Author: Adam Hawley
The vSAN team is excited about the chance to present a technical preview* of Project Magna for vSAN at VMworld as the start of a major commitment from VMware to leverage AI/ML technology to deliver the Self-Driving Data Center.
vSAN is at the heart of HCI offerings from VMware. It is also at the center of delivering the VMware Cloud Foundation to deploy a hybrid cloud, and VMware Cloud on AWS for public cloud infrastructure. Consequently, vSAN is the perfect place to start simplifying hybrid cloud operations through the use of AI/ML.
A key tenant of HCI systems is that they should make IT operations simpler to execute and scale over time. This is exactly what AI/ML helps you do: it helps you adapt and optimize your infrastructure continuously with only simple guidance from you. By automatically adapting vSAN behavior, you to get the performance and efficiency with a minimal amount of expertise or intervention from your staff.
Going forward, I’m going to refer more specifically to machine-learning (ML) or reinforcement learning (RL) rather than use the more generic “AI”. It is important to be clear that we’re talking about using the real thing here and not just putting a spin on traditional analytics methods as others are doing.
The distinction matters because real ML, and specifically RL in the case of Project Magna, can optimize and adapt itself to achieve a better result – can make itself better – without manual intervention or reprogramming. Traditional analytics can be very powerful. But they are also not able to adapt themselves, particularly if they encounter a scenario that was unforeseen by the programmer. ML is inherently better at dealing with complexity and ambiguity to determine the best way to achieve your intention (e.g. your SLA or your performance KPI, etc.) even if the rules for getting there were not explicitly programmed in.
With Project Magna, what we’ll be showing at VMworld is how we are using an ML technique known as “Reinforcement Learning” to optimize the performance your applications experience from vSAN.
The easiest way to explain how reinforcement learning works is to liken it to a new video game you are learning where the object, i.e. your “intent”, is to maximize your score: As you play and take action, you quickly learn what to do, and what not to do, if you want to maximize your score. In other words, you are guided in your learning through the “reinforcement signal” of your score going up or down based on what actions you take.
With Project Magna for vSAN performance, we have embedded a reinforcement learning model into the solution, leveraging vRealize Operations, that can manipulate aspects of the caching and storage performance tier to constantly adapt and optimize to your intentions. It takes simple guidance from you about whether you want to optimize for read latency, for write latency, or some balance and then continuously adapts the system even as the workloads continuously change profile across the hours, days, and weeks.
Project Magna can reduce the time burden on your staff to get the most out of your HCI system with vSAN: Even if you already have an excellent level of performance-tuning expertise, you don’t have the time to sit in front of the system and manually optimize the parameters 24 hours a day, 7 days a week, 52 weeks a year. The application workload profile in your environment is constantly changing, and it is not realistic to expect manual tuning to stay optimal for long. The system will drift away from the optimal point, potentially even within minutes or hours. But the reinforcement learning technology is on the job non-stop.
Of course, we are not stopping here. But Project Magna is a solid first step on the path to the Self Driving Data Center.
For more context on Project Magna and VMware’s vision to deliver the Self-Driving Data Center, check out this blog: Project Magna for vSAN Adaptive Optimization
For more detail information about Project Magna, please visit website.
For any other questions or more information, please don’t hesitate to email us at: firstname.lastname@example.org
* Please note that there is no commitment or obligation that technical preview features will become generally available.