Prashant R Jain

vCloud Director is an application that was developed in Adobe Flex. It lets users provision a virtual machine (VM), define and edit the VM parameters, and delete it when not needed.

I was one of the UI engineers on this project. The memory of the Chrome tab running the UI would go up to 2 GB and eventually crash. The memory leaks that vCloud Director was experiencing were a problem for many other UI applications developed within VMWare using Adobe Flex. I was assigned the task of fixing the performance issues in vCloud Director.

I started debugging the memory leaks when a view was removed from the DOM. Going from View1 to View2, I learned that all the objects in View1 would leak. Studied the framework code and built a generic view garbage collector that would de-reference all framework-created references, including parent or child references, event listeners, and data bindings. A hack in the flash VM also helped kick the garbage collector every 20 seconds.

The end result was the generic module that helped GC 80-90% of objects. This module was adopted by every business unit at VMWare and enabled them to ship their products.

Lacework is a late-stage cybersecurity startup. I was hired as a UI engineer to work on the management and monitoring UI applications.

There were many performance problems in the applications. Engineers refused to own these issues, and these performance issues also impacted sales. I was assigned to get to the root of these issues and find the performance bottlenecks.

I started with getting two APM/RUM tools in our environment, Dynatrace and Honeycomb. The data from these tools exposed performance issues in the entire stack, including infrastructure, UI, UX, back-end services, and database queries. This data was shared with the whole engineering organization.

On the infrastructure side, we enabled HTTP2, CDN, and browser caching of static resource files using webpack content hashing. The database and API teams worked on improving the performance in their layer.

As a result, customers who refused to renew licenses because of performance issues have renewed their yearly contracts, the UI runs much faster than it used to, and test cases were created for old APIs. Every new API now has to be backed by a performance test case.

Salesforce runs its own data centers. I was part of the infrastructure analytics team that monitors the servers' performance, including application, database, file storage CPU, and disk space in the data center.

I was assigned to build a weekly heat map that visually represents the predicted performance of the servers for the next two months. The data science team provided the performance data and future predictions.

I used AGGrid to build a grid-based heat map. Some of the features of this grid heat map include:
• Weeks represented in columns
• Sortable, moveable, and resizable columns. The column states were stored in the local storage and applied on every user visit.
• Columns pinned to the left and right
• Providing a column selector to show and hide the columns
• Weekly columns with gradient red, green, and yellow colors based on severity.

As a result, users were happy with the overall UX because it made their life easy. The data was well represented using gradient colors, making it easy to identify the week when a server would start peaking or run out of disk space. Also, a lot of code written was modularized for reuse and was reused in other parts of the UI application.