By Aloke Guha, CTO, Cofounder, OpsCruise

I am honored to be speaking at this year’s IEEE Infrastructure Conference on October 8th, virtually of course. Having had a long association with IEEE beginning as a student member and then becoming more involved as an attendee and presenter at many of its conferences and workshops, speaking at this conference is both nostalgic and yet novel. Why?

The conferences in past including ones I have participated in or spoken at were in multiple areas of information systems --  computing, storage, networking, and big data/ ML -- focused on new and interesting work in each of those specific areas

The IEEE Infrastructure conference is both topical and interesting in how we have come full circle where infrastructure topics such compute, storage and networking are no longer isolated disciplines.

This is not “your father’s infrastructure.”

To paraphrase an antique tagline, today’s infrastructure is not what it was a decade ago. Cloud computing and the new application architectures have fundamentally changed how infrastructure resources will be managed. We’ve moved from monolithic to microservice-based applications that are complex distributed systems, and modern infrastructure that are not just simple virtual machines with networks and storage, but complete virtual environments.

Organizations that moved to cloud to leverage agility and scale have initiated the fundamental shift from monolithic to microservices applications. While they improved modular development, testing, and iterative improvements, microservices also created significant challenges for Operations (Ops) teams managing application performance.

These challenges result from application structure and behavior variability, shifting bottlenecks, limited visibility and control of cloud resources. Traditional performance management approaches such as using queueing models are not feasible for microservice applications, given the lack of in-depth knowledge of the application structure and inability to handle the complexity at scale. Heuristics that capture static relationships between performance and resources are also no longer applicable with highly dynamic virtual environments.

In my talk on Thursday, October 8th, 2020 at 1:15 PM PT titled "Model-Based Control for Applications," I will address what we have learned and built at OpsCruise. Managing microservice applications in a cloud environment requires an approach different from those that worked in the past. Ops teams are often not in a position where they can instrument application code, and clearly not in the case of SaaS (software as a service) or FaaS (function as a service), nor have the visibility into the obfuscated layers of virtualization created by container orchestration and virtual cloud infrastructure.  Instead, relying on existing open source monitoring data, i.e., not in the data or I/O plane, we discover and build the application structure.

Subsequently, using both ML and a priori knowledge of known services, we can automatically build predictive application behavior models at scale.

When deployed in runtime, the models enable early detection of performance problems and interpretive causal insights. Those insights can be used to isolate the cause, and recommend remedial actions. Besides avoiding the cost of additional monitoring frameworks and non-invasive instrumentation, this model-based approach results in lower MTTD problems and MTTR with a path to autonomous operations.

The talk will be also recorded and posted to OpsCruise YouTube channel.

Please join me and register for the IEEE Conference here

Learn more about IEEE Conference and other speakers, the world’s largest technical professional organization dedicated at https://infrastructure.ieee.org