The company is headquartered in San Jose, California, with a software group located in Beaverton, Oregon. Around the fourth quarter of 2001, the company plans to upgrade its present "limited" facilities.

Cinta currently employs 130 staff, of whom approximately 75 are engineers, 13 or so of those having optical-related PhDs. Depending on market conditions, by the end of 2002, Cinta expects the total to reach around 200.

Cinta presently has a strategic partnership with Solectron, which is building 75% of Cinta's boards. By the end of the third quarter of this year the percentage should increase to one hundred and include optical assembly. Cinta does not therefore intend to manufacture any product itself. ADC provides Cinta with post-installation and field support.

Cinta has a direct sales organisation within the U.S. but expects third party sales support throughout the rest of the world.

The company has "about" ten different patents in the process of being filed; a couple on the system architecture itself, others on use of the optical components, and a further few on restoration algorithms. By the end of the year, Cinta expects to have filed about 15 patents.

Cinta Networks has raised $75 million in series A and B funding; $51 million in equity, the rest in debt or lease financing. Sequoia Capital and Battery Ventures are the single largest shareholders. Comdisco, who was also part of the financing package, has since filed for Chapter 11, but this is not expected to impact Cinta, as it received "everything bar a few dollars" that had been committed. The present level of funding will take Cinta through to early 2002, and round C funding is being sought.

According to John Vaughan "In a purely optical domain, you cannot do that - only in an electrical domain can you do an insertion. Therefore, by having the one O-E-O, a lot of additional network elements, handoffs and costs are cut. The most unique thing about us is that the initial cost (of the first lambda) can be up to 65% less than any other solution that is currently on the market."

When fully populated, with 1024-ports at 10 Gbit/s, which Cinta currently offers in the first release of WAVEJunction, the cost is still claimed to be over 50% less than alternative solutions.

"It is pretty simple. The DWDM and switching are pretty logical network functions to integrate, because every switch has to have DWDM in front of it - you cannot have one without the other. This is because, in our long-haul space, what feeds the switch is transport, or DWDM. This is what makes the integration logical."

Mr. Vaughan said that this is a unique approach only because optical switching is a fairly recent development. With reference to his past employment, Mr. Vaughan said that Cinta's system is performing the same kind of function, in the purely optical domain, as the Titan 5500 product from Tellabs. "But at OC-48 and above, the most efficient way of managing the bandwidth is in the optical environment."

Cinta believes, in fact, that it provides the best of both worlds, with the electrical function for performance monitoring and the actual switching fabric being optical. MEMS-based optical switching has not been used, as Cinta does not believe the technology to be reliable or proven.

Another factor that John Vaughan claims differentiates his company from the rest is Cinta's ability to grow switching one wavelength at a time, from a handful all the way up to 1024, a port card at a time.

Mr. Vaughan gave an example of the capability by describing a service provider having 64 wavelengths coming in to the node on a single fibre, but only needing to switch 4 wavelengths, the other 60 being pass-through point-to-point. In that situation, Cinta inserts one card to switch the 4 wavelengths, and the other 60 are allowed to pass on through. Cinta enables the service provider to implement the WAVEJunction system as transport only, and then build the switching capability up a card at a time, as the requirement grows.

"This gives the service provider ultimate operational and financial flexibility. We have a 10 Tbit/s switch, which is enormous by today's standards and clearly has a bandwidth manage capability. It does all network functions such as provisioning, reconfiguration and restoration. It does all of this at the wavelength level."

Network management
Cinta takes a distributed approach to managing and monitoring the network. For a service provider with say a widely distributed 50 node network, each node includes a control card containing all the intelligence, enabling configuration down to the component level and traffic activity control. The network can therefore be managed or monitored out of any node. "This also allows us to operate a mesh or a ring topology, with SONET/SDH restoration times."

Cinta is patenting its restoration algorithms, and notes particular success in the mesh environment. "The only thing that keeps it from being at 50 milliseconds or less, is the traffic length, it is not a function of the system." For example, on a nationwide network, if the break is on the west coast, and restoration takes it back through the east coast, the issue is one of distance, which the system cannot overcome.

Mr. Vaughan also commented that despite analyst reports, "If you don't have a SONET-based solution, you don't have an answer. SONET is not dead and buried, there is $75bn dollars invested in it already in North America alone, in equipment, procedures, training etc. With ten years development and continual refinement, there is still no better standard for restoration than SONET or SDH."

Tuneable components
Initially the adoption of tuneable devices was regarded as a technology risk, though Cinta believes the risk element is now behind it. One customer lab trial has been completed with tuneable devices, a second trial will start at the end of July and a third is planned in September.

John Vaughan said there were two reasons why Cinta decided to employ tuneable lasers. Firstly because of the amount of sparing and inventory, with Cinta having only two card types, one for 2.5 Gbit/s and one for 10 Gbit/s. Secondly, Cinta's solution has a non-blocking switch; if there is any kind of contention, traffic can be re-routed dynamically with tuneable lasers.

Mr. Vaughan said that without tuneable lasers Cinta would not have been able to cost-effectively carry out the integration as a low risk design for its product. "Without question, the tuneable laser is a key component of our architecture and our product design." Cinta said that to date it has bought "easily hundreds" of these lasers.

Cinta sources tuneable lasers currently from Agility and Altitun whose components can be interchanged at the manufacturing line because they are identical in terms of form and size. Cinta chooses to have at least two suppliers for all its key components, and often three, with the ability to interchange between them a consistent feature of its manufacturing approach.

"Our engineering team is very adept at working with a multitude of vendors and keeping track of who has got the best and the lowest price". As a result, Cinta has set as a corporate objective that, at a minimum, it will drive 35% out of the product cost each year. This year Cinta says it will beat that objective.

Long-haul
While WAVEJunction presently can reach 600 km without regeneration, Cinta has set itself the target of achieving 2000 km by the end of the year. Cinta notes that above this range, the ultra-long-haul space is a very small addressable market with a different technology and different cost model. After the 2000 km milestone, Cinta therefore intends to focus on the lower end of the metro market.

A further development will be to enhance WAVEJunction with reconfigurable optical add/drop muxes.

"These milestones will not add more functionality necessarily, but enable new ways to be found of dramatically lowering the network cost for service providers."

Testing
The actual testing of an optical system is not long in execution, Cinta said, though preparation takes time. However, the largest chunk of elapsed time in the process is interoperability testing with other vendors' equipment.

Even if a customer has a production system already running with live traffic, all the intelligence can be offloaded from each card to the virtual system, enabling customers to "play whatever games they want with it."

Mr. Vaughan noted that only 5% of fibre, in aggregate, is lit in the U.S., so there is a lot of fibre still to be utilised. As this new fibre is lit and as the service provider looks increasingly towards ROI instead of revenue growth, he asserted that Cinta has by far the most cost-effective approach.

The "perfect application" for Cinta is where a whole new route is being built, either by new installations or the leasing of dark fibre. If a customer does not need wavelength switching for say 6 to 12 months on any given node, the fibre can be lit with point to point transport, and the switching capability controlled one port a time. As Cinta employs universal cards, the inventory can be used on any node in the network for any of the wavelengths, which Mr. Vaughan described as a "hell of a feature".

Cinta also believes it can offer a competitive solution for existing networks. "If you've already got a lit fibre, odds are you don't have an optical cross-connect on it, but you will have DWDM as the transport. If you need switching capabilities, you can install a product such as those available from Tellium or Lucent's Waverouter. However, these are very expensive. We can go in with a scaled down transport and terminate using a short reach interface into our switch."

The customer is, in this situation, paying for Cinta's DWDM as well, and therefore carrying some of the overhead, but the transport can be scaled down and the switch scaled to 1024 ports.

However, Mr. Vaughan said this was understandable, carriers want network reliability, and are not prepared to waste time on a company that may not be around in 6 months. Cinta asserts that it has only just turned that corner.

With this type of product the sales cycle is naturally a long one. At the moment, Cinta said it is responding to four or five RFIs, but notes that the long distance carriers particularly, talk in quarters rather than weeks or months.

A further issue for Cinta is the uniqueness of its architecture. "People have to stop and think about it. If they do think about it, they think it is logical and helpful." Because potential customers had concerns about Cinta's ability to scale and post-sales support, Solectron and ADC were brought in.

Mr. Vaughan also noted that softness in the market means that carriers are looking and doing a lot of evaluation, and that Cinta intends to position itself so that it is a recognised player by the time the market turns around.

Cinta perceives its general market as being North America, and countries like Korea and Japan, with Europe slightly less of a priority.

Cinta's target is to hit just one PO by the end of 2001. "Nortel and Cisco look at a two million dollar contract and laugh at it, but to us every pound is a pound, and it adds up to real money real quickly."

Cinta also claims a unique performance parameter for WAVEJunction's integrated systems approach, which enables service providers to come to market quicker.

Whilst this integration and the product's scalability would appear to give an increasing advantage with larger systems, Cinta also claims it is highly competitive at the single lambda level. What the key elements are which have enabled Cinta to achieve this DWDM/optical switching integration or why others have not chosen, or so far achieved, this particular route is not too clear.

The company's major opportunity, rather than looking at a niche, is in aiming to compete in the multi-billion dollar market for mainline optical systems with companies such as CIENA, Nortel and Sycamore. The corresponding down side of pursuing this substantial opportunity, as Vaughan has acknowledged, is the problem of competing with the very solid brand images of these companies as well as their hefty marketing clout.

Cinta appears to be claiming that the alternative architecture of WAVEJunction presents no technical limitations within optical networking, except in ultra-long-haul applications - though conceivably a lack of the latter functionality could be a drawback for larger operators keen on working with a single supplier.

Assuming no inherent undisclosed limitations of the product itself, Cinta's problem is essentially one of marketing and promotion. Cinta also needs to make sufficiently credible its operational capability and support, including interoperability with other installed systems, over an inevitably limited design window. This window Cinta does not attempt to scope but realistically cannot be assumed to be more than twelve to eighteen months.

However good the product may be, to develop that degree of marketing momentum, from virtually a standing start, against companies such as CIENA that have been in the business many years and are highly esteemed, is a pretty tall order and will require a marketing programme of very high quality.

Indeed, if there is a criticism of Cinta to be made, it could be said that the marketing aspect of the business has been launched rather late considering the focus on the mainline network business. If for example Cinta had invented a niche for itself, this would not have been as much of a problem, but to elbow Nortel, Sycamore, CIENA, and perhaps Corvis, out of the way will take some doing.

Probably this is a fairly black and white issue - either Cinta can convince the market that the company is a technical market leader and can not only invent but survive long enough to deliver on time, in volume, with full operational support - or it cannot.

No doubt Cinta has been very focused on getting the product issues perfected, but in the time-compressed and ultra-competitive environment of today, the marketing programme probably needed to have started even earlier than the engineering programme - it takes much longer, and more money, to establish a major brand than invent and build a super or even super-super widget.

Some might consider this statement surprising given the fact that the company has been widely noticed and included in lists of "most likely to succeed" startups. However, there are an awful lot of startups, and Cinta requires much more than a few patronising pats on the head by analysts who have no particular track record of picking winners.

In the circumstances, Cinta's recent recruitment (July 9th 2001) of Bob Cantarutti, the former Vice President of optical networks and access sales for Nortel, as Vice President of Sales, is a great move. This is about as close as the company is likely to get to the optimum, other than recruiting Patrick Nettles, in its attempt to give the business a good chance of success.

As Vaughan has also acknowledged, there is an element of luck at play here related to how long it is before the market for optical networking systems recovers. If the start of that recovery lies significantly outside the twelve to eighteen months that we think Cinta may have in which to make itself credible, then the company's prospects of success will (along with many others in this sector) obviously be weakened.