AHhaha, and All, I find all of this "potential" capacity- and bandwidth-handling- discussion rather interesting, considering what most of the larger MSO/CableModem providers' are contemplating and actually doing already to effect "fair weighting" among their users.
Now, there's a concept that you are not unfamiliar with... fairness. Perhaps you like to address the administration of fairness in the last mile?
What we will see, in contrast to this discussion taking place here, and with increasing frequency, is actually the limiting, the rationing of throughput by throttling back on available access speeds to individuals, and through the prohibition of certain forms of link utilization.
This is not an uncommon strategy, as far as the Internet best effort mentality is concerned, and it's an acceptable expedient in this regard, as long as some form of relief is visible in the other direction at some point. If it breaks, fix it. Nothing new here. It's not the approach that was common in the more deterministic modeling of times past, but we adjust due to the overall benefits derived. Usually.
As I've stated upstream on numerous occasions, studies are underway at CableLabs and elsewhere to examine the best forms of relief (re-engineering the distribution plant in the last mile), but I don't know if or when those will be realized.
If I had to guess, these measures will be implemented at some point well beyond the time the pain threshold is reached, where most operators are concerned. Said pain is avoidable, but the price-performance metrics vis a vis quarterly reporting does not support it well.
A real problem that we must contend with going forward, IMO, was the shortsightedness on the parts of the providers in their original rebuild estimates. They will prove to be inadequate sooner than most MSO engineering departments estimated two or three years ago (which is the design-to-implement time frame of what we are seeing rolled out today).
Where last mile bottlenecks are concerned, the Cable Modem model hasn't really experienced egregious performance, yet. Soon, though, I expect to see this become a widespread problem for users whose expectations have been inflated without due cause. The severity of this will be proportional to both subscription uptake and the proliferation of new multimedia applications which are poised to be rolled out, effective almost immediately.
Relief at that point can only be achieved through well thought-out and time consuming re-engineering and reconfiguration initiatives.
DSL doesn't share this kind of prospect, for, with DSL, what you see is what you get more often than not. When it tops out, you can look into the mirror and wonder to yourself where to go next.
But don't look for relief from any technology that will be cost effective for some time to come, other than VDSL at the next plateau, but this will require major fiber builds to the pedestal or other forms of neighborhood nodes, too.
Right now I'm at home using a 56 k modem. It's really quite satisfying compared to most of the LAN-attached 'net access arrangements I use throughout the day, which are attached by multiple T1s at the first tier, and by T3 upsteam. Sharing of resources on Ethernet Lans falls under the administration of fairness algorithms, too. Some of these, like the "back-off" algorithm are spelled out in the very Ethernet protocol itself, while others are under the direct control and discretion of the LAN administrator.
But despite all of this control and gimmickry, they are still inadequate, especially during peak periods, in comparison to my measly 56k's response time at home. They will remain inadequate until they are bolstered with new resources.
Prioritization in the LAN is more easily achieved than in the last mile, by various methods, due to the lack of distance and sharing constraints. For example, upper-management users usually see this first (switched 100 Mbps dedicated links to each desk), while the lower echelons are relegated to more miserly provisions (dozens, sometimes hundreds of users on a shared 10 Mbps segment).
A total upgrade for all users in many large corporate LAN settings is often viewed as overly burdensome from a cost perspective, yielding little in the way of ROI, until things grind to a virtual standstill. When that first major order is missed, or that trade didn't go through fast enough, then light bulbs go off, and some poor slob who has been warning of this along gets his or her ass hauled onto the carpet.
In fairness, I have to say that when I visit the office on weekends and holidays, the screen is lightning fast. But to counter this, these are the times that I probably need quickness the least.
Happy Surfing, and Best Regards, Frank Coluccio |
