SevOne & Turbonomic—IBMARM I & II
Role: Lead Product Designer
Tools: Sketch, Invision
Project Overview
Turbonomic originally only focused on VM and cloud resource rightsizing, automating workload optimization across virtualized environments.
SevOne's focus was network performance monitoring, collecting and visualizing performance data, surfacing application behavior (response time, latency, throughput) through the network lens.
Neither side alone had the complete picture when it comes to the relationship between applications and the resources they consumed. The first collaboration after their merger was to combine Turbonomic's resource data with SevOne's performance telemetry and create a new product line called ARM (Application Resource Management).
The Hypotheses
1.
Certain specific resources have a measurable impact on application performance
2.
Adjusting the consumption of these resources can significantly improve application performance
3.
Optimizing those adjustments reduces overall resource consumption cost
The ARM product was based on a set of untested hypotheses, which we needed to prove before building out the full product. This meant the project was broken into 2 phases.
Getting Phase I right was the prerequisite for customer trust, and for the 6 months of testing that would define Phase II.
Phase One
The foundation of the of the project was based on the theory that adjusting the consumption of certain resources can provide noticeable improvements to application performance. Thus, it is vital to provide evidence, not only internally but also to the customers, to showcase this relationship. This helps to provide trust through tangible evidence, and drive user adoption.
To achieve this, we used historical customer data gathered by SevOne through their management system. My role was to work closely with the SevOne product team to devise a visualization of the data to compare between resource consumption and application performance.
Left: Above: Some examples of the data visualization we initially tested. It helped us figure out what worked and what didn't.
User Experience
The design was mostly focused on the data display and its interactions, as that tells the most important story to the customers. We quickly landed on a time series line chart as the best data representation. Due to the large amount of datasets, we also designed the ability for users to select which of the datasets to display. Once the development team implemented the charting UI, it was quickly evident which resources had the largest correlation with application performance.
Resource Actions
Simply establishing correlation does not prove causation. In order to test causation, we mapped points of interest onto the time series line chart where resource consumption adjustment actions were taken. While the complete datasets can go back several years, we decided to focus on data that spanned the most recent 24 months. By comparing datasets from large pools of user data, we were able to see which resource actions had the most impact on application performance.
Above: Some examples of the eleven (11) different design iterations we went through.
Phase One Result
After eleven (11) iterations, the end result was ARM I—a monitoring system that gave the users a top down, holistic view of their environment. The user is able to drill down from the top level view of all of their applications into each individual application and view all of the resources it has consumed overtime. By toggling the markers on the time series line chart, the user is also able to view a history of all of the resource actions.
Below: Final ARM I designs.
Phase Two
About six months after the launch of ARM I by SevOne, we started phase two of the project. The objective of phase two was to translate everything we learned so far into specific parameters to automatically generate resource adjustment actions. The time in between the two phases allowed SevOne to gather more real time data, which helped Turbonomic’s product team to further refine the automation framework.
My role was mainly to design the widgets that summarized these new actions. As Application is a new topology type, I also needed to design the drill down dashboard views. This included the Global view of the environment from the application perspective, individual application summary dashboards, and further breakdown view of each transaction/interaction within a single application. This was packaged and launched as ARM II, the final piece in the ARM product line.
Below: Final ARM II designs.
Final Thoughts
ARM helped Turbonomic enter a new direction and expanded its customer base. Its success, from conception to implementation, was thanks to the incredible collaborative efforts by the product teams from both Turbonomic and SevOne. ARM was a testament in the importance of data analysis. Without ARM I, ARM II would not have been possible, as it was the foundation which the whole product was built upon. It also helped, through real world evidence, to build customer trust and drove faster customer adoption.