2 ms·
First I'd increase my TCP window size. Otherwise I'd be latency limited :) Being an IT Director I try to think of what I can't do without my local 1Gbps LAN. I
by johngalt 17y ago
First I'd increase my TCP window size. Otherwise I'd be latency limited :)
Being an IT Director I try to think of what I can't do without my local 1Gbps LAN. I'd say a big one is computer re-imaging. With that type of band you could build a good reference OS install and push it anywhere in the world that had a similar band.
I'd also turn this around and say "how will the content providers deal with that kind of demand?" Serving thousands of users at those speeds is a non-trivial task. The bandwidth demanded from the server is a multiple of the bandwidth increase to each client. If you solve this, your customers will have big money on the table. Rather than figure out the "what" will go over those links I'd figure out the "how".
With the current applications I'd say latency is a bigger issue. 100Mbs at <50ms is more useful than 1Gbps at ???ms. I don't buy the idea of "dumb terminals" AKA "web OS" without less latency. The biggest reason I use RDP, VNC, or Citrix is to trade a bandwidth issue for a latency issue, so why would more bandwidth == more terminal usage? If anything I'd expect the reverse.
- wmf 17y agoServer performance has outstripped Internet performance for so long that I don't think this will be a big problem. A $2,000 server can probably serve 10 Gbps from RAM... if you actually use your processor power instead of wasting it. Throw in cheap private peering and CDNs, and you can solve the bandwidth and latency problems at the same time.