khasin@canstar.UUCP (Kha Sin Teow) (10/02/90)
This is a request for information related to IBM I/O channel. I would like to apologize ahead if this is not the right news group. I'm not aware of any other more appropriate group. If you know of one, please e-mail me the info for my future use. PROBLEM: I'm involved in solving a problem which involves the IBM I/O channels and many I/O devices connected to the channels. There is an existing communication configuration as follow: +--------------------+ | IBM 370 | | | | +----------------+ | Bus&Tag Cable | |c h a n n e l A |#[]---------------(I/O devices which are | +----------------+ | daisy-chained) | | | +----------------+ | Bus&Tag Cable | |c h a n n e l B |#[]---------------(I/O devices which are | +----------------+ | daisy-chained) | . | | . | | . | +--------------------+ There is a proposal/suggestion to replace the existing 370 (and channels) with a network of vendor independent high-powered micro/mini computers such as Sun workstations, MIP machines, etc. The network must also be vendor and protocol independent. However, the I/O devices on the "bus and tag" side of the channel must remain. The question that I have is this: (1) Is there any product out there that emulate the 370 channel and works with a system mentioned above? (Please don't tell me it's IBM/Amdahl XXXX mainframe...) There are several channel related products that we are aware of but they are only for the "bus and tag" side and not capable of emulating the channel. You may contact me via e-mail khasin%canstar.uucp@ai.toronto.edu or voice: (416)756-4100 or Canada Post: 3900 Victoria Park Avenue, North York, Ontario M2H 3H7 Appreciate any pointer.
lars@spectrum.CMC.COM (Lars Poulsen) (10/04/90)
In article <186@canstar.UUCP> khasin@canstar.UUCP (Kha Sin Teow) writes: >I'm involved in solving a problem which involves the IBM I/O channels and >many I/O devices connected to the channels. > >There is a proposal/suggestion to replace the existing 370 (and channels) >with a network of vendor independent high-powered micro/mini computers such as >Sun workstations, MIP machines, etc. >The network must also be vendor and protocol independent. >However, the I/O devices on the "bus and tag" side of the channel must >remain. The IBM-360 data channel is also known as the "FIPS channel" because the US Government mandated that all mainframe vendors must support it in order to protect the existing investment in disk and tape drives etc without locking data centers into a single vendor. It is, however, an obsolete standard, and I think most IBM peripherals are not even attached that way any more. The SCSI channel is sort of a modern-day successor to the IBM channel. The signaling scheme of the channel is straightforward and well documented, and can be implemented by any 8-bit microprocessor with a couple of parallel I/O ports, but in order to be useful, they must also implement a number of higher-order protocols, which may not be practical for small systems. In particular, many control units (what sits out on the channel, attaching preipherals to the BUS and TAG cables) usually do not have much, if any, buffering (not much means 2-8 bytes of fifo), and thus the channel must be capable of absorbing/generating data at device speeds. For disks, this means keeping up with the rotational speed of the disk as it moves under the heads. Also, some devices require that the channel do "command chaining", where the channel issues a new command IMMEDIATELY when certain "prep" commands terminate. Many channel emulators cannot do this. When it comes to remote access to a farm of IBM peripherals, I guess the leader would be Network Systems in Minneapolis, who built the "Hyperchannel" as a local area network that looked like a cable full of control units to the mainframe, and looked like a mainframe to the control units plugged into it. This allowed data centers to set up high-speed printers and terminal clusters at client sites. Versions existed that would link up over T1 lines, too. What peripherals is it that you want to retain ? Old disk and tape drives are just not cost-effective compared to the never, smaller footprint, larger capatity drives. Old magtape drives are not very useful for the modern computing environment. And old IBM terminal clusters cannot really be made compatible with the UNIX environment. The most useful of the old stuff is 1401 line printers. There is a company called SPUR that has a unibus-to-channel converter emulating a DEC LP-11. You might be able to put that on an ULTRAX machine in your network. Hope some if this is useful to you. -- / Lars Poulsen, SMTS Software Engineer CMC Rockwell lars@CMC.COM
lstowell@pyrnova.pyramid.com (Lon Stowell) (10/05/90)
The IBM channel protocol is unfortunately NOT easily implemented with "an 8-bit micro". Attempting same is guaranteed to cause Interface Control Checks! There are some sequences which require a MAXIMUM timing of less than 6 microseconds to execute a fairly complicated series of tests and branches...... Most (channel secondary) interfaces use Bit Slicers or other RISC chips for signalling control....just for this reason. The possibility of using a Channel Extender is quite interesting.... These are boxes that come in two halves... One sits at the mainframe and emulates a peripheral.....the other sits at the remote site and provides emulation of the IBM channel itself.... The two are connected over 56Kb to T1/T3 serial lines and/or fiber optics..... If you can find a vendor willing to let you eliminate the first of the units (the one at the IBM host) and speak directly to the T1 or fiber interface to the slave unit, you should be able to drive an IBM peripheral...... This will NOT be a trivial effort, but one actually wonders if anyone has actually done this..... Vendors of such channel extender boxes are Paradyne, of Key Largo Florida,.....Network Systems of California (415 area code..),,,,Storage Technologies in the Denver area......you may also want to check with McData in the Colorado Denver suburbs. Good luck.