Rule & Data Processing Improvements in Gainsight

  • 20 December 2022
  • 8 replies
  • 257 views

Userlevel 6
Badge +1
  • Lets put your data to work!
  • 823 replies

In our continuous endeavor to improve our rules and data processing platform, over the last two quarters we have done many changes which have significantly improved the average wait times across our customer base for rule runs. This post shares a brief on the changes we have done and the impact it has had on execution times.

We have moved rules into our micro service architecture and have also scaled our infrastructure to meet the growing number of data processing executions. Additionally, we have also worked on improving context switching and micro batching of jobs resulting in reduced wait times for the rule to be picked up for processing. Because of these changes, we were able to increase the parallel rule executions for our customers.

We pushed these changes in a phased manner starting early September 2022 and completed in October 2022. We have seen 51.87% improvement in average rule execution times from its peak in Aug 2022. Rules execution time is calculated as the time difference between when a rule was added to the queue to when it completes execution.

Here is a visual representation of the average execution times for rules since Feb 22:

seh40QZTJ_GqHRDRiMfHrHr6TXow09dRmi0SV9aftXIXIh6XB6W92NAt4O6uiJZaUK_1s485H-VmmC7sNMe07iPEUznd8wTiEKdnSv7m41mVg0XOzRImrdKWWGRAjIxEU41AWNOJ06xu94eAGlzFaPq4725He4mupHoY9DtALdHu-tzVEcCska5KL39NLw

Majority of these improvements were in the wait times of the data processing stack which is responsible to run the data preparation such as extraction, transformation, and merges.

Here is a visual representation of the average execution times for the data processing portion of rules since Feb 22:

tkaJXdld_SJNgqQyCdklGrEAwJHVq9XW_X7PSsJEUfnuLFfefD0NiKMTbcd8Ro8j1W45LrBq09C3spHOM3b1YF-VtNwNuMC4qTvXwt9a-Eza_RJNiR7bY0as2CmP_KnAPty8rnRxdKiapELEtTuTWSPj0t4aW2NhinL3WgEc9QJUuCsPnISyg-zB1GvIdw

Similar to Rules, other product areas in Gainsight like Data Designer, JO, connectors, and Adoption Explorer all use the distributed data processing stack. Improvements in the data processing execution times also contributed to better execution times for all these product areas as well.


8 replies

Userlevel 6
Badge +1

A lot of times, what is visible gets a lot of attention and credit, but products like Rules are like icebergs - whats visible in UI is only a small portion of what it really is. While we have released Horizon Rules Beta (PFA the overview video below) the above post makes me feel equally if nor more proud of our team. These changes are live in production and some of you might have already noticed it. 

 

Userlevel 5
Badge +8

Thank you @rakesh!  We have seen great benefits by running rules in parallel.  This Rules Horizon Beta looks interesting.  How can we be added to the Beta?  What is the time commitment and process for providing feedback?

Userlevel 7
Badge +10

Definitely noticed the change.  Thanks, @rakesh and team for all the hard work!

Userlevel 6
Badge +1

@angela_domenichelli , So glad to see the interest in the beta program! You will need to be nominated by your CSM to participate in the Rules Engine Beta Program!

 

Userlevel 5
Badge +8

@tessa_horn let’s discuss with Denise this afternoon if there’s room on the agenda.

Userlevel 7
Badge +6

@rakesh and team - awesome work on this. Absolutely be proud of the behind the scenes work, glad you showcased it here. 

Badge +3

Incredible work @rakesh and team! 

Userlevel 2
Badge +1

Much needed, Appreciate this initiative! 

Reply