30 July 2011

Token ring multistation access unit

A physical ring topology is prone to fail when any of the nodes on the ring fail.  Suppose the beads below are nodes and the string represents network links.  A break at any node or any link breaks the ring.
The solution is to wire all nodes on the ring via a central 'hub' that bypasses any node that is not responsive or not present.  Below the nodes are linked though such a hub.
If a node becomes unresponsive, the hub simply bypasses that node and the ring remains a ring.

The 'hubs' through which token ring networks are wired are known as multistation access units (MSAUs or MAUs).  Below is an 8-port IBM 8228 MSAU.  It can operate at either 4 or 16Mbps using the IEEE 802.5 standard.  It also has RI and RO (ring-in and ring-out) ports enabling it to be connected to up to 31 other MSAUs making it possible to build a ring of up to 256 nodes.  (Later MSAUs also supported 100Mbps, and plans were made to build 1Gbps MSAUs, but were abandoned before they were built.)

The ports on the IBM 8228 use the IBM Data Connectors (IDCs).
The connectors are interesting because the 'plug' and the 'socket' are identical.  For this reason the data connectors are often called hermaphroditic.  As one may observe in connector number 7 above, two pairs of copper strips are exposed in the middle - one for sending and one for receiving.  To plug another connector onto this one it would be turned by 180°.  The transmission pair of the one would automatically mate with the receiving pair of the other and vice versa.

The next picture shows two cables that are about to be plugged into one another.
And voila!  The two become one.

An aspect of the connectors that may not be clear in the post thus far is their size.  The MSAU is intended to be built into a 19" cabinet.  Each connector measures about 25mm x 25mm x 40mm.  That is huge!

Here is the MSAU with the 18" (45cm) ruler from earlier posts.

The IBM 8228 MSAU was shipped with eight baluns that converted the IDC to an RJ-45 socket. Baluns, however, will be the topic of a future post.

24 July 2011

Sun JavaStation

In the second half of the 1990s a few vendors had a dream: Let's build a computer where the operating system, applications and data are somewhere on the network.  IBM succinctly described the dream in in the manual of its Network Station 300.
"The Network Station Series 300 is designed for environments that combine a need for multiple server access with browser access to applications and data that reside on a corporate intranet or the Internet. ... [It] can be thought of as the 'Internet computer'."
(IBM's Network Station will be featured later in this museum.)

These computers were known as network computers - a phrase still trademarked by one of the vendors.

Sun's contribution was the JavaStation - or the JavaStation NC (where NC indicates that it is a network computer).  The first JavaStation version looked like a brick.  The second (and 'real') version of the JavaSation had a much more futuristic appearance.
On the technical side the major innovation of the second version was the ability to operate outside the context of an intranet.  It could boot on its own, and then use applications, data and other services from the Internet, without any 'local' facilities.
On the outside there is hardly any indication of the intended use of the JavaStation.  Two 'doors' at the rear reveals something more about its intended use.
The key connector is the RJ-45 socket that enables it to be connected to the network.  Provision is also made for a (PS/2) keyboard, a (PS/2) mouse, microphone, speakers and a serial connection (and, of course, electricity).

In order to estimate the size of the JavaStation it is pictured below with a matchstick.
On the inside it has a (fast for the time) 100Mhz MicroSPARC IIep CPU and provision for up to 64MB of RAM.

09 July 2011

ISDN 'modems'

In order to connect one's network to the outside world via ISDN BRI one needs some device to do so.  In an earlier post it was explained that the NT1 terminates the telco's network and 'converts' it to an S/T bus.  Now we need devices to plug into that bus.

In general the lines coming into a company are plugged into some networking device, such as a router.  A medium-sized company may, for example, have lines coming in from various offices - they may run a copper cable to their building next-door, have leased lines to their bigger branch offices and then use (or have used) ISDN to connect to a number of their smaller offices.  This means their router (or other network device) has to be able to connect to a range of possible media.  And nobody wants to replace an expensive piece of technology the moment other technology changes - for example replacing the ISDN line to the telco with an optical fibre to the ISP should not require a new router.  The solution was obvious: put the interface to the media on an interface card that could be replaced when necessary.  We have seen such solutions in mainframes, minicomputers and personal computers.  A number of products that are currently being built ignore that lesson - they are engineered for obsolescence because often the battery is built into the unit.  When the battery dies, so does the unit.  The unit will not survive until new technology comes along.

But, back to interface cards.  The card below provides an ISDN interface for a range of Cisco routers.  It is a Cisco ISDN BRI S/T 1 port module for the 17XX / 26XX / 36XX series routers.  As the S/T indicates, it is intended to be plugged into the S/T bus (or the S/T bus into it?).  In most contexts one would simply take an appropriate cable with two RJ-45 plugs on each end, and plug the one end into the NT1 and the other end into this interface card.
In the picture, the S/T socket is clearly labelled as such.  There are three tiny lights behind the holes in the cover plate.  Those holes are labelled B1, B2 and OK.  Obviously the OK light glows if the interface card operates correctly, and the B1 and B2 lights indicate activity on each of the two B channels.

This particular module is the Cisco WIC-1B S/T.  WIC is an abbreviation for WAN interface card.  The intention is to feature several others WICs on this site at a later stage.

The word 'modem' was placed in quotation marks above because one normally uses a modem to connect to a telephone line and ISDN BRI is, in some sense, a glorified telephone line.  However, where normal telephone lines use analog signals, ISDN uses digital signals and the role of the ISDN WIC would be to provide a digital-to-digital connection.  Some conversion may be required, but it is not modulation or demodulation, and calling this WIC a modem would therefore be incorrect.

However, the following product is a bit harder to classify.  Whereas the medium to big enterprise may use a Cisco (or Nortel, or other) router, the small office or home office only wants to plug a few devices into its ISDN line.  So, let's look at a US Robotics Courier ISDN modem (omitting the quotation marks around 'modem' while we are making our minds up whether it is indeed a modem).
At the very least, the word modem appears clearly on the faceplate of this piece of equipment.  Some readers may also recognise the V.34, which refers to an old analogue modem standard.

Let's look at the rear of the unit in an attempt to figure out what is going on.

The reader should immediately notice the S/T socket.  This is where it will be connected to the S/T socket on the NT1.  To the left is a fancy socket for the power adapter, which we may ignore.  Then there are two more interesting sockets: an RJ-11 socket labelled phone and a DB-25 female plug labelled data.

The RJ-11 socket is intended for an ordinary old-style analogue telephone.  This unit will convert the analogue signals from the phone to digital signals before sending them out on the S/T bus, and it will convert digital signals arriving on the S/T bus to analogue signals that will be audible as normal speech on the analogue phone.  For this the unit acts in reverse when compared to a normal modem, where digital signals are converted to analogue signals to go out on the phone line.  Here the analogue signals of the phone are converted to digital to go out on the digital phone line and vice versa.  Does that mean we may call the unit a modem (or is it perhaps a demod)?

As usual there are complications.  The telco is really now providing two networks: one analogue and one digital.  If you use your fancy ISDN modem to phone someone who is still using the plain old telephone system (POTS) you are sending out digital signals on a digital network and they are waiting for analogue signals on an analogue system.  So, somewhere the telco has to interface its digital network to its analogue network and if it sees digital telehone data coming along an ISDN B channel it has to convert it to analogue before placing it on the analogue network, and vice versa.  Of course if you are calling someone who has a digital phone on the ISDN network (or who has a similar setup than you) that phone (or the modem there) will 'understand' the digital signals and interpret them correctly without any further involvement from the telco.,

Let's now move our attention to the DB-25 female plug labelled data.  This is a serial (or RS-232-C) connection.  For the modern reader, an RS-232-C interface is almost like a modern USB or FireWire interface; it's just much slower and much more fiddly.  Of course this is where the computer is plugged in.  It provides a digital connection between the computer and the modem, so it seems most of our problems disappear.  Digital-to-digital is usually simple - we may have to rearrange the bits a bit, add some overhead bits and so on, but it all really amounts to repackaging .  Again life is not that simple.  It is possible that the party you are communicating with also uses an ISDN line, in which case most of our problems are indeed solved.  It's digital to digital to digital - via RS-232-C, then via the ST bus, then via the U loop, then via the telco's ISDN infrastructure, then via a U loop again, via another S/T bus, an RS-232-C link and we are at our destination!

However, suppose you want to communicate with a poor soul who is still using an analogue modem on the POTS.  There seem to be two possibilities.  The one is that you use the telephone plug - convert your digital data to analogue, send it into the phone interface that will convert it to digital immediately, transmit the digital data to the bridge between the telco's analogue and digital networks, where it will be converted to analogue again - until it reaches the destination modem, and where it will be converted to digital again for consumption by the destination computer.  If this sounds a bit cumbersome (and error prone) it does because it is.  A better solution would be to keep the data in digital format as it travels from the computer, along the ISDN line, until it gets to the point where it hops over to the POTS network.  At this point we have no choice - the data has to be modulated to analogue - but it only has to happen once which is a significant improvement.  However, the rate at which data can be carried along an analogue phone line is slower than the 64kbp (and much slower than 128kbps) of ISDN.  So, if we want to use this option we have to tell our ISDN to limit speeds that can go over to an analogue network.

To discuss the matter further would require a detailed discussion of the rather complex configuration of the modem.  Amongst others it entails telling the data channel what the maximum speed it is allowed to use if it will connect to an analogue modem on the other end.  And it entails configuring the analogue B channel such that it knows whether is gets voice or data (or fax) inputs.  The interested reader will find the modem's manual online.  After working through that documentation the reader should be able to answer the question whether this is a real modem or not.


Let's return to the WIC.  As noted, the user of the router can in principle replace the WIC with any other WIC.  So, if the user wishes to start using ADSL rather than ISDN BRI, the ISDN WIC can be pulled out and replaced with an ADSL WIC.  Of course this means the user has to ask the telco to convert its line to an ADSL line.  ADSL will have to wait for a later post.  In the meantime, here is an example of an ADSL WIC.

Side by side one can see that the two WICs look rather similar.  The ADSL one uses an RJ-11 plug - like normal phones do, while the ISDN uses an RJ-45.  Other than that the status indicators are different (and the placement of the socket obviously differs).

To conclude, here are the three devices featured in this post together (and two matchsticks have been added so indicate size).

05 July 2011

Roaming costs

This is the front page of the Financial Times (Europe) of 4 July 2011:

The crux of the headline story is:

Visually, at the prevailing exchange rate, this means:


Integrated Services Digital Network (ISDN) - as the name indicates - is a digital network (provided by telecommunications companies) that may be used for various (integrated) services.  The services that may be 'integrated' in this manner include (voice) telephony, fax, (computer) data and anything else that may be represented as 1s and 0s.  ISDN comes in two flavours: Basic Rate Interface (BRI) and Primary Rate Interface (PRI).  PRI will take a E1 (or T1) line at one end and split it into 30 (or 23) 'data' channels.  BRI will take a more-or-less normal telephone cable and split it into two 'data' channels.  BRI is the topic of this post.

Below is the NT1 device supplied by Telkom if one had a BRI ISDN line installed at one's home or one's office.  (There was an alternative device that will be mentioned in a later post.)
Note that this NT1 device is not an "ISDN modem" - those will be discussed in a later post (to the extent that they exist).  The NT1 is the network terminator installed in one's house or office.  It used to be part of Telkom's network and thus belonged to Telkom.  The user's equipment could be plugged into the other side of the NT1.

Opening the NT1 provides more colour, but not more insight into its operation.

Let's rather look at the four sockets that are provided.
On the left is the socket into which Telkom's line was plugged.  This is known as the U interface.  The nice thing was that one's ordinary old copper wire running from the exchange could the used.  At the exchange they simply removed the link from the normal telephone system and attached it to an 'exchange' network terminator.  And, at one's home the telephone was unplugged and the NT1 plugged in.  Where the old telephone line could only be used for a single conversation, it now provided two data channels, either of both of which could be used simultaneously to carry data (including digitised voice from a digital telephone).

The not-so-nice consequence of this alternative use of the telephone cable was the following.  Normal telephone wires supply 50 volts of electricity.  Once the line was connected to the exchange network terminator 100 volts of electricity came down the line.  No, it was not able to electrocute someone (because the current was limited), but it was quite capable of sending a phone or modem up in smoke.  There are a couple of sad tales of people who found a 'telephone jack', unplugged whatever was plugged into it and plugged a modem in - only to immediately learn that this was no ordinary phone jack...

The next two sockets - like the U interface - are RJ-45 sockets.  As noted earlier, this is where the user's equipment would be plugged in.  It is easy to infer that each of those sockets would provide one of the two data channels, but such an inference would be wrong.  The interface on the user's side is actually a bus (known as the S bus or S/T bus) into which multiple (more than two) devices could be plugged.  Those two sockets are merely two access points to the bus.  If one had more than two devices, connections to those sockets would be split to enable one to connect multiple devices.

The observant reader may have noticed that there is a part of the cover that can be removed.  Once removed it exposes some contacts and some switches.  Let's ignore the switches and just look at the contacts.  The upper two contacts provide an alternative to attach the NT1 to the copper cable from the telecommunications company.  Rather than using a normal phone jack, the incoming line could be inserted into the top two holes.  This immediately removes the temptation for anyone to plug a modem into the phone jack, because there no longer is a normal phone jack.  Similarly, the bottom four holes provide an alternative attachment point for the S bus (replacing the need to use the two external RJ-45 connectors).  If these connectors inside the box were used, it was typical to plug blank RJ-45 plugs into the external RJ-45 sockets.

The last remark about four points to connect the S bus may yet again lead the reader to the wrong conclusion.  It is not the case that they provide two holes for each of the two S sockets.  Looking at the picture above one may (correctly) infer that each of the RJ-45 S sockets uses four wires.  Two wires are used for sending information and two wires for receiving information.  Therefore the four holes seen earlier just provide access to the four lines of the S bus.  They would be split as required.

We said earlier that BRI provides two 'data channels'.  Technically those channels are known as bearer channels, or B channels.  Each B channel operates at a (guaranteed) rate of 64kbps.  And then there is a control channel, known as a D channel, operating at 16kbps.  BRI is therefore also often denoted as 2B+D.

If any device on the S bus wants to communicate it would send a request via the D channel.  A B channel would then be allocated to it (or, if requested, both B channels could be allocated to it providing a guaranteed 128kbps connection).  If one device uses one B channel, another may use the other.  Though multiple pieces of equipment could be connected to the S bus, at most two B channels could be in use at any time - and it is possible that one device could be using both.  For incoming calls it was possible to let any device answer it, or, based on various schemes, to let a specific device answer the call.  If one had four digital phones on the S bus it possibly did not matter who answered, but if one had a fax machine, telephone and computer connected, who answers could be important.  Even when using only telephones, it could be important exactly which phone rang.

64kbps sounds slow by today's standards.  However, it should be kept in mind that the bandwidth is guaranteed.  Even a megabit per second ADSL line may slow down to almost no throughput - something which is not supposed to ever happen on an ISDN line.

We are now in a position to discuss the function of the NT1.  On the one side (the U side) is a single copper loop carrying three channels (2B+D) in two directions.  These six streams of data need to be multiplexed and demultiplexed  by the two network terminators.  In addition the NT1 has to manage the S bus: if a device puts data on the 'send' pair of the S bus, the NT1 should figure out on which channel it belongs and then multiplex it accordingly.  Similarly, if data arrives from the U interface, the NT1 should figure out for which device it is intended and then send it out on the S bus with the appropriate addressing information on the receive pair of wires of the S bus.

Just in case their is any doubt left in the reader's mind: An ordinary phone cannot be plugged into either of the S sockets.  One needs a digital phone that can put the appropriate addressing information on the S bus and respond when its address appears on the S bus.

Oh, yes - we never discussed the fourth plug.  It is a normal figure-8 connector that is used to provide electricity to the NT1.