|
|
I am about to obtain a serial port concentrator board. Can someone describe the black box that grex's concentrator uses so that I can ponder how I would go about constructing one? Does it have electronics in it or is it merely an electrical connection? What are the board settings? Ack
8 responses total.
What does a serial port concentrator do?
Grex uses a systech MTI-1600. This consists of the following parts: A host multibus card. Onboard is a Z80. It interfaces to the host via a command FIFO, which is loaded with commands to manipulate the ports. The Z80 can also interrupt the host when it has data for the host--which is read from a response FIFO, and it can perform DMA to the host memory, under control of the commands. Unix configures the ports, does DMA output from swab'd copies of clist elements, and does character at a time interrupt driven input via the response FIFO. (This is one major reason why it can't cope with much local high speed serial traffic.) The Z80 host card then communicates, via a 50 pin cable, to a black box which contains one or two cards. Each card has, on it, dual USARTs to implement 8 ports, a PIC (interrupt controller), and auxillary logic to handle decoding. In essense, the 50 pin cable is an extension of the Z80's bus to communicate with the USARTs. There is another version of the Systech that uses the same host adapter, but provides USARTs on a second companion multibus card. Perhaps this would be useful if you liked using 3B2 style modular jacks instead of DB-25's.
What does a multiplexer do in contrast, and are there any ansi standards regarding the operations of multiplexers?
The sytech *i* a sort of multiplexor. No, there aren't any standards, indeed, the whole term is rather vague and covers a long list of things. One sort of "multiplexor" is an MSI chip. It takes N inputs in (often 4 or 8), plus LOG2(N)+X select lines (say, 2 or 3, plus 1 or 2 enable inputs) and produces on its output the enabled and selected data input. As a discrete chip, it's difficult to predict just where it will crop up -- perhaps in the tape or disk controller to select between various modes of operation. Another sort is called a "statistical multiplexor". It takes N slow speed serial lines, mixes them togther into 1 high speed serial data stream, which can then be split down into one low speed line at the other end. It's called a "stat mux" because it doesn't actually provide enough high speed capacity to accept all of the potential low speed traffic at once. But for humans typing at keboards, this is a pretty safe statistic.
In other words, if you have the right device driver for one higher speed serial port attached to a multiplexer, you could increase your lines. Especially if you don't use them at full capacity all at the same time (say many peripherials, only a few used at the same time not continuously. So if you go out and buy a multiplexer for that purpose, how would you know what kind of interfacing with the device driver it needs? Does a description of the data stream and hardware signals come with the unit? This might be a little of the subject, but it would be useful to know.
Most of those statistical multiplexors hook up to *another* statistical multiplexor back in the computer room, and split it all back out to individual RS-232 lines again. I think Systech made another version of the MTI-1600, that used a coax cable (or something) instead of ribbon, & could talk to up to 64 ports. It could conceivably have a compatible host interface with the MTI-1600, in which case very simple tweaks to the Unix device driver would suffice. On the other hand, it may have slightly different block layouts to allow for 6 bits of port addressing (instead of 4), in which case the device driver might need a bit of rewriting, basically to build the new command blocks and interpret the new response blocks. When you buy one of these things, generally, documentation comes with the unit that describes basic programming. In the case of the Systech, you also get a sample Unix device driver, if you ask their customer engineers. However, in the case of the Sun, we were lucky -- Sun already includes a driver for the thing. Both these things are actually sort of obselete. The current trend is to buy TCP/IP terminal servers--these plug into ethernet, hook up to a dozen or so of your old terminals (themselves a dying creature), and talk standard protocols (usually rlogin and telnet) to get to your host. DEC makes something essentially similar except it talks "LAT". "Current trend" is not entirely correct -- actually, even more frequently, the tendency is to buy workstations, PC, Mac's, or whatever, and then to acquire networking hardware for those. Then again, the mainframe itself is sort of a dying breed.
The latest trend I've heard of is the 'total control hub' which takes as input an ISDN PRI, and the output is ethernet. In one box, it emulates a telco channel bank, ISDN TA, V.34 modems, and a terminal server of the type Marcus described. It'll accept calls from modems or ISDN BRI TA's, it'll auto-negotiate a PPP or shell login, and connect to the server of choice. If the back-end is a high-speed data link, the box becomes a self-contained POP site for an ISP. Nice piece of hardware, and a bit out of grex's budget, I'm afraid. Just wait a few years, and we'll be pulling them out of the local dumpsters!
Total Control Hub is USRs trademark for that type of box. All the major terminal server maufactureres are now making them. The Ascend Max series, Livingston PM3s, Cisco AS5200s, Bay 5399s (I think that's the model number) and a number of other products all do pretty much the same thing. They also have routing capabilities, and can generally handle various leased line connections in addition to the dial-up stuff. Those are all terminal servers, though, not serial multiplexers. Their terminal server functions would be a lot more comparable to the Chase IOLan terminal server Grex is using now than to the serial port multiplexer board Grex was using in 1991.
Response not possible - You must register and login before posting.
|
|
- Backtalk version 1.3.30 - Copyright 1996-2006, Jan Wolter and Steve Weiss