X
Tech

Fixed-mobile convergence switch race begets a Ferrari-Hummer hybrid

The telecos might be willing to spend $6 billion over the next few years on this stuff, but they want ROI ... big time. TCO, too. So all the telco parts (parts is parts?) suppliers -- he called them TEMs -- are thinking, you know, like, blades.
Written by Dana Gardner, Contributor

So I got the weirdest voice mail of my life today. Now, we've all heard about boards of directors leaking to the press. But have you ever heard of a telecom switch dialing up an industry analyst all by itself and -- I swear, it sounded just like Stephen Hawking -- leak pending news about a multi-core chip architecture tag-team that allows for faster-better-cheaper fixed-mobile convergence processing?
And what's, like, totally, even weirder ... I received the exact same voicemail at the exact same time on both my mobile phone and my fixed land line. Neither ringed, it went right to voicemail. If they can do this, why can't I get them to combine my phone bills?
Okay, so this computer-generated voice says it's a network switch calling, said it's name was Netra, and it said it was about to be transformed, or fork-lifted in, or something really drastic. It said that new chips plus new hybrid platforms (ATCA) plus IP Multimedia Subsystems (IMS)-based  applications would, like, change the telecommunications world -- which is going gaga over a new generation of networks, partially so users can do this fixed-mobile convergence thing.

Anyway ... the processing needs are huge. There are humungous packet streams to be super redirected and massive billing issues about how to charge for minutes over your land-line from your mobile -- or, like, how many packets can dance on the head of a nanosecond? If it's all IP, I said to myself, what in Nokia's name do minutes have to do with it? How about you give me "24" for $0.50 on my damn cell phone? I said to no one.

Anyway ... then the freaky voice got really excited. It went into, like, a history of telecommunications or something. Used to be that big, huge, switches sat in central locations to process direct-circuit calls and billings. Fine. Now, high-end FMC applications (to create innovative and highly optimized FMC services such as triple-play voice, video and data and many additional components addressing the latest requirements for 3G and IMS deployments) are needed, but the hardware can't be big and honking anymore.

The telecos might be willing to spend $6 billion over the next few years on this stuff, but they want ROI ... big time. TCO, too. So all the telco parts (parts is parts?) suppliers -- he called them TEMs -- are thinking, you know, like, blades. I want my rack! the voice yelled at me. I'm thinking this one got way too hot, but I can't hang up ... it's just too creepy. Then I get this mantra about "64-bit, multi-threaded architecture," and, like, parallelism and UltraSPARC T1, and Carrier Grade Linux (CGL)? Like, wha?

I'm about ready to hurl the RJ-16 right outta the wall, when the little bugger comes up with a really compelling analogy. "It's like they put a Hummer and Ferrari on the same substrate, man, you get the heavy lifting and the super-speed -- one place, nice and cool. And they're running Sun Solaris in some threads and Wind River's CGL Linux in some other threads, and the same blade is processing the billing stuff and the data throughput stuff altogether." ROI up the whaazoo!

Okay, so I'm thinking, like, Intel and Freescale are just gonna, like, roll up and walk away from the telco space, right? Moto? Un-uh. But the father-board says this combo of a UltraSPARC T1 processor (its complete design is published by OpenSPARC.net and available to anyone under the GNU General Public License, did you know?) uses eight CPU cores running a hypervisor that provides 32 physical threads into a single ATCA compute blade (it supports the emerging PICMG 3.x ATCA specification), which is, like, 28 more than any other general-purpose multi-core processor in the market today.

I'm thinking, this sucker is, like, two years ahead of the other guys. Their bad. The SPARCy is getting 5-9s of Linux powering those packets, and total Solaris applications doing the back-office stuff, all across these threads; totally managed; contained. Whoa, dude.

I'm thinking, why buy this stuff? Lease it from GE, for crissakes, like it's a freakin locomotive. Let them insure it. Pay by the month, baby.

Then the line went dead. My Max Headroom moment was over. Then it crackled again, and said: "Support for the next generation OpenSPARC processor will be available in the spring 2007 release of the Wind River Platform for Network Equipment, Linux Edition." Whoa.
Disclosure: Wind River is a sponsor my BriefingsDirect podcasts; Sun Microsystems is, like, not.

Editorial standards