search

phylotastic / phylo_webservices

This repository contains all Web Services that NMSU team created
1 stars 0 forks source link

Identify and fix scaling bottleneck #62

Closed arlin closed 6 years ago

arlin commented 6 years ago

From load testing results here:

https://docs.google.com/document/d/1mgal-qCKOlGGrHC0EsEve2KzxP_LCy_q2Y2_ytOeM6c/edit#heading=h.fzugdk5hu7q2

we believe that there is a bottleneck in scaling services at the level of the NMSU wrapper or the underlying R function calls.

Find out what is wrong, and fix things so that we can support 100 concurrent scaling threads.

abusalehmdtayeen commented 6 years ago

The bottleneck of the scaling service is the current Python-to-R interface that have been used to call datelife methods. It was not able to accept multiple connections at a time and was processing requests sequentially. The solution is to use a different R interface to access the datelife library.

abusalehmdtayeen commented 6 years ago

I have replaced the old Python-to-R interface with a new one. The new interface will be able to handle multiple connections to datelife R library. As a result, the datelife service will now scale better.