Interview with Scott Grout, CEO, RadiSys

COTS hardware has been around for more than a decade. What is keeping equipment makers from deploying these solutions?

A suitable COTS architecture has only been available for two years. Other COTS architectures like Compact PCI were basically enterprise technology. So COTS suitable for core telecom applications that can support five nines did not exist until ATCA availability. And ATCA availability goes back a couple of years.

I do see ATCA growing to be a multi-billion dollar business, but, as with all things in telecom, it will take time. The biggest complexity has not been about ATCA itself. It is about porting applications over to ATCA and, in particular, making a move to Linux. So moving to ATCA hardware is quite easy to do, but a lot of legacy applications simultaneously moving to ATCA are also moving from proprietary, 20-year–old, home-grown OSes to Linux.

One of the biggest motivations for using COTS hardware is the time-to-market factor. How far have vendors been able to achieve that in practice?

We have customers, who, using ATCA, have brought applications to market in under 12 months versus classic 24-to-36 month development period. We have a customer in Asia that chose our ATCA platform in Spring 2007 and, before the end of 2007, had live service up and running for a wireless application.

Can you perhaps name some of the application categorises that are being enabled by COTS hardware such as ATCA? How would you generalise the type of applications being ported over to ATCA platform?

ATCA will be the hardware of choice for a wide spectrum of applications. Over the last few years we have seen particular strength in the wireless arena: media gateway, RNC, SBC, and IMS applications like media server. IPTV has also been a very good adopter of using ATCA. From the fixed network side, we are seeing elements like access consolidator and echo cancellation as prime candidates.

What about elements like feature servers. Are they suitable for ATCA or would IBM Blade Center be more suitable there?

It depends on the TEM perspective of their COTS choice. IBM Blade Center has a number of benefits. It comes with high compute density, NEBS qualified etc. It certainly has volumes behind it. Challenges include IO that is significantly behind ATCA. Data plane processing is behind ATCA. Density is slightly behind ATCA. IBM Blade Center locks you into one vendor. On the other hand, ATCA components are provided by many vendors so there is a broad choice of suppliers. So it depends, among other factors, on how a company values open supply chain as opposed to being locked in with one vendor.

There are obvious motivations to port over to ATCA. But when you speak to vendors they find themselves having to source bits and pieces from various suppliers so in the end integration is still an issue and it does not do any good if time to market is critical to you. What is your perspective on the integration issue?

Even if a customer chooses to integrate on their own, it does represent faster time–to- market for the individual assets because we began 10 Gig switch development before any other potential customers starting developing their own. Coming back to your question, you absolutely get faster time-to-market if you look toward vendors like RadiSys for a turnkey platform. In the example I gave you earlier about the vendor achieving all within 2007, they used a RadiSys turnkey model and they focused on just the application.

What proportion of your deals would involve platforms as opposed to components?

We are seeing 50% of deals at platform level and 50% at individual asset level.

It seems plausible that if carriers are having budget constraints they would lean towards COTS options like ATCA for reduced OPEX. Do you expect the same if we see a recession in the U.S.?

Yes. In telecoms there is significant pressure on price and costs of networks, both CAPEX as well as OPEX. Cost efficiency in terms of OPEX and CAPEX is the most significant driver in the move towards ATCA.

A few questions about applications. I have been following your media server product line for the last couple of years. The type of applications that these IP media server products used to enable were things like IVR, centrex, conference, messaging etc. Has anything changed in that area over the last 12 months in terms of the applications that are being developed on top of your media servers?

We are seeing innovative applications being developed right now around mobility and video. Our Convedia media server has also been very successfully used for IMS and VoIP deployments. We are involved in lots of IMS trials. Another interesting area is the mashups where companies like EQO allow you to integrate social networking with mobile.

In the mashup space, developers that use conferencing as underlying platform prefer to have proprietary setup. Is that not a market that you could potentially target? There are literally hundreds of such companies mushrooming.

Yes, certainly we see this as an emerging market. It is still early days, but there is long-term potential. There is some pressure to look at a proprietary solution, but if we work with the developers and help them understand architecture flexibility and the openness of the platform, we will succeed in having them use commercial media servers.