It's Google vs. The Speed of Light

Interesting forecast on Google's internet plans- setting up semi-redundant compact datacenters closer to users everywhere, to allow even faster data service. From Cringely found through Internet Stock Blog.

We're talking about 5000 Opteron processors and 3.5 petabytes of disk storage that can be dropped-off overnight by a tractor-trailer rig. The idea is to plant one of these puppies anywhere Google owns access to fiber, basically turning the entire Internet into a giant processing and storage grid.

While Google could put these containers anywhere, it makes the most sense to place them at Internet peering points, of which there are about 300 worldwide.

Two years ago Google had one data center. Today they are reported to have 64. Two years from now, they will have 300-plus. The advantage to having so many data centers goes beyond simple redundancy and fault tolerance. They get Google closer to users, reducing latency. They offer inter-datacenter communication and load-balancing using that no-longer-dark fiber Google owns. But most especially, they offer super-high bandwidth connections at all peering ISPs at little or no incremental cost to Google.

Where some other outfit might put a router, Google is putting an entire data center.

At first blush, we'd find the idea of having to place datacenters all over the country, and eventually world, as a bit anti-Google in terms of remaining more of a software and services operation- lean and mean. But the above author argues that these centers could actually be extremely cheap to deploy.

And whether or not the above scenario is Google's actual long-term strategy, it was interesting to read about, and remains credible in our mind because even in these days of fiber optics, one of the biggest long-term problems the internet faces is that of Latency (communication delay) due to the limits of light speed. Light is just too slow for instant communication. The below is an explanation taken from a 1996 Stanford research paper on "interactivity" by Stuart Cheshire, who is described at his website as "Networking wizard and the author of classic tank game Bolo" As perhaps a sign of the times, we'd also like to note that the paper had been commissioned by an asset management company! Nevertheless its an interesting read.

We're not used to worrying about the speed of light being a problem in everyday life but in this case it can be. [With the goal of interactivity being 70ms or less of latency]

The speed of light in vacuum is 300 000 km/sec and the speed of light in typical glass fibre is roughly 33% less, about 200 000 km/sec. How long would it take to send a signal to someone on the far side of the planet and get a response back?

Speed of light in vacuum = 300 000 km/sec

Speed of light in glass fibre = 200 000 km/sec

Circumference of earth = 40 000 km

Total round-trip delay
to far side of the planet and back = 40 000/200 000
= 200ms

So, sadly, the best response time we can hope to get to someone on the other side of the planet is 200ms, a long way short of our target for interactivity. Interactions with people on the far side of the planet are going to need the benefit of latency-hiding techniques, and true tightly-coupled interactivity will not be possible. We will probably never be able to have a truly interactive virtual sword fight with someone on the far side of the planet.

However, we're not usually communicating with people on the far side of the planet. Most of the time we're communicating with people much closer to home. Might we be able to get true interactivity with people closer to us? Could we do it within the United States?

Distance from Stanford to Boston = 4400km

One-way delay to Boston = 4400/200 000
= 22ms

Round-trip to Boston = 44ms

This result is promising. Within the United States we should be able to achieve true interactivity. The speed of light consumes 44ms of our 70ms time budget, leaving us with 26ms for software overhead in the network, the endpoints, and the graphics systems. Not an easy challenge, but at least it's nice to know it's possible.

So while there's hope for within the US, such a problem does make a Google datacenter network seem more useful. By having information stored redundantly, but near different users, the problem of light speed can be overcome.

Comments

[AnferTuto]

Hola faretaste
mekodinosad

[how record dvd]

Note: Les billets ISS sont une espèce de pot- pourrit des découvertes faites lors de ma veille quotidienne. Pas de niches ou thèmes de prédilections, cela peut toucher tout ce qui se passe sur Internet.

[dvd copy]

I have the same problem as Lauretn. I have a USB data cable but I don't have its driver, I checked the instructions and I found that it need a driver (SER9PL. INF) and it can be found in the http:// tech. prolific. com. tw/ So I cheched the link but it does not work proparly and face some error in accesing inside this link, and can not find the file. Please direct me to the right link where I can find this file.

[dvd burning software]

D- dimer blood test. Having high levels of the clot- dissolving substance D dimer in your blood may suggest an increased likelihood of blood clots, although D- dimer levels may be elevated by other factors, including recent surgery. Drawing the blood takes just a few minutes, and the risks— which include slight bleeding or a small accumulation of blood at the puncture site— are minor. The results are available in less than an hour. Normal test results are actually much more meaningful than abnormal ones. That’ ...

[cloneing playstation dvds]

Register SIP phone (DPH- 140S) to Asterisk Make Skype call from internet to SIP phone Make SIP phone call to Skype Register DVG- 6004S to Asterisk Make outgoing calls to PSTN via DVG- 6004S Receive incoming calls from PSTN via DVG- 6004S Make Skype call from internet to PSTN line via DVG- 6004S Connecting DVG- 6004S The router to internet is properly setup and have access to the internet. The router shall also provide DHCP service. The PC with Asterisk and ssgwpe is properly setup and connected to the router. ...