-storage bandwidth is not as critical, and you should instead focus on VDI network traffic between the endpoint and the host.
Preparing your network for VDI comes down to measurement, prototyping, deployment and monitoring. Network segmentation, virtual LAN construction, QoS policies and packet acceleration can address most VDI network performance issues as they arise.
ABOUT THE AUTHOR:
Frank Ohlhorst is an IT journalist who has also served as a network administrator and applications programmer before forming his own computer consulting firm.