Microsoft KB Archive/96684

= Factors in Remote Bridge Performance =

Article ID: 96684

Article Last Modified on 7/30/2001



This article was previously published under Q96684



SUMMARY
The communications link between remote bridges has more effect on performance than the bridge's hardware design; therefore, local bridge statistics may not be applicable. (Remote bridges here are defined as interconnecting geographically separate LANs.) To measure performance, one needs to consider protocol and application overhead, speed of the physical link between bridges, and bridge processing, such as buffering, processing power used for managing data links, and internal bridge speed.

Measuring with Packets-Per-Second
Forwarding rates measuring packets per second can be an inaccurate way to measure a remote bridge's performance because the packet size can vary, from either Ethernet 802.3 minimum (64 bytes) to maximum packet size (1518), to Token Ring 802.5 (up to 17,952 bytes for 16mps card); or compare minimum Ethernet to Token Ring packets. What do you measure when the mix of packets coming through have variable sizes? Protocol differences between XNS, TCP/IP, and NetBEUI also account for varying amounts of data in each packet. A bridge may be passing all types of packets, but this discounts using packets-per-second as a measure.

Link Speed between Bridges
The combination of bridge buffers becoming full (thus waiting on the link to receive more data), and the link speed between two remote bridges being lower than the network data rate can lead to backups. Thus, even if the bridge can handle receiving packets by buffering all of them, without requiring re-transmissions at an upper level, the link is not being used efficiently. If the link speed is inadequate, filtering will have nothing to do with better throughput. Thus, it is better to measure end-to-end performance rather than the ability to handle back-to-back packets.

Dial-up lines and leased lines are common. T-1 operates at 1.544M bits per second, X.25 operates at an average speed of 19.2K per second, and Public Data Networks (PDNs) operate at up to 56K bits per second. Microwave links for short distances can get 10M bits per second (requires a &quot;line-of-sight&quot; between the transmission and reception end). Compare this speed with a 10M per second Ethernet network that seeks to send information across a remote bridge operating at 1.544m.

Bridge Error Correction
When the bridge handles error correction, performance will be better as the bridge (rather than end devices) check for errors that might lead to a request for retransmission.

Additional query words: 2.20 remote bridge link measure

Keywords: KB96684

-

[mailto:TECHNET@MICROSOFT.COM Send feedback to Microsoft]

© Microsoft Corporation. All rights reserved.