x
Optical/IP

Open Source Eyes Telecom

The open-source movement might not be making Cisco Systems Inc. (Nasdaq: CSCO) nervous yet, but it's definitely intent on becoming a serious factor in the networking sphere.

The last few weeks have seen a couple of open-source ideas step beyond cult status into the "real" business world. Digium Inc. -- which supports the Asterisk open-source telephony platform -- picked up $13.8 million in its first round of funding last week. And recently, router startup Vyatta Inc. launched the first commercial version of its software along with offerings for services and subscriptions. (See Digium Raises $13.8M and Vyatta Launches Router.)

The start of a revolution? Maybe, but it's going to take a while. In the case of Vyatta, Forrester Research Inc. analyst Robert Whiteley says he doesn't see "any wide-scale movement toward it in the next five years."

Still, as networking hardware and software get commoditized, open-source products are starting to make sense. Whiteley points to the example of Cisco. One of its primary roles, years ago, was protocol translation. But with IP and PPP winning out, that's not a necessity.

Instead, Cisco, Juniper Networks Inc. (NYSE: JNPR), and others are concentrating on areas such as application-aware networks and security, not raw IP routing -- and that's why open-source routers won't usurp either company, Whiteley says.

He likens it to the server market, where Linux has made the operating system a commodity but hasn't killed off Microsoft Corp. (Nasdaq: MSFT). "Microsoft still thrives in that environment, because from the OS up they can still add value," Whiteley says.

The telecom world is opening up to non-proprietary wares anyway. The whole Advanced Telecom Computing Architecture (AdvancedTCA) set of standards, for example, is geared at creating a generic baseline for telecom equipment, a change from the old days of OEMs doing everything in-house. (See AdvancedTCA .)

Asterisk -- used as the basis of all manner of equipment, including PBXs and interactive voice response systems -- has been burrowing into the telecom consciousness since 1999. It's got 1 million users and serves 1,000 downloads per day, according to the Digium site. Digium, which employs 60, makes its money selling services and hardware related to Asterisk.

Digium doesn't actually need money -- it's been profitable since 2002 -- but president Mark Spencer, Asterisk's creator, says he wanted to wring the most out of Asterisk's success. That's going to take extra funding and some advice from people who've succeeded with startups before.

Digium's funding came from Matrix Partners , which had also invested in JBoss Inc. , an open-source middleware company recently acquired by Red Hat Inc. (NYSE: RHT). (See Red Hat Makes SOA Buy.)

Matrix gets a seat on Digium's board, but Spencer says he's not selling out.

"I've always said, if you give up control of your company to VCs -- either by the percentage you sell or the way the board is structured -- then you're selling your company," Spencer says. "I would not do something like that."

Asterisk supports old-world TDM networking as well as newfangled VOIP applications, and that's helped it gain fans among business users. "What we found is, there are enterprises that see the opportunity in this. A lot of enterprises are not happy with their existing vendors," Spencer says.

But for all its popularity, Asterisk hasn't had much effect on the business of giants like Avaya Inc. , says analyst Sam Wilson of JMP Securities . Nor does he expect Asterisk's influence to be a bother to them: "Incumbency entrenchment is going to win out," he says.

Spencer himself isn't yet talking about toppling anybody. He notes that open-source products might just push the giants to focus more on services, which is where the real money is anyway. "The jury's still out on what this means for Avaya, Nortel Networks Ltd. , and Cisco," he says.

To Page 2

1 of 2
Next Page
larytet 12/5/2012 | 3:44:45 AM
re: Open Source Eyes Telecom There are plenty of legacy equipment supporting by small cos or without support which could be open sourced by the IP owners. This way life span of the equipment can be extended and money can be made from the delivery of the hardware, support, positive PR on LightReading web site (good karma).

Open source community proves that large high quality projects is not an exclusive area of commercial cos with VC financing.

Linksys is one of the companies which build their products on Linux and opened the source and i think gained significantly in the process.
jasanz 12/5/2012 | 3:44:44 AM
re: Open Source Eyes Telecom I might be wrong, but my understanding is the Linkys does ODM to a bunch of Taiwanese and Chinese vendors and they are the ones who design the router with Linux or else inside. Don't Netgear, D-Link, Belkin etc... do the same?

Linux is moving inside every incumbent vendor... Not only Cisco Call Manager 5.0 run over linux, but other competitors IMS solutions are using this OS in their propietary hardware...
alchemy 12/5/2012 | 3:44:44 AM
re: Open Source Eyes Telecom Most commodity residential products have reference ports provided by the semiconductor company. The code is either authoried by the semiconductor company or is integrated from any of a large number of middleware shops around the globe. If open source exists, the semiconductor companies will use that rather than pay money to middleware companies. It's not like Netgear, D-Link, et al have a pile of developers kicking around. They have terrible margins and couldn't possibly afford the NRE on products that have 6 month lifecycles.

In the dark ages, many of these products typically ran on VxWorks or equivalent. Today, most are either on Linux or eCos (web search that one, it's not a bad freeware embedded OS). Who wants to give Wind River a buck per unit for a software license when you can run a freeware OS with a much more robust IP stack and utilities.
Dan'l Miller 12/5/2012 | 3:44:39 AM
re: Open Source Eyes Telecom Unlike writing OpenOffice or Linux software, the testing platform for embedded real-time software on datacom or telecom equipment is not readily available for the amateur open-source developer at Fry's or Best Buy (or for that matter, for the third-world open-source developer at the local bazaar) as PCs are. A few meters from my cubicle is a lab that contains easily $10 million of equipment that is used to test the software that I and other software engineers write.

For open-source control-plane software to be practical, either:
1) the multitude of software contributors would need to submit their work to a relatively few testers who do have access to the data-plane hardware; or
2) the data-plane would need to be available in a home version for the weekend warrior contributor working in his basement on evenings and weekends to do his/her own testing.

Perhaps open-sourcing might work for soft-switch call controllers, network/element management systems, billing, and other software typically hosted on Sun/PC/mainframe/Linux platforms. Such general-purpose-computer-based telecom/datacom software might permit SourceForge (or equivalent) to spend the required millions upon millions of dollars to set up a virtual lab over the Internet with WWW interface for people to schedule lab time.

For the aforementioned lab-testing reasons, I don't foresee any practical way to open-source the control-plane software on routers, media gateways, and other data-plane equipment until the open-source control-plane software has a testing platform that nearly anyone can afford (a la PCs). I consider this open-source pipedream right up their with open-source processors/ASICs. When AMD and Intel are given serious competition by OpenProcessor.org (because fabing the protype processor/ASIC has become as executing an executable on a general-purpose computer) then I will believe that Cisco, Lucent, Alcatel, Nortel, Tellabs, Tekelec, and so forth have a serious competitor, because hardware would have gotten so cheap that building a data-plane lab would be trivial/cheap/practical for a basement or third-world project.

In short, open-source software so far has tackled software running on general-purpose computers (e.g., operating systems, GUI environments, word processors, IT/MIS applications), not software that is embedded with a data-plane, where the control-plane software is a mere means to an end rather than an end in itself.

Telecom and datacom equipment manufacturing is an extremely capital-intensive R&D business. On the opposite Marxist philosophical spectrum, open-sourcing is essentially a big commune. How does a commune on a shoe-string budget acquire multi-million-dollar labs that can be shared as extremely scarce resources? The answer for Linux and OpenOffice is that the commune has each volunteer provide his own testing environment: a general-purpose computer.
larytet 12/5/2012 | 3:44:35 AM
re: Open Source Eyes Telecom "Who wants to give Wind River a buck per unit for a software license when you can run a freeware OS with a much more robust IP stack and utilities."


this is not just a buck/unit. this is also pure support, lack of applications (try to find SNMP server/client for vxWorks or WinCE), expensive non-portable development tools (Tornado costs 5-10K per seat), In case of small business commercial license involved with WindRiver products makes the development process and operating more complicated.

Bottom line - Open source usually allows more flexibility. The price is a requirement all or part of the sources. In most cases this is not that dangerous if the company business model is selling hardware or services.
larytet 12/5/2012 | 3:44:34 AM
re: Open Source Eyes Telecom "When AMD and Intel are given serious competition by OpenProcessor.org (because fabing the protype processor/ASIC has become as executing an executable on a general-purpose computer) "

Sun "open source" their CPU. I think this is an interesting example how a company can try to leverage openness of the system to bring more integrators in.

Indeed backbone, SDH equipment is hardly can be developed by a sole developer. It does not mean that such equipment can not be "open sourced".

One story, for example. Terayon after many years of struggle (Cisco and Juniper were among major competitors) closed the whole line of CMTS equipment. The alternative was to spin off the department, open source the whole thing and let others to continue the development of the application (Linux runs on the main board). I do not say that this was easy, but it was possible. Terayon had/has 1000s of CMTS deployed around the world. By closing the product line they essentially said the customers - we do not support you and we are not going to allow anybody else to improve/change the product.

Linksys routers i mentioned in the first post - you can find on the net at least two projects based on their source code. Does it create value for Linksys ? no doubt. Think about 10s of message boards where people discuss how to hack the box.

Re: Testing platform.
IBM has a testing platform for Linux. and RedHat too and Debian. Their night builds run on multiple CPUs (10+) in many different HW configurations. The same for Firefox. Do not underestimate complexity of software testing of large applications and costs of running modern server farm.

The PBX project mentioned in the article is more complex than just a soft switch.

I am not saying that Open source suits every project, but in many cases can be considered. For example chip maker can decide to open source drivers to make life of their customers easier. Something which Broadcom will probably never do. You have to fill forms, sign here and there, go 10 miles on the knees to get crippled, password protected PDF file from which you can not copy a register name. Isn't open source approach better. And Open Source is not equal GPL.
larytet 12/5/2012 | 3:44:24 AM
re: Open Source Eyes Telecom I never used anything with vxWorks but GNU toolchain. Unfortunately they did not upgrade their GCC for years. Currently i compile application for MIPS with version 2 of GCC (not 4 and not 3, but 2).

vxWorks itself is not that bad. TCP/IP stack is bad. They indeed promise that in vxWorks XX IP is much better, but frankly who cares ? Who is ready to pay money again just to get a better TCP/IP ?

I wonder what is the business model of WRS these days ? Do they truly believe in their ability to get share of the Linux market ? But i guess my opinion is biased - i work too much with their products. WRS marketing and sales have ability to pursue managers to buy their products. May be that "1G flash disk for your key chain" plays a major role hehehe. MontaVista should start to send their kits on the flash disks (aka Live CD)

I would not say that vxWorks is a "complete" crap. You can reach a fairly small footprint (under 300K), relatively good performance (assuming that you do not use IP stack and timers and a couple of other things), it is relatively easy to debug if the application is not too large (it is not a router, for example) and you have JTAG/BDM debugger (costs money too).

On the other hand i would not say that Linux lacks any problems.

I read about eCOS. And I read the source code - this is the best C++ code i ever read. But without RedHat or other large co behind the project i am afraid to use the code. Another problem is applications for the OS. As you say it depends on the project. I run a couple projects - one in a chipmaker (PON related SoC) where OS is dictated by the customers (mostly Linux). Another is a health care device with LCD, USB, serial ports. For this one co chose Windows CE. I should say that Platform Builder (MS toolchain for Win CE) is the worst product on the market. But the life is sh*t, isn;t it ?
alchemy 12/5/2012 | 3:44:24 AM
re: Open Source Eyes Telecom larytet writes:
this is not just a buck/unit. this is also pure support, lack of applications (try to find SNMP server/client for vxWorks or WinCE), expensive non-portable development tools (Tornado costs 5-10K per seat), In case of small business commercial license involved with WindRiver products makes the development process and operating more complicated.

Bottom line - Open source usually allows more flexibility. The price is a requirement all or part of the sources. In most cases this is not that dangerous if the company business model is selling hardware or services.


You can run the Gnu tools chain with VxWorks and avoid paying those pirates those huge per-seat fees. My gripe is that their operating system is complete crap. Historically, their IP stack and environment has been next to useless. They did go out and buy Interpeak so there must now be some better IP stack products from Wind River.

If I were doing an embedded project today, I'd likely take a hard look at eCos and the GNU tools chain. It's free and it's much easier to build low-level things like device drivers in that sort of environment than in Linux. You won't get all the cool things that have been developed for Linux but there are an awful lot of projects that don't need them.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE