I intentionally wrote, in a short paragraph in SIP Means Change, of shortening the cable drops. I wrote this as a means to broach another difficult subject, and it wasn’t to mislead anyone, as one reader suggested. I think cable length does make a difference, and I'll tell you why.
The new cabling standards Category 6 and 6A make me ask, "what are you guys trying to 'pull'?" Whatever the cause and effect of network delay/latency, the fact remains that it can be reduced. Long cable drops just don’t make for better networks. The adopted cabling standards for Category 6A have created great performing cables, some that I reviewed over two years ago, but they are bulging at the seams, emptying budgets and adding substantial weight to cable trays and occupying more space in support systems.
Let me recount over the past couple of decades the number of times customers sought us out to fix their slow networks. How many times we ran cables from one closet to a space to add intermediate switches because of the issues experienced by users at that "end or part of the building." For many trying to move ahead into what's been coined real time communications, it won't be a pleasant journey because many existing cable plants are in fact outdated, outmoded and just can't perform anywhere near the new standard, let alone the old standard cited to me. Long term, this sizing up of cabling is a troubling trend for any enterprise, because your investment dollars in copper wiring are becoming investments with short life spans.
I’ve heard over the years cable is cable and that’s so far from reality. For years, cheap imports from offshore cable manufacturers have plagued the industry, as has cheaper termination hardware. Consistency and length are two factors in cable plants that leave much to be desired. Long cables contribute to poor network performance, and whatever the specs say, we don’t install long drops even though the standard says it will work; we avoid doing so because that’s the field practice. Long drops serve as antennas, and the longer the drop, the more noise and performance issues you can expect to encounter. Of course the long drops will work but there’s a price being paid in performance.
My other point is that specs and standards have become memorized sound bites. As I said about cable specs: examine them and you will find they are written for "room temperature," and not all your enclosed spaces in ceilings and walls fall within room temperature. Upon further examination, will your copper cable plants stand up to close inspection? Cables operate under different temperatures than room temperature, and higher temperatures increase attenuation too, about 0.4% per degree Celsius/33.8 degrees Fahrenheit for Cat5e cabling.
But there's more to cabling than meets the eye. Cables vary from manufacturer to manufacturer and over the years more Category 5 or 5E cables that were sold and installed varied in performance. For those getting caught up in the money game of buying low, it's very likely you got exactly what you paid for. Even still, higher prices don't necessarily mean better cables. I stated in other posts you’ve got to know exactly what you are buying into, and you have to take steps to view and inspect before, during and after installation.
Do long cable drops contribute to the overall scheme of the network’s performing poorly? I think they do and I'll borrow from Latency & Jitter written by Kevin Davis at Network Performance Daily:
Network latency is the amount of time it takes for a packet to be transmitted end-to-end across a network and is composed of five variables:
Network Latency = (Distance Delay) + (Serialization Delay) + (Queue Delay) + (Forwarding Delay) + (Protocol Delay)
Network Latency = (Distance Delay) + (Serialization Delay) + (Queue Delay) + (Forwarding Delay) + (Protocol Delay)
Slow networks and networks that drop packets and retransmit them once or often are adding unnecessary traffic to the wire, and let’s face it, the old networks for the most part were designed with only data in mind. One key concept I learned from the NoJitter resident guru is "data networks are not static." Those networks are living, breathing, and changing all the time. Are all your packets arriving without delay or retransmission? Is your data throughput end to end at, close to, near, somewhat near, around, about or in range of 100 MBPS, 1GBPS or 10GBPS? If not, then in management terms, this is "rework," and rework is inefficient and costs money while wasting time.
Now, examine again Kevin’s formula for network latency or read his entire white paper over at NetQOS. Folks need to examine how to decrease the times I discussed in SIP Means Change, and I think his formula on network latency is a good place to begin.
Copper is noisy, and granted, Cat6 wire helps mitigate noise and increase signal strength (per Galen Udell of Belden, Discussion Paper Category 6 vs Category 5e Cabling Systems and Implications for Voice over IP Networks). Galen points out numerous other advantages of Cat6 and the impact of noisy cables--most of which are embedded in networks. Galen adds in his paper:
According to Paul Kish, Director of Networking Systems at Belden CDT, "There is a direct relationship between the Signal-to-Noise Ratio (SNR) and the Bit Error Rate (BER) performance of digital networks. The higher the Signal-to-Noise Ratio over the available bandwidth, the fewer bit errors are generated and the higher the data throughput. For Ethernet systems, information is transmitted in frames (a sequence of bits with an address header) using TCP/IP protocol. It has been shown that a 1% frame error rate can slow down the performance of your network from 100% to 20%."
Maybe you missed Gary Audin's piece on Certifying Copper: Worth Learning. Do you know how many variables go against the manufacturer's certifications/warranty of wire? This brings me to what’s really up in the ceiling and dropped behind the walls and stuck in EC all over the planet. Don't assume because your last generation cable scanner didn’t find any issue with the cable that’s there’s not an issue. You can scan all day long but your test gear is as good as your test gear--meaning old generation testers don’t come close to the new wares and capabilities that are sold today.
Cable plants are pretty much static in the sense that they don’t normally move, physically, but what could influence cable plant are sudden shifts of weight, rodent issues (food source--normally PVC), change in the structural environment (water leak, cable cut/damage, fire), change in temperature or constant heat, distance and bundling vs. unbundling (noise, Alien Cross-talk) and external power line influence and lightning. Even damage to cables during installation may not show up immediately but may appear later. This includes over-tightening tie wraps or an incorrect bend radius and even rips/tears in the cable outer jacket. Then you may not remember an old telephone rule in the PSTN world that I think still applies today to cabling: "The longer the circuit, the more troubles there are."
I extracted some key points from the Siemon ebook on cabling:
Transmission delays between pairs can vary by as much as 1%.
In the common mode, the UTP (Uniform Twisted Pair) link acts as a long antenna, while the FTP (Foil Twisted Pair) or S-FTP (Shielded Foil Twisted Pair) cable behave more or less as a coaxial cable.
For a given length of Cat-5 cable the total length of a particular pair could be longer than others. Since the signals travel in the cable at a fixed rate (approximately 90% of light speed), the arrival times of signals can be skewed in a long cable (those that have to travel farther arrive later and the corresponding image shifts to the right).
In a few years' time, the unified wiring technology that is now recommended may be inadequate
In the common mode, the UTP (Uniform Twisted Pair) link acts as a long antenna, while the FTP (Foil Twisted Pair) or S-FTP (Shielded Foil Twisted Pair) cable behave more or less as a coaxial cable.
For a given length of Cat-5 cable the total length of a particular pair could be longer than others. Since the signals travel in the cable at a fixed rate (approximately 90% of light speed), the arrival times of signals can be skewed in a long cable (those that have to travel farther arrive later and the corresponding image shifts to the right).
In a few years' time, the unified wiring technology that is now recommended may be inadequate
With regard to that last remark: Are we going to see Cat 10E or Cat 10A cables one day? Then, how can we create less waste and more sustainable copper networks? Do you have any idea what a Cat 10E or 10A cable would look like, weigh and how it would perform? As it is now or on any given work day with traffic on your network, can you say your network communications is error free? Of course reality says it can’t be, but how close are you? Then when you consider which wire is best and whether or not to add a shield and make the extra investment in plenum for non-plenum spaces, your worries still won’t be over because you are trying to install a cable plant that not only meets today’s needs but tomorrow's. More importantly, designing out delay/latency isn’t a 100% reality, but electing to design in latency or potential problems associated with copper wiring isn't going to create the best network environment for real time communications.
I challenged you on switches as to whether they are acting like switches or behaving like repeaters. Extend those cables and I believe you get the latter but I also know what marketing did to the term "switches." This is also a challenge to every switch manufacturer. Are your switches glorified repeaters or are they really switching? Then what can you do to improve them? Fat MTU sizes or Jumbo Frames is another inkling. More songs are coming about IPv6. Jumbograms?
Not long ago customers were told "you must have Cat5E for VoIP." Now the tune is changing to “you must have Cat6 for Real Time Communications” but the choir isn’t singing of whether they mean Cat6 or Cat6A and there's a huge difference. In some cases some Cat6 cables weigh as much or less than Cat5E cables and performance varies between the two. Go figure.
Cat6A however is a different story--not only do they weigh more, they are bigger. Their size and weight depend upon manufacturer too. Cat5E per thousand feet can weigh around 28 pounds, some Cat6 comes in at 26 pounds and I’ve seen Cat6A weighing in from anywhere between 40-53.3 pounds per thousand feet (one box). What this means for you is that you’ve got to alert your architect to those electrical conduits that may not suffice for drop locations in some situations. The fill rate of cables in conduits is still 40% for 3 or more cables so check your Cat6/6A cable outer diameter, because they range from skinny to fat by manufacturer, and then you’ll know your numbers. Then, you need to budget about 20-25% more money for Cat6 cabling, 45-60% more for Cat6A cabling and then more for other supporting improvements. Cat6A are monster cables and this means less drops pulled simultaneously and that adds to labor costs, longer time to terminate, again this means more labor costs and higher costs for higher performance hardware on both ends including higher performance patch cables. Your cable trays need to be designed to carry this added weight. Then you may need to change job descriptions to state that the employee must be capable of lifting 106.6 pounds, not the customary 50 or 55 pounds. We don't carry in one box of cable and whether you use a cart or not, someone’s going to lift and carry two boxes at once. Cable folks, you will become similar to drywall guys carrying two buckets of mud.
Then I want to remind folks that "real time" isn’t real time at all because there is always "delay/latency" present. Again, marketing is to blame. The industry has also told you voice is "just an application." Real time communications means that the network must behave differently than in the past. Each step of the way something else is always required--bigger cables, faster switches, zippier routers, fatter pipes and now session border controllers to secure SIP trunks. More importantly these changes and new requirements cannot avoid or ignore constant improvement, new ideas and ways of doing things. Thus, shortening the length of your future cable drops is one place to examine, especially those tasked with adding new wiring.
For a couple of years we’ve been hearing new terms relating to peer-to-peer (P2P) and SIP such as pods, islands, communities of interests and such. Is it possible to flatten your closets and disperse them? The closets and centralized closet theme is old, and isn't it time for some improvements?
Another question I have is why don't networks and what’s running on them become more socialized? By that I mean two things. First, the gear must play well and better than before without all the lip service to interoperability. Then, not all gear acts the same with long cable drops and not all gear behaves properly when another vendor’s gear resides in the network.
Secondly, do networks really recognize those pods, islands and communities of interests by limiting their traffic within, and route traffic external to those interests only when required? Communications is no longer limited to just voice or data or point to point. Anything and everything connected for man, machine and devices of various sorts will be connecting to your networks.
These pods, islands and communities of interests could interconnect via fiber. I've pondered over the concept of using a "continuous fiber" meaning a fiber backbone laid into the building infrastructure that would be central on every floor, minimizing distance required for the pods, islands and communities of interests to connect. Connecting copper to pods, islands and communities of interests would physically reduce the copper distance, then connect to the continuous fiber.
A little more dreaming is hardware designed to be "intelligent," to absorb the function of switching, routing and whatever it takes to place traffic on the community of interests and external to it or for alternative routing. You could end up with a more dispersed network and more locations and new sets of problems, but is there a way to effectively flatten the closets? What can you do to improve the network, reduce time, waste, rework and facilitate faster turnaround of any cabling changes?
There are plenty of other ideas, and I'd love to hear about yours. I think one idea is evident--cabling infrastructures must change and we must consider how we use cabling more effectively. You may also want ponder over how much money you’re telling your company folks you need in order to sustain what you think you require to maintain the company communications. You see back in the day, cable plants lasted 30 years or more, not 10 or less. Today you have to ask--what will touch the network?
Maybe adding air-cooling fed into the ceiling grid to keep the cables cool is your next quest but then again maybe the cable manufacturers will see revenue opportunity to include a cooling feed alongside them. Seriously, if Cat6A is any indication of past trends and I think it is, then you should be able to imagine what Cat 10A cable is going to be, look like and weigh. Kind of gives new meaning to what we call "bulk cable." So readers of the last post, you’re right--I’m citing other things besides delay/latency. Then, read Latency & Jitter by Kevin Davis and examine every element including the medium(s) you use for your network and where to look to make improvements.
There’s also an interesting write up on Ethernet over at Yale by Howard Gilbert and he states:
An Ethernet station sends data at a rate of 10 megabits per second. That bit allows 100 nanoseconds per bit. Light and electricity travel about one foot in a nanosecond. Therefore, after the electric signal for the first bit has traveled about 100 feet down the wire, the station has begun to send the second bit. However, an Ethernet cable can run for hundreds of feet. If two stations are located, say, 250 feet apart on the same cable, and both begin transmitting at the same time, then they will be in the middle of the third bit before the signal from each reaches the other station This explains the need for the "Collision Detect" part. Two stations can begin to send data at the same time, and their signals will "collide" nanoseconds later. When such a collision occurs, the two stations stop transmitting, "back off", and try again later after a randomly chosen delay period.
I’ll say it again: Shorten your cable drops! I know it may be unsettling but I want you to think like a plumber for a minute. Copper is what it is and you can monkey with the sheath, shielding and wire gauge, but when you run long drops you are creating what I call a cumulative effect that ultimately slows down the network and causes re-work.
This doesn’t mean that Category 6 won't work or work well, because it will perform much better. The basic concepts I’d like you to walk away with and consider are:
* Copper is expensive , more copper usually means better cable that equates to more weight/space used and higher costs; yet copper remains in demand
* Will the network or unforeseen requirements outgrow Category 6A wiring requiring yet another cable plant makeover?
* Is the existing old cable plant and closet system adequate for next generation networks?
* Will the MACS (Moves, Adds, Changes) expense go wild over these new bulky wires and is this type of cabling sustainable ?
* How close to "real time communications" are you?
Cat 5 and 5E worked pretty well considering, but not as good as what some of you think. Remember one other factor and that is the ability to communicate with devices and machines. While you may never hit true "real time," the faster pipe with more throughput and less rework may win a deal, beat a competitor, get your late payroll data delivered just in time or defeat an enemy. Don't design latency and issues into your networks, but design it out.
Whatever the cable plant of the future looks like, it’s going to need a makeover, and sooner is better than later. Category 7 or "F class" cabling is already on the board and when and if a Cat 10 cable standard ever materializes, you can know that I'm just glad that I’m not the one 'pulling' over the ceiling what it is that you now think you need for "real time communications."
