by February 28, 2004 0 comments

If your 10/100 Mbps network is getting saturated because of the many applications running on it, then you have an option. Determine the segments generating the traffic and upgrade them to a higher bandwidth of 1000 Mbps, or 1 Gbps. Most commonly this is required, when running audio/visual or other network intensive applications. That’s where the Gigabit Ethernet switch like this one from Dax comes in. It’s a 16-port manageable 1000Base-T switch capable of providing 16 full duplex 1 Gbps LAN connections.

The front panel of the switch has 16 RJ-45 ports with LEDs indicating active status of link, link speed, duplex mode and any collision or error. The LEDs are very useful in knowing the status of a particular port. To manage the switch you have to use the serial console port with a terminal emulation program like HyperTerminal, in Windows. It does not provide management capabilities through Web browser or other software applications. This is a slight problem as the serial port console based management is bit tougher than Web or software-based management, especially when you have to manage several switches together. Also now most new switches come with Web or software-based management options or telnet at least.

But the switch offers many desirable features such as VLAN setup, port mirroring, port aggregation, and QoS (Quality of Service) priority queues based on information taken from network layer 2 to 4. But the switch, doesn’t support the Spanning Tree Protocol, which is used to avoid loops in networks that often lead to broadcast storms. It does however, have built-in broadcast storm protection functionality.

Coming to performance, we tested the switch with a three-nodes setup with two systems acting as the data transfer/receive endpoints and the third as the test console. We used NetIQ Endpoint and IxChariot to determine the throughput. NetIQ is used to test throughput and response time between the nodes, and the IxChariot console is used to design and run the tests between the two endpoints and obtain results. The systems used were an IBM Dual processor Xeon server, Intel 865GBF motherboard based P4 3.2 GHz system, and Intel 845PEBT2 motherboard based P4 3.06 system. All systems were equipped with full-duplex Gigabit Ethernet LAN connections.

First, we measured the data throughput/transfer rates with only one endpoint sending data to the second endpoint and in the second test both endpoints were sending data to each other. In the first case, we were able to get maximum unidirectional throughput of 1000 mbps with an average of about 997 mbps. In the second case, the combined bi-directional throughput came to about 1500 mbps, as against the switch’s rated 2 Gbps. We feel that it would be able to achieve its rated throughput as well, if more powerful nodes are used.

Anoop Mangla

No Comments so far

Jump into a conversation

No Comments Yet!

You can be the one to start a conversation.