Traditional network optimization always involves drive testing. Since the first generation of mobile/cellular (AMPS, TACS, NMT) networks were built drive testing has been a core component in optimizing performance. However is it possible to optimize a 2G, 3G and 4G network without drive testing? Most engineers have doubt that it can be done since the most difficult part of the network to operate optimally is the air interface or the RF part of the network. Further to that most engineers first learn optimization through drive testing. Is it really possible to optimize a mobile or cellular network without drive testing?
We all know that drive testing is costly, time consuming and difficult to implement. Having that said there are several ways to optimize a network without using drive testing. This involves the use of network recordings, statistical counters, customer mobile measurements and probes. The simple idea is to optimize the KPIs in a way that the performance gain are much better than using drive testing alone. The question is can you really optimize the RF part (antenna configuration and parameter settings) without drive testing. The answer is YES, it is possible as long as the KPIs are correlated across multiple data sources such as network recordings and statistical counters at the very least. If you can bring in probes and customer mobile measurements the better the results will be.
There are of course a very few tools or software systems and solutions in the market that can do this today, most especially correlating multiple data sources for optimization. One such tool is Actix One which relies mostly on the drive test input and another example is the Wireless Explorer that uses all available data sources and is open source. The challenge of course is to find out a solution that satisfy the custom requirements of every operator and the differences in processes and procedures.
The best optimization results are of course obtained by incorporating and correlating drive test data to the network recordings, performance statistics, probe measurements and actual customer experiences. Customer experiences are captured through detailed network or probe recordings and more accurately directly from the customer mobiles. So if you can utilize each and every subscriber in your network as a drive tester collecting massive amount of data, wouldn’t that be better than sending out several teams of drive testers? At the end of the day drive testing is still needed for trouble shooting and competitive benchmarking. It will not completely go away but it can be complemented and at the same time reduce its overall cost.
7 comments:
Optimization w/o drive testing is quite possible; vendors like Actix, Arieso and Schema offer solutions which use network data instead of drive testing collected data.
It comes down to being able to determine accurate call location. Companies listed above all come from the planning and optimization space--ergo their location is sub par. I represented Hexagon (now part of Nice), and Israeli company who has an extremely accurate location platform based on data collected from probes or network data collection systems (such as Ericsson's GPEH). They hav existing customers, and some (Israel, Singapore) have almost completely stopped drive testing for the purposes of optimization.
This company claims they can do RF optimisation without drive test and without probes - just CDRs: http://celllution.com/
I am very skeptical. I note their partner Gybe Systems seems to have disappeared recently
Thanks Mike. I can't find information about the company and their website don't say much either...
Drive test is actual customer experience, so it has big value. There are some points given below to support it:
1. Customer Experience: Drive testing actually simulates subscriber experience and drive tester is actually in the environment. Drive tester is not just getting stats from OSS which sometimes do not completely represent the issues unless other data sources are used (still complete problem may not be identified with other sources). When doing drive tests, drive testing engineer can identify the real issues on the ground and can check the site antennas or Node B as well if there is a related problem found.
2. GPS location: We do not get accurate GPS location with OSS based tools as mentioned in the comments and its location data for all the calls may be or may not be available depending upon the satellite line of sight. If the network does not have location hardware (LMU), then GPS location may be about 200 to 500m inaccurate and with location hardware, it is about 50m at best. Without accurate GPS location, it is hard to know the location of problem.
3. Store and Process Engine: A sufficiently big, powerful and dedicated store and process engine is required that can store and process calls with the ability to target desired area. As per today, only less than 20 call traces are generally available in the RNCs as this activity creates big processor load. Apart from this fact, all the call tarces are usually occupied by VVIPs or the company top executive chain. The remaining test traces are usually very small to do a country wide drive test for the optimisation department. The additional RNC load is also avoided because possible restarts of RNC will be the last thing desired.
4. An alternative to the OSS Statistics: Drive tests can support the OSS statistics results or negate these as well. If there are call drops in an area with good statistics in OSS, then we shall prioritise drive test team to look after the problem and also find why OSS does not support the drive test results. Same status may be with probes although these are designed to be performing very well but face limitations.
5. Customer Experience at Desired Location:
If we want to know about the mobile network performance on a rural road which does not have any significant traffic, we will not be able to gather sufficient samples to analyze statistically well. A drive test helps in this regard and we can drive the area even when there is no customer on the road.
6. Reliability and inter-operability of about handsets:
A regular user may use handset which may be functioning well or not. It may be working and still drop the calls due to hardware or software issues. By using drive tests, we shall be sure that the test equipment is reliable and if it is doing measurements, these are accurate as the equipment is calibrated. It will also let us know if the problem is related to a particular handset although it can be achieved using call trace as well. A recent work has shown problems with MMS being unsuccessful only with certain brands on a mobile network which was not identified previously by the operator through monitoring of statistics.
I hope that the points mentioned above clarify and show that drive testing is necessary and cannot be eliminated completely although may be supported by OSS and probe based measurement tools.
Subscriber mobile measurements are like drive test measurements and can indeed be used to optimize mobile networks. In fact, subscriber measurements and other network data are the very essence of "Self" in the Self-Optimizing Networks (SON) concept - the ability to continuously and automatically optimize networks without the need to stage drive testing.
Certainly drive testing data has its place in the initial deployment of the network when no subscribers are present, but as the network loads, subscriber measurements can do their part and facilitate SON."
ReVerb Networks is taking this a step further into intelligent antenna-based SON using KPI's based on networks measurements to iteratively optimize coverage, capacity and interference in the network through management of tilt, azimuth and beamwidth of antennas. In this self optimizing process pioneered by ReVerb, an algorithm determines the optimal antenna settings and delivers them to networked Remote Electrical Tilt (RET) controllers and azimuth and beamwidth controllers of 2D and 3D antennas to achieve an automated antenna-based SON.
More comments at:
http://www.linkedin.com/groupAnswers?trk=view_disc&gid=136744&commentID=37755882&viewQuestionAndAnswers=&discussionID=47077231
Nice reviews by the auther. Well written and really informative information for future too.software testing
Post a Comment