top of page

Risk-free optimization consulting, guaranteed results - Schedule your call today!

Want to know how to optimize your spending?: Estimate your saving here

Risk-free optimization consulting, guaranteed results - Schedule your call today!

Cloud Optimization Driven by Machine Learning: Optimizing Cloud at Runtime with Granulate

TLDR: Granulate is a groundbreaking real-time continuous Cloud Optimization platform that enhances application performance. Catering to workloads such as JVM-based tasks, Python, and Go, Granulate stands out by offering up to 45% cost savings on compute costs without necessitating code changes. Its unique agent-based solution, combined with a detailed learning mode, ensures optimal thread reordering and resource scheduling, making it ideal for businesses seeking efficient cloud optimization for their Kubernetes and Big Data workloads.



Intel-Granulate is a new-generation FinOps solution that focuses on improving the performance of applications. What sets them apart is that they can achieve up to 45% cost savings on compute costs without any code changes. All optimizations occur autonomously at the runtime level.


How Does It Work?

  • Agent-Based Solution: An agent is deployed into the environment depending on the architecture. This agent is installed on each VM or instance.

  • Workload Discovery: The agent identifies the application runtimes in the environment and installs a suitable runtime module for each application. Supported workloads include JVM-based tasks (Java, Scala, Kotlin) and languages like Python and Go.

  • Learning Mode: Once the runtime module is installed, it starts learning the resource scheduling mechanisms of the application. For JVM-based tasks, it might look at thread execution order. By profiling the application, it can determine which threads should be reordered to reduce wait time on the CPU.

  • Data Collection: The agent collects data in 20-second batches and sends it to Granulate's backend for resource-intensive learning processes.

  • Activation: After learning, the optimal configuration is loaded back to the agent. Upon activation, performance optimizations are realized.


Examples of Runtime Optimizations

  • Thread Scheduling: For JVM-based workloads, the default is first in, first out. However, by profiling, they can reorder threads to reduce wait time.

  • Crypto and Compression Optimization: The agent can finely tune compression parameters at the thread level, giving it more control and optimizing when compression occurs.

JVM Optimization


The Engagement Process:

  • They start by profiling the application with an open-source profiling tool to estimate the optimization potential. The ‘g-profiler’ agent ‘learns’ and identifies patterns in data flow, resource usage, and processing stages. It’s recommended to run “g-profiler’ on the production system, for more accurate profiling.

  • After installing the agent and going through the learning process, they will activate the optimizations. The “sAgent" is the agent used for profiling patterns and determining the type of optimization. s-agent is particularly effective for Python, Java, and Go, achieving 15 to 50% CPU optimization.

  • A benchmark is then done comparing one week before and after activation to validate the achieved optimization potential.

  • gcenter is the dashboard that allows users to activate or deactivate the agents on any workload and at any time.


For which use cases?

The 2 prominent use cases are K8s clusters Optimization, and Big Data Optimization.


Kubernetes Clusters Optimization

Granulate is compatible with EKS and Karpenter and can work alongside any auto-scaling solution.

Kubernetes Cloud Optimization

The optimization starts with workload rightsizing, followed by resource rightsizing. Then comes runtime optimization, followed by another round of resource rightsizing . It’s a continuous feedback loop.


Big Data: Spark workloads Optimization

Granulate can enhance Spark workloads, improving CPU usage and performance. They offer optimization for Databricks resource management, performing auto-scaling better than Databricks' native auto-scaling solution.

Big Data Optimization
No Wasted Space


Is it secure?


Conclusion

Granulate redefines cloud optimization by autonomously enhancing performance at the runtime level, setting it apart in a market filled with standard optimization tools.





🌐 Optimize Your Cloud Performance with Expertise!

Impressed by Granulate's groundbreaking cloud optimization capabilities? At OptimNow, we specialize in guiding businesses towards efficient cloud optimization and FinOps solutions. Let us help you navigate and implement these advanced technologies seamlessly.


📞 Contact OptimNow today for tailored guidance and ensure your cloud operations are at their peak performance!





Comments


bottom of page