Salesforce’s application performance management team faced a challenge in late 2020: they needed to improve inconsistency detection algorithms.
The performance management team was monitoring the health of Salesforce’s data centers, which emit a lot of metrics in real time, including CPU usage for any particular service. It generates metrics known as time-series data.
“If you can detect discrepancies in the telemetry metrics that you are getting from data centers, you can quickly detect what is happening in the salesforce, then you can quickly correct it, and that reduces downtime for customers,” said Adyot Bhatnagar, senior. The research salesforce engineer told VentureBeat. “So that was the original motivation for why we might be interested in the overall time range.”
For the past two years, Bhatnagar and his team have been developing an open-source machine learning library called Merlion, which conducts time-series analysis with machine learning. It was originally designed to help solve the challenge faced by Salesforce’s application performance management team. Merlion is an end-to-end Python library for many time-series tasks, he explained, including discrepancy detection as well as prediction.
How Merlion works and what time-series enables machine learning
The Merlion Project was launched as a collaboration between Salesforce research teams in Palo Alto and Singapore.
“The merlin is a mythical animal, half a lion, half a fish, which is also the national animal of Singapore,” Bhatnagar said.
As the project is named after the mythical Merlion, Merlion machine learning technology is more than just one thing. Merlion includes capabilities for loading and processing data, creating and training a wide range of models integrated under a common API, Bhatnagar said. The project also includes practices and measures for model output, as well as a framework for actually evaluating model performance.
Once the Merlion project was started, the Bhatnagar team quickly realized Salesforce’s wide range of internal requirements for time series machine learning. The original motivation for the project was the discovery of discrepancies for application performance management.
“Apart from that, we also got a lot of use for time range forecasting for a very wide range of tasks,” he noted.
For example, in the IT operation area, if there is a service that uses computational resources, such as memory and CPU, time-series-based machine learning can be used for predictive prediction. It can predict how resource utilization will change, which could help the salesforce in better planning capacity.
From idea to product consumption for Merlion
One thing to keep in mind for a machine learning library is; It’s another thing to have technology that actually works in the production environment.
Bhatnagar said that in his view, a common challenge for any machine learning library is integration into the production environment. This includes enabling machine learning tools to retrieve data as needed, with access to the necessary computer resources and the ability to read data back in the right places.
To meet that challenge, Bhatnagar said the Merlion project has added some default options, providing users with a good starting point. The project continues to simplify the overall workflow to make operations more automated.
Towards a new open-source standard for time-series analysis
Merlion is not the first open-source project to try to help solve the challenge of time series analysis.
The most popular is the Prophet project led by Facebook, which provides predictive capabilities for time range data. According to Bhatnagar, the prophet did not meet the needs of the salesforce as he had only a subset of facilities including pre-processing, modeling, evaluation and post-processing. That’s why Salesforce decided to create its own project and then open source it.
As an open-source project, Merlion can be used on Salesforce and by anyone looking for a time-series data machine learning analysis framework.
“In our view, there was a lack of a standard solution that caters to all the needs of the people for time-series analysis in one place,” Bhatnagar said. “So we thought this would be incredibly useful, not only for the salesforce, but also for others with time-range problems.”
Venturebeat’s mission Transformative Enterprise is about to become a digital town square for technology decision makers to gain knowledge about technology and transactions. Learn more about membership.