Trends in Distributed Computing

Stand-alone workstations delivering several tens of millions of operations per second are commonplace, and continuing increases in power are predicted. When these computer systems are interconnected by an appropriate high-speed network, their combined computational power can be applied to solve a variety of computationally intensive applications. Indeed, network computing may even provide supercomputer-level computational power. Further, under the right circumstances, the network-based approach can be effective in coupling several similar multiprocessors, resulting in a configuration that might be economically and technically difficult to achieve with supercomputer hardware.

To be effective, distributed computing requires high communication speeds. In the past fifteen years or so, network speeds have increased by several orders of magnitude (see Figure ).

  
Figure: Networking speeds

Among the most notable advances in computer networking technology are the following:

• Ethernet  - the name given to the popular local area packet-switched network technology invented by Xerox PARC. The Ethernet is a 10 Mbit/s broadcast bus technology with distributed access control.

• FDDI  - the Fiber Distributed Data Interface. FDDI is a 100-Mbit/sec token-passing ring that uses optical fiber for transmission between stations and has dual counter-rotating rings to provide redundant data paths for reliability.

• HiPPI  - the high-performance parallel interface. HiPPI is a copper-based data communications standard capable of transferring data at 800 Mbit/sec over 32 parallel lines or 1.6 Gbit/sec over 64 parallel lines. Most commercially available high-performance computers offer a HIPPI interface. It is a point-to-point channel that does not support multidrop configurations.

• SONET  - Synchronous Optical Network. SONET is a series of optical signals that are multiples of a basic signal rate of 51.84 Mbit/sec called OC-1. The OC-3 (155.52 Mbit/sec) and OC-12 (622.08 Mbit/sec) have been designated as the customer access rates in future B-ISDN networks, and signal rates of OC-192 (9.952 Gbit/sec) are defined.

• ATM  - Asynchronous Transfer Mode. ATM is the technique for transport, multiplexing, and switching that provides a high degree of flexibility required by B-ISDN. ATM is a connection-oriented protocol employing fixed-size packets with a 5-byte header and 48 bytes of information.

These advances in high-speed networking promise high throughput with low latency and make it possible to utilize distributed computing for years to come. Consequently, increasing numbers of universities, government and industrial laboratories, and financial firms are turning to distributed computing to solve their computational problems. The objective of PVM is to enable these institutions to use distributed computing efficiently.