Re:Three Packets Walk into a Tunnel
by ofrzeta on Thursday December 12, 2024 @12:37AM
(#42396519)
Around that time (1996) I worked a lot with ATM. We had Cisco routers and switches and Fore switches that were doing something called "LAN Emulation" (LANE, https://www.cisco.com/c/en/us/tech/asynchronous-transfer-mod...) on ATM (with a 622 Mbit/s backbone, no less).
It was bleeding edge technology and kind of a nightmare because randomly the management plane would lose the mappings from IP to ATM and the "LAN" randomly stopped working (here's a troubleshooting guide, though :-) https://www.cisco.com/en/US/docs/internetworking/troubleshoo...) . We also did some wide-area networking with ATM that worked by manually setting up ATM PVCs with some telco guys throughout Europe.
Re:Three Packets Walk into a Tunnel
by MisterTea on Thursday December 12, 2024 @09:53AM
(#42399624)
ATM was an interesting networking tech. I like how it was a hybrid of packet and circuit switching where you could dial another machine to setup a call (this is Telephone tech after all) and you were handed a an in-order point-to-point byte pipe. Made TCP unnecessary. Unfortunately Ethernet was eating its lunch in the 90's leaving it to die a slow death in Telephony.
A similar tech was the old Bell Labs Datakit, https://en.wikipedia.org/wiki/Datakit which was supported in early versions of Plan 9, http://man.postnix.pw/plan_9_2e/3/datakit. The Datakit device and dk conversation services enabled you to dial a machine and you'd get a stream which is basically an fd you just read/write. They could use this to dial a disk server using 9p over dk to boot a terminal using datakit. Since 9p multiplexes access to files and files can provide services you could then import an IP stack running on a remote machine to get IP over 9p over datakit.
Another thing I find highly ironic is how they duck tape TSN and other QoS stuff on top of Ethernet to make it behave deterministically like ATM, the technology it destroyed.
Traceroute Isn't Real
https://news.ycombinator.com/item?id=42054835
1) Computer Networks: A Systems Approach
2) Software-Defined Networks: A Systems Approach
3) TCP Congestion Control: A Systems Approach
4) Private 5G: A Systems Approach
5) Edge Cloud Operations: A Systems Approach
It was bleeding edge technology and kind of a nightmare because randomly the management plane would lose the mappings from IP to ATM and the "LAN" randomly stopped working (here's a troubleshooting guide, though :-) https://www.cisco.com/en/US/docs/internetworking/troubleshoo...) . We also did some wide-area networking with ATM that worked by manually setting up ATM PVCs with some telco guys throughout Europe.
Back then there was also quite a bit of activity for using ATM with Linux, led by Werner Almesberger (who also developed the LILO bootloader): https://www.almesberger.net/cv/papers/atm_3rd.pdf
A similar tech was the old Bell Labs Datakit, https://en.wikipedia.org/wiki/Datakit which was supported in early versions of Plan 9, http://man.postnix.pw/plan_9_2e/3/datakit. The Datakit device and dk conversation services enabled you to dial a machine and you'd get a stream which is basically an fd you just read/write. They could use this to dial a disk server using 9p over dk to boot a terminal using datakit. Since 9p multiplexes access to files and files can provide services you could then import an IP stack running on a remote machine to get IP over 9p over datakit.
Another thing I find highly ironic is how they duck tape TSN and other QoS stuff on top of Ethernet to make it behave deterministically like ATM, the technology it destroyed.