Rocket Silence

Brain-teaser time: is it wiser to invest in network upgrades or to purchase traffic-shaping equipment in a futile effort to combat the inevitable?

This is the issue I can’t seem to understand. At UC Santa Barbara, there is a pool of 200Mb of downstream bandwidth allocated for about 7,000 students. The logical response to that statement is of course, “WTF?”. Let’s run through a few hypotheticals. Assume that a tenth of the students are watching a YouTube videos or other streaming media (the average FLV stream is 300Kb/s); that consumes the entire bandwidth pool and leaves all other HTTP traffic only to slow to a crawl. Let’s assume you throw in some rogue P2P file sharing and the infinite Facebook clicks, it’s clear that it isn’t enough. The uptick in bandwidth consumption per student is not unwarranted with the shift in how individuals consume their media and it is only going to continue. Now, as a network administrator, how do you solve the issue? By spending tens of thousands of dollars (per device) for traffic-shaping equipment that prioritizes other traffic over others, deprioritizes ‘recreational’ activities to create the illusion of a functioning network.

That is the case here. A common cable equipped-houshold has access to to upwards of 10 megabits of downstream bandwidth. One can argue that this is not guaranteed, but that’s beyond the point. Let’s say that’s functional for a household of five average internet users. That would suggest the average user requires two megabits each. Let’s divide that by four since we assume most users won’t be constantly pulling data or won’t be online at the same time. By this math, a campus with 7,000 residents would need a 3.5 gigabit downstream link for acceptable performance. Compare that to 200 megabits; the current network capacity is less than one-seventeenth what it should. I don’t expect some type of miracle due to both budget and infrastructure limitations, but an honest evaluation leads on to believe that devices from Packeteer and Procera that cost upwards of $20,000 per box is not the best course of action and merely offsets the inevitable.

Links: Procera Networks, Packeteer PacketShaper, Traffic-shaping at Wikipedia.