D. Schmidt and T. Suda. Transport system architectures for highperformance communications subsystems. IEEE Journal on Selected Areas in Communication, 11(4), May 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....general purpose operating systems as well as general purpose communication paradigms for distributed systems do not cope very well with such types of applications. A major core of the problem has been found in the isolation imposed by layered system and protocol architectures [Ten89, CT90, CWWS92, SS93, XP99] Within this context, we are currently investigating a new concept based on knowledge transfer. Knowledge transfer overcomes the strict layer separation in traditional layered architectures, which is known as the fundamental reason for the architectural drawbacks. Fundamental to the ....

D. Schmidt and T. Suda. Transport system architectures for highperformance communications subsystems. IEEE Journal on Selected Areas in Communication, 11(4), May 1993.

....to achieve high throughput and resource utilization while keeping waiting times within predictable limits. The configurations considered in this study are (see Figure 5) ffl connectional parallel connections (config.c) where a processor is handling only the messages of its assigned connections [29]; ffl message parallel connections (config.m) where any processor may process (completely) any message (from any connection) 29] and ffl quasi pipelined connections (config.1p, config. 2p, config.4p) where a processor is assigned one or more protocol modules and their processing is based on ....

.... in this study are (see Figure 5) ffl connectional parallel connections (config.c) where a processor is handling only the messages of its assigned connections [29] ffl message parallel connections (config.m) where any processor may process (completely) any message (from any connection) [29]; and ffl quasi pipelined connections (config.1p, config. 2p, config.4p) where a processor is assigned one or more protocol modules and their processing is based on shepherding. In the classical pipeline [29] each protocol module is assigned its own processor. This makes the achievable ....

[Article contains additional citation context not shown here]

D. Schmidt and T. Suda. Transport system architectures for high-performance communications subsystems. IEEE Journal on Selected Areas in Comm., vol.11,No.4, May 93.

....assignments to processors) to optimize performance metrics as per connection maximum message rate, latency, predictability, and efficiency. The configurations considered in this study are (see Figure 4) ffl connectional parallelism (config.c) where each connection is assigned its own processor [16]; ffl message parallelism (config.m) where a message (in any connection) may be (completely) processed on any processor [16] and ffl quasi pipelined (config.1p, config.2p, config.4p ) where more than one protocol module may be assigned to the same processor, and where messages are shepherded ....

....efficiency. The configurations considered in this study are (see Figure 4) ffl connectional parallelism (config.c) where each connection is assigned its own processor [16] ffl message parallelism (config.m) where a message (in any connection) may be (completely) processed on any processor [16]; and ffl quasi pipelined (config.1p, config.2p, config.4p ) where more than one protocol module may be assigned to the same processor, and where messages are shepherded through all of those modules before a processor is relinquished. These configurations are chosen to cover a varied range of ....

D. Schmidt and T. Suda. Transport system architectures for high-performance communications subsystems. IEEE Journal on Selected Areas in Comm., vol.11,No.4, May 93.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC