40G long haul, Clouds, and IPv6

Communications Infrastructure

Communications infrastructure is one of the underlying systems which make daily living possible; like the supply of energy, and roadways, it makes many of the things we take for granted easy.

The current data communications infrastructure continues to develop, and to make more things possible. The process tends to operate in jumps: first it is demonstrated that something new can be done (like sending data at 40G bps between the USA and Sweden) then, provided that there appears to be the opportunity for some commercial return, the iniital ideas are developed and modified so that they can be manufactured and sold in volume. This typically lowers the price - and in the networking business, makes a 'network effect' possible, so that the worth to society of having many of a certain thing is more than just the simple multiple of the number of things.

There are at least three big change points going on in the networking communications business at the moment. Each of these areas has existing businesses competing - the really interesting opportunities are likely to be created at the junctions between them.

Long haul

It is just becoming possible to run long distance trans oceanic network connections at 40G bps in production. Each is still very much a custom installation, requiring hand tuning, but it is  possible to buy a 40G service - and it is reasonable to expect that competition will encourage both greater supply and lower prices in a year or two. The next target is commercial 100G bps long haul service.

Data centers

Changes in data center operation have been called by a variety of names, including outsourcing, cloud computing, cloud
networking,  virtualization, SaaS (software as a service), and PaaS (platform as a service). What is happening is a swing away from locating servers near to the people using them and towards centralization of servers in a smaller number of (sometimes very) large locations. This often is combined with centralization of control (e.g. many institutions depend
on Google mail).

The elements which affect the location decision are

  • Cost of people
  • Cost and availability of energy for operating and cooling
  • Cost of servers
  • Latency between servers
  • Latency between users and servers
  • LAN  bandwidth costs
  • WAN bandwidth costs
  • Server and network installation time

There is a unsatisfied requirement for a dynamic protocol enabling constant bi-sectional bandwidth network connections between thousands of servers in data centers. 802.1 Data Center bridging, TRILL, BGP#, VL2 and Valiant load balancing are all works in progress.

The public Internet

The public Internet continues to grow, making it possible to sell directly to, and to sell access to, ever increasing numbers of consumers. (Businesses like Twitter, Facebook and YouTube, none of which existed five years ago,  rely on the public Internet). The number of IPv4 addresses still available is decreasing - more Internet Service Providers are  looking seriously at supporting IPv6. This is a disruptive opportunity.

Further reading

Sprint 40G transatlantic trial press release
Tyco trans Pacific 40G press release
Microsoft data center in Dublin
VL2 - Microsoft Research
BGP#
TRILL RBridge
IPv6 transition factors