| A simple choice? |
| Friday, 28 September 2007 01:00 | |
|
Has structured cabling finally become too complex for a mix ‘n’ match approach? Paul Eo, regional manager Europe for LS argues that it has – and explains why it is time for a more simple solution.
If one reads the pages of specialist network and cabling journals, it becomes obvious what was once the simple “art” of structured cabling which more or less amounted to “throw it in and connect it up” no longer applies. It’s not that the people installing cabling networks ten or even five years ago were cowboys (although sadly some were and still are) – the main reason is that technology has moved on massively and the science of structured cabling has changed almost beyond recognition. Many IT people don’t yet fully understand this because the visible side of network cabling – the RJ-45 plug and jack – still looks exactly the same as it did ten years ago. The difference is that ten years ago its function was to transmit data at 10 Megabit/s (requiring a line frequency of 16MHz max) whereas today’s category 6A version, which looks identical, is expected to carry data at 10 Gigabit/s with a frequency of 500MHz. To put this in context, 500MHz is 31 times the frequency and is somewhere between VHF and UHF TV aerial frequency. Of course the data rate is 100,000 times faster. If we look back through history, after the earlier co-ax based systems, there was once Category 3 cabling for 10Megabit/s – the original Ethernet speed. It wasn’t much different to ordinary telephone cable and, like telephone cable, it was highly insensitive to the treatment it received. Technicians could be rough with it, other trades could walk on it, place ladders on it, kink it, stretch it and, within reason, it didn’t matter. 10 Megabit/s Ethernet was not particularly sensitive to impedance matching, return loss or crosstalk. Factors like alien crosstalk had never been heard of.Then came Fast Ethernet operating at 100 Megabit/s full-duplex. This required a line signalling frequency much higher at 70MHz. Category 5 cabling was devised. (Category 4 never existed) Category 5 was designed to offer less attenuation and the pairs were better balanced to reduce crosstalk between the individual pairs to levels that would allow this ten-times rise in data-rate. In the field, everyone assumed that Cat 5 was as resilient as Cat 3 and so continued to treat it roughly paying little heed to the termination techniques used by installers. As long as it passed relatively simple electrical tests it was deemed to be OK. While installers and manufacturers learned quickly from this, the facts were highly embarrassing to customers and suppliers alike and so most IT people never got to hear about this with the corollary that they carried on thinking that structured cabling “doesn’t need thinking about. It just goes in easily and always works”. Sadly this wasn’t true even ten years ago! Moving on a few years, engineers decided that by using all four pairs in Category 5 cable, they should be able to increase its data-rate by a factor of ten again – this time to one Gigabit/s (1,000Megabit/s). At this point, one of Murphy’s Laws came into play - along the lines of “If it can go wrong it will,” and a couple of previously unmeasured and uncontrolled characteristics of the cable and jacks suddenly became critical to Gigabit operation. A new standard Category 5e had to be created and much of the Category 5 cabling already deployed in the field was declared incapable of carrying the new ten times faster data-rate. Stepping forward to today, we now have both shielded and unshielded copper cabling systems capable of data rates up to 10 Gigabit/s. Thankfully the cable no longer looks like telephone cable – it is much thicker and less flexible - although the RJ-45 plugs and jacks both look just the same as ever. But, being required to operate at an incredible 500MHz, they are very sensitive to handling, installation and termination methods. Certainly once a ladder has been stood upon a Category 6A cable it is unlikely to ever be capable of full speed operation because the sensitive balance between the pairs will have been destroyed - creating an impedance mis-match, signal reflections, data corruption and data packet loss! Readers will perhaps be beginning to realise that what once was a very simple technology has now morphed into one where ultra high frequencies are being transmitted along complex waveguides that just happen to look like cables. (The physics of ultra high frequencies means that most of the energy actually travels in the air or dielectric space around each conductor and not actually within the conductor!) At these UHF frequencies, the amount of signal that actually gets from one end of the cable ‘channel’ to the other is miniscule and crowded out with all manner of noise signals. While same-channel noise (from other pairs) can be compensated for by signal cancellation techniques, signals from other cables – ‘Alien Crosstalk’ - are impossible to cancel out because their form is unknown. This is why many different designs of Category 6A cable now exist all having the aim of keeping cable pairs in adjoining cables as far away as possible and never running parallel. (If they did run close and parallel, the wires would act like a transformer and couple the two cables’ signals together - leading to large scale data corruption). Of course, all the manufacturers will guarantee that their cables and connectors have been designed to reduce alien crosstalk well below acceptable levels and so this does not represent a problem. The part they aren’t so keen to tell you is that, in real world situations, where 10Gigabit/s Category 6A cabling is likely to be installed alongside existing Category 5e or Category 6 links – all sorts of havoc is waiting to happen. This is because those older Category cables don’t share the physical alien crosstalk avoidance designs and hence the transformer effect will occur leading to failure of the Category 6A to be able to carry 10 Gigabits when the active equipment finally arrives! No-one is quite sure yet, but it is likely that the Cat 5e/Cat6 channels will also suffer data corruption. Of course, it is not all doom and disaster. If one understands enough about near-end and far-end crosstalk, impedances, insertion loss, signal to noise levels, alien crosstalk and a dozen or more other technical factors, it is possible to devise a rock-solid copper, fibre or combination network to take a data-centre or enterprise network comfortably through the next ten years. A structured cabling system, in our view, is just that – a system. As such it should not be put together from a selection of random components, it needs to be engineered as a system by experts to perform a specific function. It should be designed, installed and handed over as a complete, working and guaranteed system. A solution in fact guaranteed to work with standards compliant active equipment. It is even more essential in a data centre environment - where five 9’s or even six 9’s reliability are being specified - that no chances are taken with the interconnection layer of the data centre. Research by Gartner and others has indicated that up to 70 % of downtime can be attributed to faults or finger-trouble related to the cabling system! An interesting analogy is the motor car that we all drive. Twenty or thirty years ago they were relatively simple (albeit it hopelessly unreliable by today’s standards). Many of us, or our fathers at least, would happily change their own brakes, clutches, piston rings etc. Now cars are so complex that if you open the bonnet there is a frightening array of wires and pipes - and you can’t actually diagnose anything without the manufacturer’s software on an attached computer. In fact my latest car has a cover over everything in the engine compartment except the oil and water fillers! A very strong hint that it’s too complex for an amateur like me to meddle with. After all at today’s motorway speeds to meddle with these systems could lead to a life or death situation. Well, in my view, the same is true of a data centre structured cabling system. Not life and death – but given the financial and reputation costs of any downtime in a Tier 4 or Tier 5 data centre – additional downtime caused by the cabling system is totally unacceptable. Worse still, given that the cabling only accounts for 10-12% of the IT budget, it makes no sense to install inferior mix and match cabling which doesn’t carry a system design and system warranty. So our opinion at LS is that you as data centre and network managers should not need to understand all the technology that goes inside a structured cabling system any more than you need to understand the piston technology in your car. You should only need to know, and specify, what you want your network to do. The equivalent of specifying whether you want a racing car, motorway cruiser, off-roader or shopping run-around. You should then be able to pass over all the responsibility for designing, installing and warranting the system. In some cases you may even wish to pass over the ongoing maintenance and moves, ads, changes to us and our ‘main dealers’.
|




