How Granulate supports Intel’s high-performance computing efforts

Enterprise

We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 – August 3. Join AI and data leaders for insightful talks and exciting networking opportunities. Learn more about Transform 2022


In the latest move to enable high-performance computing for enterprises, Intel has scooped up Israel-based Granulate, a startup that uses AI to perform workload optimization in real-time.

The deal, which is reported to be valued at $650 million, gives a major push to the chip giant’s portfolio of optimization tooling that ensures its architecture delivers maximum value when dealing with present day’s challenging compute demands in both cloud and on-prem environments. 

“Granulate’s innovative approach to real-time optimization software complements Intel’s existing capabilities by helping customers realize performance gains, cloud cost reductions and continual workload learning,” Greg Lavender, CTO, senior VP and general manager of Intel’s software and advanced technology group, said.

So, what does Granulate do?

Granulate’s platform uses algorithmic models to learn specific resource usage patterns and data flow associated with an application and identifies contended resources, bottlenecks and prioritization opportunities to work upon. Once this process ends, it uses all the gathered information and executes operating-system-level scheduling and resource management decisions to improve the infrastructure’s application-specific performance. 

“The Optimizer takes four days to learn the application data flow and the processing stages, and then, once activated, optimizes resource management decisions,” Asaf Ezra, co-founder and CEO of Granulate, previously told VentureBeat. As part of this, it works on aspects such as memory allocation, process swapping, cache and thread scheduling, among other things. 

The whole process works without any human intervention or code change and directly improves performance in terms of cost, throughput and latency. In fact, a number of enterprises, including PicsArt, Coralogix, Nativo and Singular, have already reaped these benefits by signing up for Granulate. The solution’s software agents can be installed on any Linux server in data centers or cloud environments, including virtual machines.

The Intel association

Intel, on its part, has been closely associated with Granulate since its early days. The startup was the inaugural member of Intel Ignite, an accelerator formed by the chipmaker to create and support select groups of promising, early-stage companies. 

During this program, Granulate closely collaborated with Intel to demonstrate how its solution improves the performance of 2nd Gen Intel Xeon Scalable processors when handling key Java-based business software workloads. Then, building on that work, the companies debuted a commercial workload optimizer as part of Intel’s cloud tools for optimization and management

This offering did exactly what Granulate promised (performance gains and reduced costs) but only for enterprise customers relying on Intel’s Xeon deployments. 

Mobileye, an Intel subsidiary that implemented the optimizer in its hybrid cloud environment, saw as much as 45% faster job completion and a 44.5% reduction in cloud instance costs. In some cases, the solution can even increase throughput by 5 times while decreasing cost by up to 60%, the company said. 

Full-stack optimization

Now, with the deal set to close in the second quarter of 2022, Intel will look to continue the work it has done with Granulate on workload optimizer, scaling it and taking it to more enterprises across its data center and cloud portfolio.

“Together with Intel, we believe we can help customers achieve meaningful cost reductions and five times the throughput across workloads,” Ezra said. He also indicated that the offering will be expanded under Intel’s umbrella but did not share specific details.

According to Gregg Siegfried, a Vice President, Analyst on Gartner’s I&O monitoring team, the move gives Intel a strong portfolio of tools to optimize at the CPU, operating system, cluster and application level and ensure that workloads running on its platforms exhibit superior performance.

“Granulate works at the operating system level, and can continuously optimize the Linux kernel to maximize the performance of a workload at any given time, which can materially benefit users of public cloud services by requiring resources in aggregate – particularly in concert with the Densify-based Intel Cloud Optimizer (also a part of cloud tools for optimization and management),” he said.

At the application layer, there is an Intel Ventures company called AIOBench, which addresses the optimization of application workloads via benchmarking. For CPU, tools such as Performance Maximizer and Turbo Boost enhance the performance chips with an eye toward the specific workloads of gaming – where overclocking is common – and real-time or embedded applications.

“Currently, AI-driven workflow orchestration space includes Turbonomic (IBM), Virtana, Densify, as well as newer players such as Sedai and aforementioned AIOBench. Not all of these are appropriate for every workload or application, but being able to optimize for cost in some cases, and performance in others, at multiple layers of the stack using Intel branded tools certainly keeps the Intel platform attractive for both cloud providers and developers,” Siegfried noted. 

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.