Message routing latency-minimizing method in an ASIC design for distributed cooperative communication protocol processing
✍ Scribed by Hiroyuki Yamashita; Toshihiko Suguri; Shingo Kinoshita; Yasushi Okada; Kiyoshi Oguri
- Publisher
- John Wiley and Sons
- Year
- 1998
- Tongue
- English
- Weight
- 501 KB
- Volume
- 29
- Category
- Article
- ISSN
- 0882-1666
No coin nor oath required. For personal study only.
✦ Synopsis
In the field of distributed systems, research proceeds on reducing communication overhead by using a ring-type high-speed network to interconnect nodes and to reduce the number of transmissions, and by migrating cooperative functions such as load distribution and fault tolerance from conventional applications to lower-layer communication protocols. To raise the overall performance of such distributed systems, message routing latency at each node must be reduced, and to this end, we have incorporated communication protocol processing in an LSI chip. This LSI uses the received message and system-status data stored on the LSI to perform input-message path control (decide whether to transfer the message to the host or relay it) and to update specific fields in the message to be relayed. In the conventional method, processing of necessary message fields began after they had all been received and placed into a buffer. In this LSI, however, a set of modules are activated in sequence as fields that require various types of processing are received. This scheme is achieved by adopting a datadriven-type distributed control system and implementing all functions through wired logic control. As a result, message updating and relay latency has been minimized to 2.55 ms. The LSI has been achieved by a gate array of about 400 K gates using a 0.35-mm CMOS process.