What's Behind Cisco's SIP Buy

Earlier this month, Cisco Systems went company-shopping again, this time snatching up Dynamicsoft for $55 million in cash. Officially, the acquisition gives Cisco a credible VoIP offering to sell to wireless and wireline vendors. Unofficially, it gives Cisco key expertise needed to further its enterprise VoIP solution.

Dynamicsoft is a 104-person company that makes carrier-grade VoIP servers aimed at enabling the creation of advanced voice and data services over wireless and wireline infrastructure. Cisco has long offered a carrier-grade Session Initiation Protocol (SIP) proxy server, but has lacked the feature server critical for delivering advanced services to those markets. The acquisition of Dynamicsoft is expected to provide Cisco with that portfolio.

But not without some pain. Cisco will likely have to further invest in fixing the internal code structure of Dynamicsoft's products. "A lot of times, given the choice to make long-term enhancements to the product that would make it more flexible, stable, and maintainable, we would instead be instructed to add a specific feature in the hope that it could be used to land a specific account," says an engineer close to Dynamicsoft.

More fundamental flaws in the code were ignored because the "marketing guys" couldn't see the advantage of addressing functions that weren't externally visible. The net effect was that it took a lot longer to add features to Dynamicsoft products.

How much longer is hard to quantify, according to the engineer. "My gut feeling is that had we spent a good six months doing those kinds of improvements, we could have made that time up within a year or so," he says. So instead of spending a couple of months to add a feature, he estimates that Dynamicsoft could have added the same feature in a matter weeks to a month.BED FELLOWS

Ironically, architectural limitations may make Dynamicsoft a perfect fit for Cisco. Cisco's enterprise IP PBX, Cisco Call Manager (CCM), has its own history of architectural limitations that have made adding new features a significant challenge.

Cisco continues to push its own Skinny Client Control Protocol (SCCP) within CCM. "CCM is architected in a softswitch model so that it has individual and direct control over every session," says one engineer close to Cisco. "This means it takes years to add a feature." He points to Music on Hold as an example. "Cisco was only able to add that feature with two years' worth of work."

Even today, SCCP offers no presence construct. Like other features within SCCP, getting that capability is possible, but not without some work. Cullen Jennings, a distinguished engineer within Cisco, notes that while there's "no classic presence construct," most of the information needed for presence can be derived from the SCCP environment because the call controller is the master of the environment. "We think SIP/SIMPLE [SIP for Instant Messaging and Presence Leveraging Extensions] is the right way to do that," he says.

JEKYLL AND HYDEIt's this kind of sentiment that gives Cisco a most Hydian relationship with SIP. On the one hand, SIP is almost nonexistent within CCM. SIP devices aren't supported, locking users into an all-Cisco environment. Only SIP trunking for connecting up to a SIP server is provided.

On the other hand, Cisco continues to lead the protocol's development. Jennings wrote many of the SIP drafts. Rohan Mahy, a network architect in Cisco's Voice Technology Group, co-chairs the SIP Working Group.

The acquisition of Dynamicsoft should only solidify that leadership, and if anything perhaps raise concerns over Cisco's dominance in the area. Dynamicsoft's CTO, Jonathan Rosenberg, co-authored SIP. Dean Willis, a fellow with Dynamicsoft's Office of the CTO, co-chairs the SIP Working Group with Mahy.

At first glance, it doesn't seem as though the lack of SIP support has hurt Cisco much. During the second quarter, Cisco's market share increased by 6 percent over 2003, providing 29 percent of the 1.5 million lines shipped to enterprises, says Matthias Machowinski, directing analyst at Infonetics Market Research. The next closest vendor was Nortel Networks, which shipped just 17 percent.

THE FUTURE OF CCMYet as the market demands new features-features such as presence that are readily available from Nortel-Cisco will be forced to respond. Adding those capabilities along with other new features may stress the CCM architecture.

It's hard to imagine Cisco giving up on CCM and its proprietary protocol so easily. The competitive advantage is just too great. Besides, Cisco tried that one already. Dreadnought, a project within Cisco to develop an enterprise SIP proxy, was killed off, supposedly because of internal politics. According to the engineer close to Cisco, a team of six Cisco engineers was able to deliver 80 percent of CCM's features within three months, with the goal of delivering 90 percent of the CCM's features within six months-and that was with CCM having a six-year head start and several hundred engineers working on the project, he says.

On the other hand, the Dynamicsoft acquisition offers key technology that together with Cisco's SIP proxy could address many of CCM's limitations. Cisco has for a while now spoken about the need to integrate presence with the rest of its product line. Dynamicsoft gives Cisco key expertise in that area, notes Jennings.

As for whether the combination of Dynamicsoft's technology and Cisco's existing proxy server will serve as a basis for a future CCM release, Jennings only had one thing to say: "Call me back in a month on that one."