Difference between revisions of "CYCLADES"
m (+links, cat) |
(Move some technical content about the advantages of the CYCLADES approach here) |
||
(One intermediate revision by the same user not shown) | |||
Line 1: | Line 1: | ||
'''CYCLADES''' was a ground-breaking, and extremely influential, early [[packet switching]] [[data network]], done in France in the late 1970's. It was the key intermediate technical step between the [[ARPANET]] and the [[Internet]]. | '''CYCLADES''' was a ground-breaking, and extremely influential, early [[packet switching]] [[data network]], done in France in the late 1970's. It was the key intermediate technical step between the [[ARPANET]] and the [[Internet]]. | ||
− | CYCLADES was the first network to move reliability into the [[host]]s attached to the network | + | CYCLADES was the first network to move the responsibility for the reliability of data sent across it into the [[host]]s attached to the network. The technical chief of the project, [[Louis Pouzin]], decided that the best way to build a network was to accept that the network itself was basically somewhat unreliable, and then move ''all'' the responsibility for reliability into the [[host]]s on the ends (unlike in the ARPANET, where that responsibility was borne by the [[packet switch]]es that made up the network, the [[Interface Message Processor|IMPs]]). |
− | In addition to making the | + | They did that by using the now-standard mechanisms of [[sequence number]]s, [[acknowledgement]]s, [[timeout]]s, and [[retransmission]]s. In addition to making the packet switches in the network much simpler (a major gain, and one that would be crucial in the development of [[internetwork|internets]]), it was also a step toward the [[end to end]] model of networking. It introduced the idea of 'fate-sharing', where the state necessary for the [[connection]] is co-located with the [[application]], so generally 'they all go together if they go' (away). |
− | The actual physical network (links and switches) was named '''CIGALE'''; CYCLADES includes the whole ensemble, including the hosts and higher-level [[protocol]]s and [[ | + | The actual physical network (links and switches) was named '''CIGALE'''; CYCLADES includes the whole ensemble, including the hosts and higher-level [[protocol]]s and applications. |
+ | |||
+ | ==Advantages== | ||
+ | |||
+ | That approach, of having the network provide only ''unreliable'' [[packet]]s (sometimes called [[datagram]]s) has several huge advantages: | ||
+ | |||
+ | * The network itself is much simpler, because it does not have to go to any extra work to ''guarantee'' that data is not lost. It just does its best (hence another name for the service it provides, 'best effort'), and focuses on its main job - moving data around. | ||
+ | |||
+ | * This approach allows ''all'' the intermediate packet switches ([[router]]s, in an internet) in the overall network to carry ''no'' state about the connections travelling through them. This has been the single biggest reason why the Internet has been able to grow ('scale', in engineering jargon) to the stupendous size it has. | ||
{{semi-stub}} | {{semi-stub}} | ||
+ | |||
+ | ==Further reading== | ||
+ | |||
+ | * Louis Pouzin (editor), ''The Cyclades Computer Network: Toward Layered Network Architectures'' (North-Holland, Amsterdam, 1982) | ||
[[Category: Networking]] | [[Category: Networking]] |
Latest revision as of 11:59, 16 January 2024
CYCLADES was a ground-breaking, and extremely influential, early packet switching data network, done in France in the late 1970's. It was the key intermediate technical step between the ARPANET and the Internet.
CYCLADES was the first network to move the responsibility for the reliability of data sent across it into the hosts attached to the network. The technical chief of the project, Louis Pouzin, decided that the best way to build a network was to accept that the network itself was basically somewhat unreliable, and then move all the responsibility for reliability into the hosts on the ends (unlike in the ARPANET, where that responsibility was borne by the packet switches that made up the network, the IMPs).
They did that by using the now-standard mechanisms of sequence numbers, acknowledgements, timeouts, and retransmissions. In addition to making the packet switches in the network much simpler (a major gain, and one that would be crucial in the development of internets), it was also a step toward the end to end model of networking. It introduced the idea of 'fate-sharing', where the state necessary for the connection is co-located with the application, so generally 'they all go together if they go' (away).
The actual physical network (links and switches) was named CIGALE; CYCLADES includes the whole ensemble, including the hosts and higher-level protocols and applications.
Advantages
That approach, of having the network provide only unreliable packets (sometimes called datagrams) has several huge advantages:
- The network itself is much simpler, because it does not have to go to any extra work to guarantee that data is not lost. It just does its best (hence another name for the service it provides, 'best effort'), and focuses on its main job - moving data around.
- This approach allows all the intermediate packet switches (routers, in an internet) in the overall network to carry no state about the connections travelling through them. This has been the single biggest reason why the Internet has been able to grow ('scale', in engineering jargon) to the stupendous size it has.
Further reading
- Louis Pouzin (editor), The Cyclades Computer Network: Toward Layered Network Architectures (North-Holland, Amsterdam, 1982)