Data Distribution Analysis Metrics Using JavaScript

One of my Alcatel-Lucent projects was to author methods and procedures to collect and analyze data collected from the execution of  LTE cellular data throughput downloads under various testing scenarios.  Ajaxtest was the JavaScript application I authored to perform download throughput tests and analyze results.  The reporting module contained a JavaScript application class that analyzed the raw data set producing a Histogram and normalized standard deviation/error plots.

<CLICK HERE> to load a demo that analyzes a user-provided data set, displays results and a histogram.  There are numerous additional options used to validate and format the data which is not included in this simple example.  Hint:  Inspect the JavaScript for code that actually calculates results…

In our tests, we often collected 100 samples that would provide a “normalized” distribution.  The more samples collected, the better chance Ajaxtest would return a normalized distribution.  Often, you must review result data to prune invalid and/or outliers as LTE TCP downloads are sometimes affected by uncontrolled factors due to the variable nature of the internet.  Some features that are not contained in the Ajaxtest demo above include methods to automatically detect and prune outliers using standard deviation and other calculations.  Ajaxtest reporting also analyze 2 simultaneous test results comparing results between the TCP optimized and non-TCP-optimized flows executing at the same time.

These tests were executed by Alcatel-Lucent / Nokia system test and field installation teams to validate performance improvements when routing LTE data traffic over the “TCP split-tunnel” network equipment that our company developed and deployed into Verizon’s LTE network.  This equipment  increased data throughput (decreased load time) between 10-50% over non-optimized traffic depending on the size of the data sample as validated by Ajaxtest.

Comments are closed.