how-to block ads
|
|
Share Topic
 |
 |
|
|
 espaethDigital PlumberPremium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
 ·Clear Wireless
| reply to funchords
Re: BT comments Redux said by funchords:This is a very recurring theme. It comes in waves, but often struggled with something new sucking up the capacity. In the late 80s, it was large FTP archives. In the early 90s, it was images (pictures). Five years later, it was the web-boom itself and the popularization of online music. This "crunch" IS NOT ABOUT P2P*, it's about video. It's not just about P2P, it's not just about video. It's about ANY technology that causes a drastic shift in the oversubscription model on which networks are built.
said by funchords:How did we get out of each one? We grew the network's capacity. Nothing else truly works -- it all amounts to denying or delaying service in one way or another. To meet increased demand, increase supply. To solve your debt problem, acquire more money. If only it were always that simple.
You can only grow your network once the necessary interfaces become available, and occasionally you have to push through large capital expenditures when a full refresh is required to improve capacity.
For example, right now I'm stuck building out capacity between data center locations using N*10GigE links because that's the only common link protocol my intermediate DWDM providers will support (outside of ungodly expensive OC768 options). The thing that is annoying about this is that traffic hashing algorithms aren't perfect, and to be able to sustain 22gbps of traffic I actually need 4*10GigE links in order to get the traffic to balance out. (on a 3 link hash one of the links would usually end up hitting in excess of 9.5gbps) 40GigE and 100GigE will make my life better, but the standards aren't set to be finalized for those options until next year, and even when they are finalized I'll still need to replace entire chassis for cases where my switching backplane doesn't have room to support 40/100GigE interfaces.
Upgrades take time, just look at DOCSIS 3.0: that spec was finalized in August 2006 and we're just seeing the rollouts starting now.
You're right in that increasing capacity is the only long term solution, but the problem is that today you can only deploy technologies that exist as actual products, and they need to fit within your capital budget. | |
 funchordsHelloPremium,MVM
join:2001-03-11
Yarmouth Port, MA
kudos:5 | said by espaeth:It's not just about P2P, it's not just about video. It's about ANY technology that causes a drastic shift in the oversubscription model on which networks are built. Well that's not very new, either. I think ISPs picked up on the fact that users were pretty keen about surfing, but they also knew a large number of users were also keen on putting up their own websites and photos. ISPs responded to that by providing well-hosted sites and kept that heavier and repetitive upstream traffic off of the last mile. It was a tactic and it worked ... for a while.
The uplink demand was there, and the ISPs mitigated it. But their decision depended on the Internet users' uplink tastes not changing.
But I'll disagree even further up the argument and reaffirm that it is about video. If you removed video from the net, there's plenty of room for everything else -- down AND up. I may gripe that the ISPs haven't grown fast enough, but they've grown.
said by espaeth:said by funchords:How did we get out of each one? We grew the network's capacity. Nothing else truly works -- it all amounts to denying or delaying service in one way or another. To meet increased demand, increase supply. To solve your debt problem, acquire more money. If only it were always that simple. You can only grow your network once the necessary interfaces become available, and occasionally you have to push through large capital expenditures when a full refresh is required to improve capacity. Which is also not at all new. All of those truths of today were truths during those crises as well.
And we did try different other things to defer the time and expense. We put up proxies and caches. We crunched images. Ultimately, though, we grew the network and turned off those compromises.
--
Robb Topolski -= funchords.com =- District of Columbia -- KJ7RL | |
|
