dorl@vms3.macc.wisc.edu (Michael (NMI) Dorl) (02/19/88)
I read the article in 'Data Communications' on the QPSX 802.6 metropolitan LAN yesterday and am a bit confused. Maybe one of you could put me right. The article talks about two unidirectional but opposite directional buses, A and B. A node can write information on either bus but its a one way write so only downstream nodes see the data. So how does a node know which bus it should write data to send to a given station? It seems it would have to write to one bus for some stations and another bus for other stations. I suppose it could write data to both buses. The article talks about 'active' and 'passive' failure of a node? What do these mean? How will these be detected and bypassed? Who would be possible clients for such a net? Off hand, it appears that equipment to interface with the framming protocol will be quite complicated. Will this restrict possible uses to high volume customers or are connections to low cost equipment planned?
ooi@otter.hple.hp.com (Stanley Ooi) (02/25/88)
> ... how does a node know which bus it should write data ... The addresses on the station are in order i.e. if one node wants to send to another node with an address greater than itself, it will use the "ascending" bus and vice versa for a node with an address smaller than the initiating node. > The article talks about 'active' and 'passive' failure of a > node? What do these mean? How will these be detected and > bypassed? The way QPSX works is by having its nodes attached to "access units" (AU) which are nothing more than a shift registers. When a node wants to transmit data, it will "OR" the appropriate bits into the data stream going through the shift register. So, if a node dies, then as the data stream does NOT go through the node, the rest of the network is unaffected (i.e. passive MAC failure). However, if the power to the AU should disappear (say), then it will be an "active" failure i.e. the network will have to reconfigure in order to function (as the shift register will stop shifting and the data stream is essentially stopped). > Who would be possible clients for such a net? PTTs - it is proposed as a standard for a metropolitan area network and that's the homeground of the PTTs. It could probably be used as a backbone network connecting several (slower) Local Area Networks within a large campus or manufacturing site. Stan "why not QPSS?" Ooi
mw3s+@andrew.cmu.edu (Martin Weiss) (02/26/88)
I am no expert in this area, but my understanding is as follows: >So how does a node know which bus it should write data to send to a given station? My understanding (based on second hand information) is that this is still being discussed in the committee. There are several possible ways of doing this, including having a node "learn" the relative locations of other nodes dynamically, and developing a numbering scheme such that the address of the node would indicate the direction. Each method clearly has benefits and problems. There may also be other techniques. >The article talks about 'active' and 'passive' failure of a node? What do these mean? How will >these be detected and bypassed? I don't recall exactly what the article said, but I believe that this issue was addressed in the MAN article or its companion article on QPSX in the same issue. I suggest you re-read both again (I will as well for my own information). >Who would be possible clients for such a net? The article makes the point that local telephone operating companies as well as corporations would be potential users of this MAN. I hope this helps.