The speed of backbones is growning faster than the speed of many routers architecture.
The approach of Pluris is to make use of Massively Parallel computation engines and interfaces.
This provides built-in reliability, no single point of failure, and hot-swapping capability.
This architecture is scalable and is believed to have a useful life of 6 to 8 years.
The processing power of the initial product offering will be well above OC-192. The aggregate capacility is expected to be around 2.5 Terrabits/second. Backbone line speed is expected to be around 1.2 Terrabits/second. Of course, there is no available lines that run that fast right now.
The box consists of a number of single board computers interconnected via a scalable backplane interconnect.
It was long believed that this type of this box could not be used in the Internet because of packet re-ordering issues. However, in this box, there is a two step forwarding process. The first process determine which channel to send the packet down and a second process detemines which interface to dump packet out on.
Most point of presences are using a lan technology that hooks up a set of routers used for aggregation or transport. With the Pluris router, only one router is really required. It can be just as reliable and reduce latency varience.
The NAPs could also use the Pluris router since it can be broken down into independent functional units which can operate independently of each other and can talk to each other inside the backplane.
With the single node, 120,000 packets/second. In the prototype, there are 72 of these nodes. Each node contains a computer and an interface card.
The fabric is based on the buttlerfly switch. Pluis has modified it to have some fault-tolerant capabilities.
The fabric is used to connect computers together and interconects other router shelves.
If these routers are used with raw fiber, then backbone capacity will not be a problem. The problem will be the tail circuit capacity.
Vadim's comments are in this font. -- Ed.
When will this stuff be available?
Volunteers are being taken now.
How do you cope with routing problems?
We believe in simplifing the routing complexity.
What about QoS?
The software manages this. We do have some queuing capabilities.
Who are projected to be the customer?
Telephone companies are expected to be the main customers.
For more details, see the white paper.