EXPERT RESPONSE
To address your question, improvements in speed and performance of your LAN and WAN depend on several factors, budget and technology. With the LAN, I am going to recommend a few general guidelines and best practices on design to ensure reliable performance. For the WAN, there are several factors to consider that are more cost prohibitive. Decide what works best for your enterprise and environment.
To address potential performance issues on the LAN, you want to identify and isolate any bottlenecks due to technology. On LANs, you are able to achieve better performance results with high speed, low cost technologies like Ethernet. Ensure that all of your servers in the Citrix MetaFrame farm are capable of at least 100Mbps at full duplex. Oftentimes, I recommend that the switches be hard wired for this configuration (or higher if applicable) to reduce any potential problems with auto negotiation. A good LAN design also ensures that the right VLAN and Spanning Tree implementations are in place to reduce multicast and broadcast congestion and bridging loops. If not currently implemented, ensure that the servers are on a managed switch that supports multiple VLAN configurations and Spanning Tree. By implementing managed switches in the environment, the switches will be able to assist you with control over the server ports and reduce collisions. In summary, ensure that you have managed switches on the network that can control VLANs and bridging loops. Also ensure the server switch ports are set to the right speed and duplex settings. These are the two major rules to follow when developing better performance on the LAN side of the enterprise.
For WAN issues, Citrix MetaFrame clients are a good start. Citrix MetaFrame clients are quite effective when circuits are already heavily congested and you need to reduce the amount of traffic that the users are transferring over the WAN. Unfortunately, Citrix tends to have a performance hit when it comes to the end user experience. This is due to the fact that depending on the Citrix MetaFrame deployment you have implemented, the client when launched will open a single TCP session. With this single TCP session, the screen updates and refreshes are sent to the client's machine at the remote location. While this consumes less bandwidth, any sort of delay or dropped packet causes the end-user response times to degrade significantly. If there are any sources of delay within the network and a packet is lost, the TCP window for the single-session client will drop significantly. To compensate for this ensure that the links are not congested and special treatment (QoS) policies are given to Citrix clients to ensure packets are given the best performance. However, you must also be aware that by doing so, you effectively compromise the performance of other applications on the network. There are tools that allow you to troubleshoot and analyze the performance of WAN-deployed applications. If you would like more information on these tools, please let me know and I will happily supply that information.
|
