Now, however, Corvis has gotten some direct competition. A U.K. startup called Ilotron Ltd. has developed a similar (and possibly better) package of technologies with exactly the same goal.
Both Ilotron and Corvis say that the only way carriers can build a genuine all-optical network is to buy a complete, integrated package of optical switches, ultralong-haul DWDM (dense wavelength-division multiplexing) transmission equipment, and management software from the same vendor.
This is because every route across an optical backbone has different characteristics, and those characteristics can change over time. As a result, the idea of point-and-click provisioning of connections over optical backbones has to go a lot further than simply mapping out a route between switches. The transmission equipment on each span needs tuning and may need regular adjustments after being set up.
Right now, that’s a laborious manual task requiring engineers at either end of each span to tinker with equipment while monitoring the signal quality with test equipment.
Ilotron and probably Corvis (it won’t talk about its technology) have eliminated this requirement by embedding measurement probes in their equipment so that the tuning can be done remotely as well as locally, via management systems.
While Ilotron and Corvis appear to have identical goals, Ilotron is taking a totally different (and refreshing) stance on explaining its technology and allowing folk to take a close look inside its switch, which is undergoing lab trials at British Telecommunications PLC (BT) (NYSE: BTY).
Light Reading has witnessed these trials and has seen Ilotron’s management system being used to reconfigure its switch and tune transmissions. In other words, it’s definitely real and it works.
All the same, Ilotron is clearly some way behind Corvis. It only got its first round of funding -- $10 million –- one year ago (see Ilotron Ltd.), and right now Ilotron only has a single prototype switch, the one at BT’s labs. However, it’s close to finalizing a second round, which will enable it to build four more switches for lab trials with other carriers that have already been selected, according to Stuart Barnes, Ilotron’s director of engineering. These carriers include “global, European, and American” players, he adds. At present, Ilotron has about 90 staff.
In comparison, Corvis has actually shipped commercial products to at least three customers. As usual with Corvis, it’s tough to pin down specifics. Broadwing Communications (NYSE: BRW) appears to be the only carrier to have bought and deployed a complete package of switches (at least one, maybe more), transmission equipment, and management software. Broadwing is running production traffic over Corvis equipment, according to a senior source in Corvis, who asked not to be identified. Light Reading has asked Broadwing to give details on two occasions in the past month but hasn’t gotten a response.
So, how does Ilotron stack up against Corvis from a technology point of view?
The lack of information about Corvis makes this a tricky undertaking, but Ilotron appears to have a significant edge in a number of respects. The best way of approaching it is to look at the different elements in each vendor’s package.
Table 1: Corvis Gets Some Competition
|Number of fibers||16x16||6x6|
|Fully redundant?||Yes||Not disclosed|
|Switching fabric||MEMS||Not disclosed|
|Status||Operator lab trial||Shipping|
|Wavelengths per fiber||160 @ 10 Gbit/s||160 @ 2.5 Gbit/s or 40 @ 10 Gbit/s|
|Maximum distance||2,000 km on legacy network||3,200 km on new network|
|Status||Operator lab trial||Shipping|
|Monitoring of signal quality||Optical signal-to-noise ratio plus bit error rate||Not disclosed|
It’s important to realize that both Corvis and Ilotron have made switches that sit in the core of optical networks, switching wavelengths. They shouldn’t be compared with switches that pack lower bandwidth connections into wavelengths before switching them, from the likes of Ciena Corp. (Nasdaq: CIEN) and Sycamore Networks Inc. (Nasdaq: SCMR), which have electrical cores in any case.
Ilotron’s switch has an optical core that’s based on arrays of tiny tilting mirrors -- 2D MEMS (micro-electro-mechanical systems) subsystems from OMM Inc., currently the only vendor to be shipping commercial product of this type.
Corvis also has an optical core but won’t say what it’s based on. It has bought OMM subsystems itself but warns against reading too much into this. “We’re testing all sorts of technologies all the time,” says the source in the company who wants to remain anonymous (sigh).
Ilotron’s switch can have up to 16 input and 16 output fibers, while Corvis’s switch has up to six of each.
To be fair, Ilotron is banking on most carriers wanting 8x8 because that’s what OMM is focusing most of its production on, according to Barnes. Ilotron will get a limited supply of 16x16 subsystems, he adds.
At first glance, this makes both Ilotron’s and Corvis’s switch look really small. However each fiber can carry as many as 160 wavelengths –- so Ilotron can handle up to 2,560 wavelengths and Corvis can handle less than half of that, 960.
Things get more complicated still, unfortunately, because it’s not possible to convert one wavelength to another without changing the optical signal back into an electrical one and retransmitting it. This is an integral part of the Ilotron box, but it’s an after-thought with the Corvis switch. Another box has to be bought to accomplish this.
Ilotron makes a big thing out of this, pointing out that it ensures that retransmitted wavelengths dovetail nicely with other streams of traffic that have whistled through its switch without needing any conversion. It’s hard to identify the opposite argument for Corvis’s approach, because it won’t say how any of its gear works.
Ilotron also makes a big thing out of having a fully redundant switch. This means that it’s got two switching cores, one live and one backup, and another bunch of switches and control software to shunt traffic from one to the other at the slightest sign of trouble. It’s also got more granular mechanisms for automatically shifting traffic from one line card to another, so that switching capacity is used efficiently.
Corvis flat refuses to talk about redundancy, saying that this would give clues about its secret switching fabric.
The Transmission System
Ilotron’s current transmission system is designed to carry 160 wavelengths, each supporting an OC192 (10 Gbit/s) Sonet connection. It’s planning on adding support for OC48 (2.5 Gbit/s) because many carriers haven’t upgraded to higher speeds at present.
Corvis can only manage OC48 over 160 wavelengths and has to scale back to 40 wavelengths to carry 10-Gbit/s traffic.
Corvis has a significant lead in terms of breaking distance records. It’s managed 4,000 kilometers in a trial and says it can stretch to 3,200 kilometers in field conditions.
Ilotron is claiming a more modest 2,000 kilometers but claims that it can do this over legacy fiber infrastructure, where EDFAs (erbium doped fiber amplifiers) are typically spaced at 80 kilometer intervals and losses reach 23 decibels per span.
Barnes says that Corvis’s transmission system won’t work on legacy fiber. It’s designed for new networks built with modern fiber, EDFAs at 60km intervals, and losses of 20dB per span. Guess what? Corvis declines to comment on this issue.
Ilotron, it should be said, has yet to demonstrate that it can achieve what it claims. Barnes maintains that it is possible to forecast performance with transmission systems. He contends it’s standard practice with subsea systems.
Ilotron has developed a two-tier management system. At one level, it enables individual switches to be configured and transmission characteristics to be tuned on a span-by-span basis. On another level, it gives operators the big picture for point-and-click provisioning of strings of wavelengths.
The performance of the transmission systems is monitored by two types of built-in probe. One gives the optical signal-to-noise ratio and the other gives the bit error rate. Both are required, according to Barnes.
Corvis? It would be tough to compare management systems even if both sides gave full information. Unsurprisingly, Corvis declines to say how (or even if) it monitors signal quality.
-- Peter Heywood, international editor, Light Reading http://www.lightreading.com