Proven scalability of data collection and meter value handling

In January and February 2013, our customer CEZ in the Czech Republic conducted tests looking at scalability of data collection and meter value handling.

For this purpose, a test environment was established at the University of Brno, based on a combination of virtual and physical servers. A third party developed a simulator using a DLMS based protocol that simulated a heterogeneous concentrator-based solution. The values were pushed by the virtual concentrators continuously based on close to realistic circumstances.

Both communication errors, uncertain and missing values were simulated based on statistics from CEZ’ AMM project. In total, 99 values per meter were gathered per 24-hour period (one 15-minute profile, three 24-hour period values), which is high compared with existing infrastructures in Europe.

The task was to collect, validate, estimate and store in the shortest possible time.

To begin with, we chose to carry out a small “warm-up test” on 100,000 meters, in order to check how the simulator, the test environment and Powel ELIN worked together. This test was run with only one communication server, as opposed to the ten communication servers planned for the “big” test. Without further optimisation, we conducted the test with an average processing time of just under seven hours. In other words, one communication server managed to gather, validate, estimate and store around ten million values, in just under seven hours.

Based on this test, we adapted some of the routines for the configuration of the test, and in mid-February, we kicked off the test with 3.5 million meters. Powel ELIN was run on ten communication servers (virtual servers with six CPUs) versus one physical database server with 32 CPUs and 100 GB RAM. The test lasted for three days, in which time, Powel ELIN collected, validated, estimated and stored upwards of one billion values. During periods of high pressure, we were able to process around 16,000 values per second. Average time of processing per 24-hour period was roughly eight hours, which is very fast considering the exceptionally large amount of data.

One of the keys to this success is that Powel ELIN validates and estimates all the time series before the data is saved to the database. Because of this, the total process is achieved much more rapidly. We also observed that it was possible to substantially reduce the size of the database.

Our customer, CEZ, also conducted the same test with another supplier. This supplier is a world leader within Smart Metering, and is known for having an extremely scalable system. Overall, Powel ELIN completed the test slightly faster than this other supplier did. This is an achievement we are very proud of. It shows that both Powel ELIN and as Powel’s metering solutions are ready for great responsibilities – both in Norway and in the rest of the world.