Optical startup Coriolis Networks Inc. (www.coriolisnet.com) today unveiled details about its products, which it says enable service providers to recover 75 percent of stranded bandwidth generally found in SONET networks and to parse out bandwidth as desired for voice and data applications.
Bob Castle, president and CEO, explains that the SONET add/drop multiplexers and optical networks in place in today's public networks are very inefficient in providing bandwidth. This is true both in the large chunks of bandwidth they provide to all applications - whether those applications need all that bandwidth - and in the manually intensive way in which providing occurs, Castle says.
Coriolis addresses those problems by providing service providers and their customers the ability to access bandwidth on SONET and WDM connections at a more granular level and with far greater ease, he says.
The company's OptiFlow Network line of products is based on a architecture Coriolis calls Optical Spatial Division Multiplexing, which shapes the way information is packaged inside SONET and WDM transport. Voice still is slotted into virtual tributaries as with traditional SONET solutions. But instead of using only traditional SONET hierarchies for data, Coriolis recovers stranded bandwidth by breaking down the SONET or WDM bandwidth into a contiguous block of bits so that customers can set up bandwidth "frame relay style" for committed data rates plus bursty data.
The products statistically mux bursty traffic so a service provider can oversubscribe the bursty data. As a result, service providers can offer different classes of service, or can look at packets themselves and discover the kind of service required or who originated it, and decide on the fly what priority to give it. The system prioritizes traffic using MPLS and DiffServ quality of service standards.
Coriolis' products also do domain wide bandwidth management, explains Castle, noting that a domain is an entire metropolitan ring. This technology looks at all the data streams from all the nodes on the ring and manages requests for bandwidth instantaneously from any node.
"In our system, instead of just a series of nodes on the network, our approach is to have a logical star, with one node with intelligence that doles out bandwidth on an as-request basis," Castle says. "So what we're able to do is accumulate the best effort variable traffic from many nodes and we could oversubscribe an OC-48 by a factor of 2, 3, 4 or 5."
Coriolis products also have the ability to transport information in native protocol form, Castle says. "We are not converting different kinds of services to some alternative form for purposes of transport from egress to ingress" as is done in ATM, he says.
However, the products do offer interworking between such protocols as Ethernet, frame relay, ATM and time division multiplexing (TDM). The difference is that Coriolis "decouples" the services that the carrier presents to the customer from the backbone the carrier uses to connect its sites. That's so the carrier, rather than the end user, is the one that has to do the management of the transport protocol.
Castle says what's happening in the optical network using products such as those from Coriolis is analogous to what happened a few years ago with the transition from dedicated to switched carrier services when customers used nailed up circuits and paid for them regardless of whether they were using them. The switched model allowed customers to send traffic through the network "cloud" and pay only for the bandwidth used.
Likewise, bandwidth on SONET traditionally has been provisioned for certain bandwidth between two points on the ring and once a SONET add/drop multiplexer is provisioned to offer a set amount of bandwidth it tends to stay that way, says Castle, so there's a lot of wasted bandwidth. The new approach is treating the optical network as a "cloud" where all nodes connect to the network and any location can request bandwidth as needed.
Castle offers up a couple scenarios of service providers that might find that approach attractive. Consider a CLEC who today has heavy TDM business but some data services. That CLEC is concerned because new generation service providers in its area are offering high-capacity Ethernet loops. So the CLEC needs a platform to protect its TDM business and yet offer a scalable, flexible approach to packet services with the same cost structure as the next generation carriers. Castle says the Coriolis solution fits that bill.
The flip side looks at it from a next-generation provider standpoint. That service provider might be concerned about gigabit Ethernet connectivity becoming a commodity service and be looking to offer higher margin offerings. The Coriolis product line would enable them to support services such as voice over IP with the carrier-class functionality of traditional voice networks that allow for 50-millisecond delivery and SONET-like reliability.
And because of the network management software Coriolis offers, a service provider also could provide bandwidth to multiple retail carriers, dividing an OC-48 arbitrarily into chunks for various carrier customers. The network management system would enable each carrier customer to securely manage its own bandwidth.
The network management software also could be used by corporate customers of service providers to provision their own bandwidth and make sure their service provider was delivering bandwidth as promised, Castle notes.
The company expects to deliver its products to its first customers this quarter, with general availability planned for the first quarter of 2001. The products include two trunk unit options, which would do aggregation and act as the central node of a logical star network, and three service unit options, which would sit at the customer premises to serve a single customer or an entire building.
The products support up to 16 OC-48 rings (or eight protected rings); or two WDM rings with 16 lambas; or a mixture of OC and WDM. Cards in the trunk unit and service units can allow a service provider to upgrade individual locations to WDM without having to change cards in any other nodes in the network, Castle says. Packet and circuit switching fabrics also exist in each of these nodes, reducing requirements for switching in central offices, he says.
"Our goal is to provide only big chunks of aggregated traffic to COs," Castle says.
The service nodes start at a cost of below $10,000. A total system can be bought for less than $6 a megabit. "These are pretty aggressive numbers," Castle says. "A lot of people are trying to figure out how to get below $10 a megabit."
