Mobile operators around the globe are investing huge amounts of cash to boost the capacity of their networks as the number of subscribers accessing high data volume Internet applications continues to grow.
Already there are an estimated 186mn global subscribers to mobile data services and some experts say mobile data traffic is doubling every six months. Informa estimates a 25 fold rise in data traffic by 2012. About 10 per cent of mobile users can account for 80 per cent of the data traffic, according to some operators. Many mobile data users also use the wireless network just as they would their fixed broadband line as their main access point to the Internet. Informa also recently discovered that the costs of delivering data consumption outweigh an operator’s ability to improve network infrastructure based on the premise that two per cent of mobile data subscribers consume 50 per cent of network capacity.
But the capacity squeeze facing operators today is caused by more than the sheer volume of data traffic: the type of mobile data application and even the device from which the application is accessed can all impact the management and capacity of the network.
This double burden of increased traffic and more complex applications to manage presents a very real challenge for operators. If they don’t adapt quickly they risk impacting the high quality of service that mobile customers have come to expect from their network operator. Unfortunately many mobile broadband subscribers already are experiencing quality of service issues including sluggish connections and service outages for example. There is potential for the problem to extend to mobile voice subscribers too since the network will not distinguish how and where it squeezes capacity. In addition, increasingly overburdened network control and signalling channels are common to both voice and data services .
In a highly competitive market, no operator wants the service experience to be the reason behind any customer defecting to a rival operator. Fixing the mobile data capacity squeeze is the key agenda item of the day.
Operators however have mixed views on how to solve the problem. Some are upgrading to 3.5G HSPA or even 4G technology or adding hardware capacity at the radio access level of the network. Unfortunately such upgrades will not eradicate the problem entirely – after all, as any highways authority will tell you, the more lanes you add to the roads, the more vehicles will come along and clog them up. Further, the available spectrum to accommodate additional radio access channels is anything but limitless.
Some operators are looking at effectively offloading traffic onto fixed line ADSL etc. networks using picocells or femtocells. Although these will boost capacity by essentially adding another layer of network to the existing infrastructure, they also have management and set up challenges and costs. Other operators are trying to gain more capacity from the existing spectrum through reducing the cell size of individual radio sites or optimising the management of the network. It is this last option that offers potentially the most sensible route for operators.
Optimising the network to deal with the variety as well as volume of data being transmitted across it – from file sharing and mobile email to mobile video, gaming and podcasting - will provide an immediate solution to the capacity jam problem. This will be a relief to operators that would rather avoid more cash expenditure in their networks. But it is also something few operators used to providing simple voice and more basic data services are equipped to deal with for themselves.
Pinpointing trouble spots
First, it’s crucial to understand exactly where the network is inefficient. The more mobile data services an operator provides, the more complex the daily management and monitoring of the new, multi-service network becomes. It would be a mistake to make assumptions about where the problems might occur. The only way to really understand what’s going on is to audit the performance of the network and perhaps also benchmark the results against the collective performance of other operators’ networks through carrying out the various test sequences in parallel across the ‘home’ network as well as on each of the competitors’ networks at the same time. Sophisticated and automated mobile test beds, the front ends of which are simply an array of mobile phones, one per network under test, make the measurements either whilst being driven around a pre-defined route or stationary at known ‘hotspots’.
New measurement parameters
Since the end users’ actual quality of experience (QoE) becomes more important than quality of service - connection times, dropped call rates, latency, and speech quality for example – operators need to add many more KPI’s (Key Performance Indicators) to their measurement programmes.
For instance measuring bandwidth provisioning as well as usage becomes essential. Of course bandwidth usage can vary wildly depending on the type of services and files that are being accessed. A video file will be more intensive than a web page for example. The upstream/downstream bandwidth requirements also need to be taken into consideration since mobile data applications tend to have variable capacity requirements in either direction.
The key is to simulate as many possible real-life usage scenarios and protocols as possible – which will help overcome the unpredictability of always-on mobile data applications. For example, a typical measurement sequence might run through FTP file download and upload, HTTP web browsing, end-to-end PING turnaround etc. – mimicking a typical user’s normal activities of sending and receiving emails, uploading and downloading files, browsing web pages such as Facebook, viewing video on YouTube and so on.
The focus of the work should be on end-to-end application optimisation on the one hand and on understanding the interworking between the radio access bearer (UMTS, HSDPA) and the core IP network and applications on the other.
In one operator’s audit we performed low-load tests to prove the radio bearer and IP throughput could reach system limits and then did the same test under poor radio conditions, under high-load conditions and whilst driving. The test results revealed a whole range of other problems such as resource management, scheduling and mobility management.
Acting on measurement results
Post-processing of test session data that is acquired from such extensive auditing can yield a vast array of results ranging from simple end-to-end throughput values to highly sensitive engineering parameter traces that indicate the status of the most critical as well as the some of the more esoteric, configuration parameters in the network. This can help to identify weak spots that can be re-engineered as well as help predict emerging service issues which will help operators with future capacity planning. For example, by correlating real-time analysis on user and group-level behaviour with actual network and node behaviour, the operator might predict and alarm on emerging service issues such as network congestion, service breakdown and roaming/handoff problems.
It’s also worth noting here that it’s vital in any such exercise to provide a level of reporting that rises above the technical detail and speaks to the management audience. So while the ability to drill down into the bits and bytes with comprehensive ‘tracing’ of individual transactions and sessions is crucial for the engineers, equally the management seeks easily assimilated top-level performance analysis presented in a reader-friendly format.
Operators have many challenges to overcome as they roll out their mobile data offerings. Most importantly, they will only be successful if the quality of the end user experience is not compromised. Creating new measurement paradigms will help operators to optimise performance and get the most efficiency and capacity from the available infrastructure before it becomes necessary to invest further.
Tony Gray, Middle East Regional Business Director of P3 communications GmbH