What if we put routers in satellites?

In this industry, there's a lot of talk about branch offices and satellite sites, but the notion of satellite sites are seldom taken literally. However, we recently received a query from a reader asking for our opinion on putting routers on satellites and what type of productivity we might see.

Let’s start with some basic arithmetic. A geosynchronous satellite - one that orbits the earth at the same speed as the earth rotates so it appears to stay at the same place in the sky - needs to be roughly 22,500 miles above the earth. And the radio signals travel to and from the satellite at about 186,000 miles per second. This means that a round trip to the satellite and back to earth (about 45,000 miles) takes about a quarter of a second. And for a transmission to make a full round trip - originating point to a destination router and back to the original point (or another destination) - takes roughly half a second.

So the reader’s hunch was right. If the router were on the satellite, then the information could be sent from the originating site to the ultimate site in a single quarter-second trip rather than having to make the extra round-trip back to the ground to be routed.

Even so, we see some severe caveats and think this might be limited to special situations. The inherent delay can wreak havoc on protocols such as TCP where a defined amount of information can be outstanding and unacknowledged. This number of “bits in the pipe” is calculated by taking the transmission delay and multiplying by the transmission speed (time multiplied by bits-per-time). So as speeds increase, the amount of outstanding unacknowledged data increases proportionately.

And even though you can define a very large windows size, it’s still a tradeoff. You can have more bytes outstanding, but if you have to retransmit, you're totally up the proverbial estuary without viable means of propulsion. Even with selective retransmission, you have to wait for that information to be retransmitted, still introducing significant delay.

So we think this technology will be primarily limited to:

- Traffic that is not extremely sensitive to delay.
- Geographic areas where optical fiber and similar terrestrial transmission is not practical. (It appears that the inquiring user was in a group of islands in the South Pacific.)

Also we must point out that whatever router is used in this type of application has to be hardened for extreme temperatures, and that swapping out a board can be an inconvenience, to say the least.

But we love imaginative questions and encourage you to keep asking them. And if you really want to let your imagination run wild, refer back to our discussion from April 1, 2004 of graviton-based transmission for core networks.

Read more about lans & wans in Network World's LANs & WANs section.

Steve Taylor is president of Distributed Networking Associates and publisher/editor-in-chief of Webtorials. Jim Metzler is vice president of Ashton, Metzler & Associates.