AskLytics: Enticing Cloud Optimizations for Business Amplification

Tiago Poola, Founder & CEO IT organizations have been investing a fair amount of their resources in equipping themselves with robust application performance and systems management solutions in order to optimize end-to-end performance and avoid unforeseen system failures and outages. Few organizations today have DevOps teams specializing in making their applications perform better by optimizing their infrastructure and platforms based on system behavior under changing load conditions. Moreover, problems are persistent for the DevOps especially during continuous delivery (CD) of applications and deployment cycles where it gets difficult to detect the changes in the infrastructure needs when moving to newer versions of the applications. When factors such as marketing campaigns, breaking news, and seasonal business activities cause unexpected workload changes on applications, it further intensifies the problems for DevOps. This is where AskLytics is a game changer with its unconventional approach of carrying out optimization from an infrastructure & platform standpoint to enhance application performance while providing insights into system behavior to the DevOps teams using machine learning algorithms specific for system management. These algorithms are designed to help DevOps identify and resolve system issues before they adversely affect the company’s revenue and reputation. Tiago Poola, CEO of AskLytics says, “We make it easier for organizations to optimize the infrastructure and platform, as optimized infrastructure results in better application performance and potentially lowers infrastructure costs.”

According to Poola, the applications in the market vary in nature, mostly because of the data being ingested, the workloads, the non-homogeneous clients, and multiple other aspects that vary substantially. From an infrastructure standpoint, applications can be under workloads with different characteristics during different time periods that have different system resource utilization patterns. The performance goals for some applications could be low latencies while for others it could be higher throughput. In order to respond to the effect of such dynamic changes on the health of the applications and to maintain service level objectives (SLOs), the Operations Team conventionally overprovision the infrastructure. On the other hand, AskLytics’ goal is to empower Operations through use of case-specific and infrastructure-specific optimized solutions while creating a performant environment that complies with the changing nature of applications. In terms of optimization, a multitude of providers in the market often prefer to optimize from an ‘end-user perspective,’ using deterministic approaches, without ‘optimizing from an infrastructure perspective’ as well. This necessarily requires unparalleled expertise and results in increased expenditure. Large organizations usually employ a dedicated team to carry out infrastructure optimizations. AskLytics targets to provide infrastructure optimization to the mid and small organizations eliminating their budgetary constraints and helping their applications to thrive within optimized infrastructures. AskLytics built its machine learning platform with their team of engineers and data scientists that have unique expertise and knowledge in IT systems domain.

AskLytics started off with providing APIs for machine learning algorithms in systems management.


We make it easier for organizations to optimize the infrastructure and the platform, as optimized infrastructure results in better app performance and potentially lowers infrastructure costs


It closely observed the plight of DevOps who could not afford the time and effort to build their own solutions using APIs. AskLytics then came up with solutions to help them streamline their work more easily. “We figured that the best way to help DevOps is to provide intelligence about application’s infrastructure needs at earlier stages of the software delivery workflow than the production stage.” AskLytics then decided to focus on the Performance & Load Testing (PnL) stage, and on pre-production, as opposed to production itself. The company focuses on a three-dimensional view of factors which model application behavior and performance that mutually impact each other. The first dimension is the workload on the server. The second is the application behavior in terms of the resources it consumes to do the work, and the third is the quality of service. After the functional testing stage, when the application moves to the PnL stage, the developers are aligned towards code optimization. The Ops, on the contrary, tend to focus on the end-user perspective of the application, working on the system performance, quality of service of the application, such as latency and throughput, infrastructure sizing and platform configuration.

AskLytics brings QA and performance engineers to collaborate with both the developers and Ops creating a bridge between the two. One way the company achieves this is by enabling them to understand the characteristics of their application before production. With the help of the ingested data and machine learning algorithms, AskLytics classifies the workloads and characterizes the resource usage and consumption, along with determining how the quality of service is being impacted for different workloads. Through this, they create a relationship between all the three-dimensional aspects of application at runtime. Poola explains, “The value that this three-dimensional analysis offers our customers is that it enables them to compare the infrastructure performance of different application builds. This also helps them to choose the right production infrastructure by running tests safely in their PnL and pre-production environments.”

Illustrating one of AskLytics’ successful client engagements, Poola shares the story of a large electronics conglomerate who wanted to gain visibility and analyze the impact of different applications on the infrastructure. In the past, they used to follow a more conventional process where they had a testing phase explicitly allocated for sizing and figuring out how the application behaves with a significant load. AskLytics stepped in and used its application characterization approach and provided them with a formal relationship based on workloads, resource consumption, and the latencies that are associated with that application. This enabled them to visualize and compare different versions of their applications. This gave them the most needed critical insight into the application behavior even before they began with regular sizing and testing activity.


Since AskLytics’ strength is its statistical model-based algorithms, it focuses on solutions for DevOps that use a probabilistic approach rather than the deterministic one. The company enables the collaboration between performance engineering and Ops by providing a platform with infrastructure-centric application performance analytics. It assists in the creation of workloads that are similar to those seen in a production environment, giving credibility to PnL tests & analytics. The company uses its core statistical models and machine learning algorithms for empowering technical teams to streamline their software delivery and ops workflows by providing insights, forecasts and estimation of risk of SLO violations.



The algorithms also provide warnings if the infrastructure and platform need to be re-optimized based on the changing nature of application behavior and of resource requirements compared to previous versions of the application. These pre-emptive warnings not only offer guidance during the application delivery lifecycle but are also highly significant for the CIOs who need to ensure that the application releases are quickly deployed in production environments, with no loss of service and impact on revenue. In the future, DevOps will be able to include data from CMOs, such as those related to marketing campaigns, as inputs to forecasting and predictions models which will allow to estimate the application usage in their planning cycle. Poola says, “With a workbench for Ops team that includes forecasting and prediction models, by adding the capability to ingest information from different sources, the collaborative circle will get wider from an organizational viewpoint.”

"We at AskLytics want to empower DevOps with insights into IT system performance & optimization using statsbased machine learning solutions"

Accentuating AskLytics’ future roadmap, Poola says that the next phase of the company will be to target the Ops and CIOs. The company plans to expand its product suite over this next stage of its journey by providing Ops with solutions for production environments and analytics for CIOs. Going forward with its geographical expansion plans, AskLytics continues to strengthen its organizational footprint across North America while spreading its reach far into the global markets of APAC region. “We see tremendous traction for Infrastructure as a Service and Platform as a Service in APAC, and that means more opportunities for us to proliferate in those markets. Nevertheless, catering the markets of North America constitutes our primary target,” adds Poola.



Introspecting on the very foundation of AskLytics, Poola mentions that the real focus is the people, and the technology is an enabler. According to him, while creating statistical models and machine learning algorithms is an exciting venture from a technical perspective, the real reason of AskLytics’ existence is to make the work life of DevOps teams more efficient and seamless. “We want to make sure that DevOps teams make use of the intelligence gained from machine learning solutions to improve their productivity. If we can empower 10,000 DevOps teams with smarter solutions, we have made a big difference by making their lives easier,” concludes Poola.

Company
AskLytics

Headquarters
Sunnyvale, CA

Management
Tiago Poola, Founder & CEO and Vladimir Volchegursky, Chief Data Scientist

Description
Empowers DevOps with insights into performance & optimization using systems management specific machine learning solutions