IP protocols/software

The Dwindling Case for Saving TDM

Much of the talk around replacing TDM switches has focused on the many benefits of IP transformation -- new service opportunities, fixed-mobile convergence and more efficient and lower-cost network operations.

But there's an increasingly urgent catalyst for TDM switch replacement projects that doesn't come with any feel-good factor: The rapid retirement of TDM-trained personnel and the shrinking availability of spares and software patches for these massive voice systems, many of which are finishing their third decade of service.

Already, carriers are turning to a refurbished equipment market for spares, says Jim Hodges, senior analyst with Heavy Reading . In a recent Light Reading webinar, IP Evolution for Mobile and Fixed Core Networks, Hodges noted that "it's unclear how long this will be a viable option."

The circuit-switched brain drain is a critical issue, though, as carrier staff with TDM expertise find themselves increasingly in the firing line. AT&T Inc. (NYSE: T) cut 5,900 jobs during the first three months of 2011, many in the sagging landline business, while Verizon Communications Inc. (NYSE: VZ) has shed 16,000 workers during the past 12 months, many of whom are older employees taking voluntary retirement packages. Other U.S. service providers are also trimming their workforces, notably CenturyLink Inc. (NYSE: CTL), which has absorbed Embarq and closed the acquisition of Qwest on April 1. (See Merger a Boon to CenturyLink Business? and Qwest, CenturyLink Plan $22.4B Marriage.)

Those opting for early retirement are taking with them much of the expertise used to run the Class 5 circuit switches, which have been the lifeblood of the telecom industry until recently. Newer employees are more likely to be trained in IP and Ethernet.

"The telco skill sets are changing," Hodges says.

To be sure, new features and the potential cost savings of newer technology (including reduced power and environmental costs) are still the primary drivers behind migration strategies. The older switches require six to eight times the power of next-generation systems, Hodges estimates. If power tariffs increase, as expected, to cover the cost of building the smart grid, "we will be reaching a critical turning point" when maintaining older switches becomes cost-prohibitive.

But there are also risks and complications to the TDM replacement process that may slow this process. Maintaining existing emergency services and ensuring network synchronization are reasons given to go slow on the switch away from TDM. (See Ethernet Europe: Busting the Backhaul Bottleneck .)

But telcos are increasingly aware that while maintaining the status quo is usually the safe path, keeping their services tied to ageing TDM switches carries its own set of risks.

"Most of the interest is in delivery of new services," says Mark Woollam, senior director, technical solutions, at Genband Inc. "But there is a growing awareness of the risks association with some of the ageing equipment."

— Carol Wilson, Chief Editor, Events, Light Reading

paolo.franzoi 12/5/2012 | 5:05:53 PM
re: The Dwindling Case for Saving TDM



The issue is (I think) beyond what you said.  New semiconductors are so expensive to build that nobody is going to speculatively build new TDM switch fabrics.

So the only technologies that are being built are the ones that either some company is real sure that it can get HUGE volume on (to do 45 nM devices) or a chip company sees an real market in (for example Ethernet).



fgoldstein 12/5/2012 | 5:05:53 PM
re: The Dwindling Case for Saving TDM

Thers's nothing wrong with TDM technology, and it still works better than packet for many applications.  The problen is old gear.  Embedded TDM switches are mostly from the 1980s and 1990s, using ancient semiconductors.  The 5ESS, with its analog crossbar stage, was obsolete before commercial volume shipments began in 1983.

New stuff exists, uses very lower power, and fits a tiny footprint.  But since the capital demanded "all IP" during the Internet boom, very little is being developed.  Telcobridges has some nice TDM+IP stuff that shows what can be done in the hardware.

cnwedit 12/5/2012 | 5:05:53 PM
re: The Dwindling Case for Saving TDM

Interesting perspective, although it's hard to imagine anything slowing down the IP train right now, especially for those carriers that want to converge their wireless and wireline networks.

Duh! 12/5/2012 | 5:05:52 PM
re: The Dwindling Case for Saving TDM

I haven't dug into it deeply but strongly suspect you probably would build a new TDM fabric in FPGAs rather than ASICs.  Probably not even particularly big or fast FPGAs, either.

Fred is right... TDM is much more optimal for voice than IP,   to the extent that anybody would optimize a network for POTs and ISDN these days.  We all remember the day when POTs was king and data was along for the ride.  Now it's reversed.  And Bellcore... er... Telcordia... is on the block again.

Besides, Carol raises the key point:  market perception is a far more important determinant than mere techological optimization.  Otherwise, the entire world would be ATM based by now.  That, and the fact that VoIP is regulated differently from POTS. 

yike 12/5/2012 | 5:05:50 PM
re: The Dwindling Case for Saving TDM

With everything over IP with the complicated typical encapsulation FCS + TDM/Payload + UDP(16bytes)/TCP + IP (20bytes) + EtherType (2 bytes) + VLAN (4 bytes) + MAC (12 bytes) + Preempt 12 bytes, too many over-headers, more bytes means more power consume/cost in the switching/routing/transport and less efficency, than the just memory copy based TDM switch, even further, when 20 bytes IPv4 was used up months ago, let's use the 44 bytes IPv6...Is all IP the right or at least wise direction? I am doubting ...

fgoldstein 12/5/2012 | 5:05:50 PM
re: The Dwindling Case for Saving TDM

TDM doesn't require much in the way of expensive chips.  Zarlink, for instance, makes a one-chip TDM matrix good for 32K simultaneous DS0 phone calls.  It's cheap because TDM switches are really just fast memory.  Write the incoming stream when it arrives and read it when it's time to send it.  A 125 microsecond TDM frame is a long epoch for a modern chip!

IP requires a ton more processing and power consumption.

Since most users still want analog phones for their wireline (they do work better than IPphones, especially at home), the line termnator circuit (BORSHT) remains intact either way.  Whether it's fed TDM (GR-303) or VoIP is a minor issue.


Duh! 12/5/2012 | 5:05:47 PM
re: The Dwindling Case for Saving TDM

All very true, but moot.

Is UDP/IP efficient transport for voice?  Of course not.  Is TDM highly optimized for voice?  Of course.   Could a modern Class 5 switch be built for significantly lower materials cost than a 5ESS or DMS100?  Of course.  Would it consume significantly less power? Of course.

But so what?  By every relevant measure, the POTS business is dying.  No telco executive in their right mind is going to sign off on new investment for a POTS switch upgrade.  And if they were inclined to, the analysts are so indoctrinated with VoIP as to foment a shareholder rebellion. 

By the way... those of us with a few gray hairs remember that back in the day, 10AM on Mothers Day was high busy hour for residential long distance.  The Bell System (and later the Baby Bells) used to warn consumers to expect blocked calls and beg them to spread their calls out during the day.  On Monday, there' be a press release touting the record number of calls handled on the previous day.  It's been a long time since I've seen any of that.

Sign In