The Great Softswitch Revival
Softswitch makers must deal with every-thing from new form factors
to incorpora-ting the features of other network slements.
by Richard Grigonis
Back in 2002, the softswitch business was on the brink of disaster,
along with the whole telecom industry and the rest of the economy in
general. The following year, however, worldwide softswitch sales
grew by 42.4%. Today, VoIP and wireless are the two hottest spots in
telecom, and service providers, ISPs, CLECs, IXCs and ILECs realize
that, while VoIP continues to gain a foothold, the PSTN will also be
around for the next 15 years or so, and the principal network
component that can seamlessly interconnect IP broadband, PSTN, and
wireless networks is the softswitch.
Softswitches are cheaper to install and maintain than conventional
TDM equipment, and yet they enable the development of a vast range
of revenue-generating nex-gen services. This is why IP services
wholesaler Global Crossing, for example, continues to decommission
old TDM-based Class 5 switches in its network.
Ironically, the all-important softswitch is actually a conceptual
or "logical" network element made up of various physical
(real)
elements, the most important of which is the MGC (Media Gateway
Controller) that mediates call control and controls access from the
IP environment to-and-from the PSTN. Other softswitch elements
include the SG (Signaling Gateway) that interfaces to the SS7
network, the MG (Media Gateway) the packetizes/depacketizes voice
traffic as it passes between IP and PSTN networks, and the IP/ISL
(IP/Internet Services Layer) where media/application servers and
other devices provide the `value added' or `enhanced'
services to
the network.
Some or all of these physical elements can be housed in a single
CompactPCI or AdvancedTCA rackmount computer, or they can be
separate boxes running in different locations (in a softswitch, call
processing is totally separate from the physical switching performed
by the MG). Softswitches require some kind of gateway functionality,
though standalone gateways exist in the network having softswitch-
like call control functionality. Moreover, application servers can
be placed anywhere in the network, or inside a single-box softswitch.
Thinking Inside the Box
Todd Wynia, VP of marketing for telephony board maker Artesyn
Technologies (www.artesyn.com), says: "We view the softswitch
market
as double-digit annual growth over the next five years, but not 200%
a year or anything outrageous. That's based more on RFQs [Request
For a Quote] and activity at the telcom OEM level."
"Softswitches are therefore not a huge market for us, compared to
other network elements, but it is something we see as a growth
market," says Wynia. "The types of products we provide for
softswitches are boards/blades that are PICMG 2.16-based–also
called
the "Ethernet backplane" or "Etherplane"
technology–and they run on
Linux and they're for CPU-intensive applications. The RFQs and the
particular design wins we have are based on PowerPC processors."
"Our network customer's switching application runs on each one of
the 15 blades in a rackmount chassis," says Wynia. "They use
a
protocol stack and an application on top of Linux that enables them
to create a softswitch out of those PICMG 2.16 boards. There's not a
lot of I/O required other than the Ethernet connection. It's
basically a cluster on the backplane held together by the Ethernet
backplane."
"The great equalizer here is Linux," says Wynia. "Nothing
necessarily ties this application to the PowerPC; the PowerPC was
chosen for this particular app because of the PowerPC's power-to-
performance ratio, as compared to Intel's. The PowerPC does handle
Ethernet traffic more efficiently than the equivalent Intel
processor. The PowerPC is also good for the I/O architecture of the
softswitch and the PowerPC's power/ performance ratio allows us to
squeeze a lot of processing power onto a single shelf that you
wouldn't be able to achieve with an equivalent Intel processor."
Wynia prognosticates: "As for the future, we've seen a number of
RFQs for AdvanceTCA-based softswitch architectures. Some of the RFQs
are based on PowerPCs, some on the Intel architecture, and some,
believe it or not, are split. For example, they'll do the data plane
portion of the softswitch as a PowerPC while the control and
application plane portion of the softswitch resides on an Intel
architecture blade. In the past, mixing those two architectures on
the same chassis would have been a big issue. Fortunately, these
newer technologies such as Linux and the 2.16 `Etherplanes' have
alleviated some of the issues associated with mixed architectures.
With them, we can take advantage of the best features of Intel CPUs
(they run legacy software) and Power PCs (lots of them can be
crammed into a chassis, because they have low power consumption)."
Thinking Outside The Box
Emergent Network Solutions (www.emergent-netsolutions.com) provides
advanced telecom software for emerging communications networks.
Their ENTICE (Emergent Networks Telecommunication Infrastructure
Control Environment) solutions portfolio includes a softswitch,
session controller, VoIP gateway, converged tandem/international
gateway and various enhanced service solutions. Their new E-REV
(ENTICE Residential/Enterprise VoIP) solution is an end-toend VoIP
infrastructure package for CLECs, ISPs and wholesale carriers
wishing to offer residential VoIP or hosted IP Centrex services.,
Emergent's CEO, Nathan Franzmeier, has an interesting conception of
a softswitch: "When I talk about a softswitch, I'm referring to
something that does call control for gateways," he says.
"Maybe it's
doing endpoint services too. But what immediately became apparent a
few years ago was that everybody who was building a softswitch also
needed SBCs [Session Border Controllers] or at least SBC
functionality. Our session control product has all of a softswitch's
capabilities plus the unique ability to handle RTP [Real-Time
Transfer Protocol]. Everybody is really buying into that combo
solution. We actually don't call it a softswitch, but a session
controller. It handles both call control and RTP streams so we can
do things on both the wholesale and retail sides; we can
do `topology hiding' like a NAT [Network Address Translator] or a
proxy server and we can do firewall traversal too."
"Also, once you deploy a softswitch into any market, you soon
discover that you need to provide media services," says
Franzmeier. "If you don't have control over the media stream,
then
how can you redirect it here-and-there, and play prompts, music, do
CALEA or perform whatever voice-type service the customer wants? You
absolutely must have control of the media streams. So, we don't just
do call control. In fact, I don't think we've sold a `pure'
softswitch in over a year."
"To be a session controller company is to pigeon-hole
oneself," says
Franzmeier. "You then have to grow up and become a softswitch
company and handle the whole network, or your functionality has got
to be rolled into routers and other devices. Still, it was apparent
in the emerging market that SBC functionality was needed, whether it
was in the softswitch or worked into the protocols, or wherever.
What we really offer is a network operating software system that
controls all network gear, which makes us a sort of union of both
softswitch and SBC functionality."
Trends, Challenges and Prognostications
Vendors' vigorous efforts to coerce the softswitch into dealing with
all aspects of today's hybrid "converging" network have made
what
was an already imprecise, logical network element into an even more
flexible (or nebulous) `what-the-customerwants-
in-this-particular-
instance,' type of technological critter. Still, one can make out
some trends, challenges and possible predictions.
One trend is that softswitches are integrating with SBC
capabilities, such as handling RTP streams. However, would a vendor
of such softswitches do an immediate revision in its hardware
platform simply because a new SBC feature became available? Probably
not. Even so, the distinction between SBCs and softswitches is
blurring.
The influence of 3G wireless and its IMS (IP Multimedia Subsystem),
the would-be universal service architecture for wireless and
wireline networks, is increasingly felt. Services must reach all
endpoints, so they must be capable of moving across multiple
networks. Even today, some vendor architectures are able to deliver
services seamlessly (well, almost) across network boundaries.
The 3G standards (3GPP and 3GPP2) also bring to the table more
transport layer security and similar security-minded methodologies.
You'll see more of that creeping into the requirements for network
deployment. Right now, security is almost a secondary concern. As
more spectacular attacks occur, people will wake up and implement
more transport layer security, as well as using anti-hacking
algorithms on SIP servers so crackers can't grab people's phone
accounts and make unlimited calls.
Softswitches will always need to take into account the "retail
side"
of carriers selling Class 5 emulation service directly to
subscribers. Hosted solutions continue to grow too; everyone wants
to deliver all kinds of services out to the endpoints. One the other
side of the network, application servers could in theory be
replaced. Subscriber-oriented services would still exist as some
kind of service component.
Softswitches need to have clear paths to interface to all equipment.
As Emergent's Franzmeier says: "Wholesale networks tend to be
isolated from the requirements for doing traffic redirects;
operators try to implement that stuff locally with servers and what-
not that handle the retail traffic. You need to easily deliver
services off to those retail subscribers or enterprise subscribers.
For example, if a carrier wants to offer a `ring-don't-answer'
solution for an IAD [Intelligent Access Device] working with a PBX,
then where is that delivered in your network? Is it delivered at the
softswitch side as a network level feature? Or do you expect the
application level server to handle it? Class 5 features can now be
deployed at smart endpoints. So, some features can be delivered at
the edge, while other features should be delivered in the middle.
Whereas it makes sense to do three-way conferences out at the
endpoints, larger conferences may require conference
servers–where
do you put those servers? On the softswitch side? If you do, then
how do you integrate them with the endpoints, etc., to deliver a
nice seamless service to everyone?"
More interoperability testing will be needed (e.g., when a provider
sets up a Brand X softswitch with a Brand Y applications server to
deliver endpoint services).
Increasing transactional complexity will make the scaling up of
softswitches difficult. First pointed out in a 2003 presentation by
Nortel Product Manager Steve McKinnon, he illustrated the problem in
terms of Presence: "30 friends and/or colleagues = 30
subscriptions.
If everybody has 30 friends, then every local event causes 30
notifications. Each subscription results in notification at some
average frequency resulting in a system state change every 10
minutes or so. If multiple devices [probably about 3 or so] need to
be updated, then that becomes a multiplier. 30 friends thus results
in 3 Devices * 30 Notifications * 6 Notifications per hour = 540
network events per hour, and that's just for one person/ endpoint.
Moreover, we've only tabulated the Presence state information that
we can see, not any background processes."
Perhaps the most amusing prediction comes from Emergent's
Franzmeier: "You'll see core level softswitches, application
layer
softswitches–application servers that look like a little
softswitch
but they will concentrate on a particular piece of the network such
as IP endpoints or enterprises, etc. For example, you've got
BroadSoft and Vocaldata, which generally emphasize application areas
instead of wholesale solutions. You could thus envision a hierarchy
of softswitches which, ironically, brings to mind the old Class 4,
Class 5 circuit-switched hierarchy! Interestingly, the technology
appears to be moving in that direction."
In theory, a long time from now powerful SIP-based endpoints will
communicate with each other in peer-to-peer fashion, thus making
softswitches obsolete, right? Well, not for quite a while. For a
discussion on this, see this month's Editorial by Yours Truly
elsewhere in this issue.
vonmag.com |
