Mobile data traffic growth has put the spotlight on the need for cellcos to adopt efficient and flexible all-IP architectures. But cellcos and IP experts at a online discussion in January say that IP networks should evolve only if the economics are justified. The event, moderated by Telecom Asia global technology editor John C Tanner and sponsored by Cisco, was held via telepresence at sites in Hong Kong, Singapore, Sydney and Mumbai.
John Tanner: What types of devices are customers using and what impact is that having on your IP resources and your network?
Christian Daigneault: Dongles are generating about 80% of all our traffic at this point. The other 20% is mostly smartphones, and traffic for both has grown by about the same factor - 18 to 20 times - since we launched HSPA+ nine months ago. And this is not stopping. When we talk about the iPhone, and those type of phones, we see new devices like this coming out on a monthly basis, so this is not going to stop.
Stephen Chau: In our network, more than 80% of our traffic is generates by the dongles. Also, we are offering a residential fixed-broadband-type service [using HSPA as the last-mile link]. So from that perspective, if you include that with the dongles, that actually represents more than 90% of the total data traffic volumes.
Anthony Goonan: I can relate completely to that 90% figure. And within that, about half of our data is just straight http-type traffic. Around 10% of it is YouTube and other streaming-type services. High-end smartphones, for Telstra at least, are contributing seven to ten times as much data as a standard-feature phone. And our data traffic in Telstra is roughly doubling every eight months.
Lam Hong Kit: One thing to add on the point of http traffic, from what we see, that's not only just for http web, but also progressive downloads. I don't know the exact split, but a big portion of http comes from the video via the progressive download, and YouTube is one part of that.
Tim Mark: Our colleagues in the US, and even some of them outside the US, are seeing traffic doubling every six months because of video and telepresence - all the traffic has been doubling. We're really seeing that from the infrastructure side.
Jayesh Easwaramony: Basically what is happening is you're creating a much bigger addressable market for service providers. So if you look at mobile broadband, instead of having one household broadband connection, you are actually selling four connections - each laptop will have a connection. Also, we talk about smartphones and netbooks, but there's also going to be another entire set of devices - which could be tablet PCs or e-book readers - which are going to further the gap between netbooks and smartphones.
Hong Kit, how much is this driving internet traffic on the international links?
HK: Before 12 months ago, a lot of mobile operators were under their mother company, which also owned a fixed-broadband service, for example, and they were bound together because the traffic was relatively small. After 2008-2009, they began to have high enough growth demand that they were able to spin off, and they could come directly to us to buy IP transit. And in the past 12 months, some of the customers have doubled their bandwidth. Traffic growth for our mobile customers has been 100% to 120%.
What kind of impact is all this having on the mobile IP architecture?
CD: Well, we have HSPA+ and all-IP to the cell sites, and if you have fiber or high-capacity microwave, there's no bottleneck with the radio access or backhaul. For the core, at this stage it has been a bit easier to grow the capacity. There will come a point where you want to simplify your core, but I don't think the complexity is at this level today.
SC: I agree. In Hong Kong, there's not as much concern about network loading because we've already been building up the network to accommodate the traffic to meet the broadband need, not just the mobile usage.
AG: One thing that we found in Telstra, because we've got a common core for our 3G network and HSPA+ network, as we went through the speed evolution, we would find parts of our IP network that were in fact causing bottlenecks, and they weren't exposed until the air interface speeds went up. So there are bottlenecks within your core that you just don't know are there, and we need to look at how we can identify those and remove those in a cost-effective manner.
Given all that, just how urgent is it for mobile operators to migrate to a flat-IP architecture with LTE?
CD: I think it's a question of volume and throughput. And right now at 21-Mbps, I don't see it as a bottleneck. There is the claim that with flat IP or LTE, we'll get better 10-20 millisecond latency vs the over-100-millisecond latency we have now. That would make a big difference, but I think a lot of it is due also to the LTE radio access and NodeB.
AG: No one's going to pay the engineers to make it flat for the sake of being flat. We're not loading things onto our existing network, we're unloading things off. So we're doing incremental things at the moment and it'll take a technology change to really move to true flat architecture.
SC: It has to be something that's justified. When LTE is ready, should we then jump on it? My question is still: why? By all means, if there's a good reason to get a much more efficient network architecture, you can always do that. It's not a question of technology; it's a question of business evaluation.
JE: I think this flat-IP decision is more like when you buy a car, you want to first look at its looks, and then look at the fuel efficiency. So I think it's that kind of a decision - it's not necessary as of now. I guess it is a desirable state, but is it an essential state? Maybe not at this point of time. In terms of really understanding cost efficiencies, I don't think operators have completely grasped the cost economics of their network. I mean, they can do it technically, but if you present a business case to the CFO, I think that modeling goes a bit haywire, so I think that's something which operators need to work on.
AG: I think you make a really good point there. The engineering group want to have the latest and greatest devices in customers' hands, the device distribution people want to give the cheapest ones out there, and the pricing folk want to drop the prices to be cheaper than what you can move a bit around the network. So you've got these three parties with diverse models and you have to make sure those three different needs get aligned, and it's a real challenge.
Christian, does your CFO get the economics?
CD: Yeah. [laughter] We have a CFO and a CEO who is also a previous CFO, so he's very focused on the dollars and the business case. Our challenge on the technology side is that we need to reduce the costs, because that's the only thing in our control. We increased our traffic by 18 times in the last year, maintaining the same backhaul. So we have reduced our cost by 18 times for the backhaul. It also means very strong negotiation with our vendors. Vendors are coming with all those claims of reducing costs by doing it this way. Very often you don't see the cost savings. So you really need to understand the economics of your network costs to be able to drive the costs down.
SC: This is why, as a CTO now, I always call myself the most non-technical guy in the company [laughter]. The reason is that I want to be even more upfront on the overall cost structure rather than just focus on the technology, because there's a lot of factors that affect the whole cost structure, not just the stuff we're buying. It's the whole value chain and how we work together with the international bandwidth provider, the domestic leased-line provider - they have their own costs they need to adjust as well. But my job is to make sure when we build anything, we understand the business rationale.