re: bandwidth, cache and internet time: the perishables of domination
Ray, the browser connection is especially interesting to me, and now that they have been acquired and undergone a form of entropy, of sorts, losing their sovereign corporate identity and all, I can comment a bit.
I consulted to a group that was setting them up, actually repositioning them onto the 'net, about three years ago, and wrote the RFPs for their wide area Sonet and CAN (Campus Area Net) networks. Talk about demand creep. At the time they were going over the top with between 100 to 115 million hits per day, due to the nature of the browser wars and borrowing from the nature of users' first screen login dynamics, primarily, but what a load it created at their I/O points, back then. It seems like twenty years ago to me, now. Internet time. It's as though when we did this I was listening to the second coming of The Doors during the Late Seventies.
I had some difficulty convincing them that they would max out before the coming of the next moon if they limited themselves to an OC-12 outside envelope, which is what they wanted to do. Investors were breathing down their necks, as I recall, and cost was the primary motivator to them. At the very least, it afforded them a defense with which to poo poo my bandwdith orientation, and "paranoia."
"Everything" had to come back to CA, in their eyes, and the concept of mirroring and caching [M and C] content were still deemed bleeding edge and NIH, or not-invented-here, which was a very important internal political consideration to them, and others, at the time. And who is to say that in the end they will not be proven right, when there's sufficient bandwidth to justify this mode.? Not me. [Hi Ed!]
In retrospect, I could see why they may have thought that I was burning both ends of the candle. I was both advising them to up the ante on their ring sizes, AND, at the same time, I was talking about mirrors, reflectors, and caching. I could see now where I may have been sending mixed signals. But, both were needed eventually, so I don't feel so bad. I needed to say that. For catharsis.
You might recall this as the time when the larger players were involed in a new form of race, which persists to this day, to see who could draw more hits per day. Who do you think their competition in the hit parade was in '96-'97? Was it MSFT? .GOV? Does anyone here recall who the top five were?
The scale was unprecedented. And M&C were only vaguely understood, and even when understood, only remotely necessary at the time, as they saw it. Throw more bandwidth and back haul all traffic, everything, to the lot where all the IPO cars and MRVs were parked, was their motif. At the time, DWDM was only first being demo'ed at SUPERCOMM, with a topping-out capability of a paltry 4 to 8 lambdas as near term futures, by most vendors. Times: They are changing.
They very politely rejected my admonitions, initially, and given the newness to them of the turnpike principles they helped to validate, I really couldn't be too vociferous nor blame them. But I took the initiative to spec in enough head room in their cabinetry and carrier framework to allow them to expand to at least OC48, anyway. And guess what? The OC48 itself turned out to be not enough, in the end.
Bandwidth. What the hey, it's perishable anyway. Right?
Regards, Frank Coluccio |
