& cplSiteName &

Ample Crams in the Ports

Light Reading
News Analysis
Light Reading
8/26/2002

When a revision of Light Reading's Top Ten Private Companies was discussed at the Opticon 2002 conference last week, a rather dark horse -- Ample Communications Inc. -- popped up as a likely new entrant to our list of startups most likely to make it to a "liquidity event" (getting acquired or maybe even staging an IPO).

Ample, we decided, tickled our fancy for three reasons:

First, the startup is developing chips that will enable system vendors to offer a fourfold increase in capacity of their existing equipment, by cramming four 10-gig interfaces onto a linecard in place of one. That's music to the ears of carriers trying to improve their return on existing investments.

Second, Ample's founders have an impressive track record. Vish Akella, CEO, was a founder of Kalpana, which developed the first Ethernet switch. He was also a founder of Stratacom, which developed the first Frame Relay and Asynchronous Transfer Mode (ATM) equipment. Both startups were acquired by Cisco Systems Inc. (Nasdaq: CSCO). Ravi Sajwab, Ample's CTO, was a founder of Acclaim, a developer of high-speed compression and encryption processors that became part of Intel Corp. (Nasdaq: INTC) in a roundabout way (it was acquired by Level One Communications, which was itself acquired by Intel.)

Third, money doesn't appear to be a problem. Ample recently announced an additional round of $21 million, bringing its total funding to $38 million (see Ample Gets Ample Funding). It's already shipping product for revenue and boasts of multiple design wins with Tier 1 system vendors. It also has some bluechip backers: Center Point Venture Partners, InterWest Partners, Storm Ventures, and Rho Ventures.

Details of Ample's latest product -- a 40-Gbit/s "frame processor" -- were announced this morning (see Ample Unveils 40-Gig Frame Processor). The device, called "Blackbird," makes it possible to cram four 10-Gbit/s interfaces (OC192 Sonet (Synchronous Optical NETwork) and SDH (Synchronous Digital Hierarchy) or 10-Gig Ethernet) onto a single linecard by consolidating the functions of PHY, framer, pointer processor, mapper, and datalink processor chips into a single device. This reduces space requirements by a factor of four and reduces power consumption (and thus cooling requirements) by a similar amount, according to Ample.

"Over the last few years a lot of new systems have been deployed. But because of silicon limitations, most of the systems are only utilizing 25 percent of their capacity, meaning fewer customers per linecard," says Marek Tlalka, Ample's VP of marketing. Swapping out single-port cards for Ample's four-port cards will enable carriers to quadruple the number of customers on their network, without adding any new boxes.

"What used to take four linecards, now you can do on one," says Tlalka. "And since it's a single chip, the software is much easier, too."

Blackbird can also help system vendors improve their own return on investment by making it easy for them to take advantage of the arrival of 40-gig optical components.

Right now, for instance, a single Blackbird might be used on a linecard to process signals coming from four 10-gig transceivers and going to four 10-gig network processors. In the future, the four transceivers could be replaced with a single 40-gig transceiver without making changes to the rest of the linecard. This is because Blackbird boasts 40-gig architecture. Without this architecture, vendors would have to glom together four OC192 framers using an OC768 interleaver and an aggregation ASIC (application-specific integrated circuit). Such an arrangement would be much more expensive and power hungry, and wouldn't support concatenated OC768 interfaces, according to Ample. Concatenation -- binding together smaller channels to create one big virtual pipe -- is a key requirement for packet-processing gear such as routers.

Smoothing the migration path to 40-gig doesn't cut much ice with Rick Thompson, principal at PointEast Research LLC. "If you just think about it as 40-gig, it's somewhat questionable at the moment," he says. "But doing multiport 10-gig on a line card -- I do think that's valuable." Although Ample has only recently started raising its profile, it was founded nearly two years ago, in September 2000, and has been shipping its first products -- 10-gig frame processors called Nighthawk and Skyhawk -- since 2001. Future products (each with an avian appellation) will be capable of handling channel granularity down to the DS1 (1.5 Mbit/s) level.

Ample identifies two types of competitor. Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC) and Infineon Technologies AG (NYSE/Frankfurt: IFX) offer high speeds and low numbers of ports, while PMC-Sierra Inc. (Nasdaq: PMCS) and Agere Systems (NYSE: AGR) offer more ports at lower speeds. Ample claims to be unique in combining high speeds and high port density.

It may not be long before this claim is put to the test. AMCC, for one, is planning to announce some chip developments in this field in the coming week. We can't tell you more, but watch this space. In any case, Amit Bannerjee, AMCC's director of marketing, contends that system vendors are leery of buying chips from startups. "They always come back to us when the startup goes belly up." Ample, however, shows no sign of becoming an Ex-Ample of this at present.

— Peter Heywood, Founding Editor, and Pauline Rigby, Senior Editor, Light Reading
http://www.lightreading.com

(7)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
jamesbond
jamesbond
12/4/2012 | 9:52:34 PM
re: Ample Crams in the Ports
Wouldn't you need an 80Gig (40 Rx, 40 Tx) backplane/switch fabric to go with this?

Acclaim was a dud. Didn't amount to anything.
fabric_man
fabric_man
12/4/2012 | 9:52:33 PM
re: Ample Crams in the Ports
> What are you talking about? You can't put 4x10G
> into a 10G slot which is what I think almost
> everyone has today.

Well said, skeptic. The framer is a relatively small part of the overall 4x10G line card. Even if you're talking about pure SONET/SDH switching, the fabric to hook these up will not be trivially cheap. And the minute you get into systems that need to do protocol processing or protocol interworking, you need several NPUs, TMs, etc. as well as a packet/cell fabric.

We have customers who are designing 4x10G line cards, but they are purpose-built, carefully managed in terms of chip count and certainly aren't simple "upgrades" from 1x10G cards. The backplane connector alone ... don't get me started :-)

fabric_man
skeptic
skeptic
12/4/2012 | 9:52:33 PM
re: Ample Crams in the Ports
First, the startup is developing chips that will enable system vendors to offer a fourfold increase in capacity of their existing equipment, by cramming four 10-gig interfaces onto a linecard in place of one. That's music to the ears of carriers trying to improve their return on existing investments.
----------------------
What are you talking about? You can't put 4x10G
into a 10G slot which is what I think almost
everyone has today.

Its another tribute to light reading's disjointed
thinking that you can fall all over yourselves
praising a component company like this while
at the same time promoting the doom of the core
router market.

If the core router market is shrinking and there
is only room for two vendors and no startups,
who is going to be beating down the door for
high-speed framers?

redface
redface
12/4/2012 | 9:52:31 PM
re: Ample Crams in the Ports
It seems to me that what AMPLE is doing is very incremental progress and nothing revolutionary. I used to think revolutionary stuff should be what a startup is about. Now that there is such noise about a company with very little to contribute, I have to wonder that maybe making small moves should be the right approach to success for a startup instead of making big, risky moves. This way, progress happens faster, it does not require much investment, and it is less risky. This is how people like the AMPLE founder and Wu-Fu Chen could sell multiple companies to the likes of Cisco for big buck in record short time. I wonder whether this is still the best approach for a startup in the current business environment.
Belzebutt
Belzebutt
12/4/2012 | 9:52:31 PM
re: Ample Crams in the Ports
The only use I can see for this is creating interface cards that are 4x oversubscribed, and even then I'm sure the equipment will have to have been designed with oversubscription in mind.

Maybe Juniper can use this to make 40 Gig cards for the T640, since that's one of the few boxes that actually has 40-gig slots. But it's far from turning a 1-port 10-gig into a 4-port 10-gig.

optical_maverick
optical_maverick
12/4/2012 | 9:52:30 PM
re: Ample Crams in the Ports
Anyone notice the Nortel signs at Pac-Bell park. I guess you have to be sleeping to miss them... they are MASSIVE. Anyways, I'm a Giants fan and don't remember seeing those signs last year. Just curious to know who signed off on the millions of $$$$$ they paid for the signs.

Mavi.
MrForceful
MrForceful
12/4/2012 | 9:49:51 PM
re: Ample Crams in the Ports
Actually most of the equipment designed in the last year or so (that I know of, anyway) IS designed with 4x OC-48 in mind. This is also true of some of the stuff architected as long as 4 years ago.

While the issues with forwarding packets at 10 and 40 Gig do definitely exist, many solutions, especially aggregation solutions, anticipate statistical multiplexing and need very high port density to get to the desired equipment costs.

Mr. Forceful
Featured Video
Upcoming Live Events
November 5, 2019, London, England
November 7, 2019, London, UK
November 14, 2019, Maritim Hotel, Berlin
December 3-5, 2019, Vienna, Austria
December 3, 2019, New York, New York
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events