Emad Benjamin

Virtualizing and Tuning Large Scale Java Platforms

9:00 AM MDT

The session will cover various GC tuning techniques, in particular focus on tuning large scale JVM deployments. Come to this session to learn about GC tuning recipe that can give you the best configuration for latency sensitive applications. While predominantly most enterprise class Java workloads can fit into a scaled-out set of JVM instances of less than 4GB JVM heap, there are workloads in the in memory database space that require fairly large JVMs.

In this session we take a deep dive into the issues and the optimal tuning configurations for tuning large JVMs in the range of 4GB to 128GB. In this session the GC tuning recipe shared is a refinement from 15 years of GC engagements and an adaptation in recent years for tuning some of the largest JVMs in the industry using plain HotSpot and CMS GC policy. You should be able to walk away with the ability to commence a decent GC tuning exercise on your own. The session does summarize the techniques and the necessary JVM options needed to accomplish this task. Naturally when tuning large scale JVM platforms, the underlying hardware tuning cannot be ignored, hence the session will take detour from the traditional GC tuning talks out there and dive into how you optimally size a platform for enhanced memory consumption. Lastly, the session will also cover vfabric reference architecture where a comprehensive performance study was done.

Virtualizing and Tuning Large Scale Java Platforms

11:00 AM MDT

The session will cover various GC tuning techniques, in particular focus on tuning large scale JVM deployments. Come to this session to learn about GC tuning recipe that can give you the best configuration for latency sensitive applications. While predominantly most enterprise class Java workloads can fit into a scaled-out set of JVM instances of less than 4GB JVM heap, there are workloads in the in memory database space that require fairly large JVMs. In this session we take a deep dive into the issues and the optimal tuning configurations for tuning large JVMs in the range of 4GB to 128GB.

In this session the GC tuning recipe shared is a refinement from 15 years of GC engagements and an adaptation in recent years for tuning some of the largest JVMs in the industry using plain HotSpot and CMS GC policy. You should be able to walk away with the ability to commence a decent GC tuning exercise on your own. The session does summarize the techniques and the necessary JVM options needed to accomplish this task. Naturally when tuning large scale JVM platforms, the underlying hardware tuning cannot be ignored, hence the session will take detour from the traditional GC tuning talks out there and dive into how you optimally size a platform for enhanced memory consumption.

Virtualizing and Tuning Large Scale Java Platforms

1:30 PM MDT

The session will cover various GC tuning techniques, in particular focus on tuning large scale JVM deployments. Come to this session to learn about GC tuning recipe that can give you the best configuration for latency sensitive applications. While predominantly most enterprise class Java workloads can fit into a scaled-out set of JVM instances of less than 4GB JVM heap, there are workloads in the in memory database space that require fairly large JVMs. In this session we take a deep dive into the issues and the optimal tuning configurations for tuning large JVMs in the range of 4GB to 128GB.

In this session the GC tuning recipe shared is a refinement from 15 years of GC engagements and an adaptation in recent years for tuning some of the largest JVMs in the industry using plain HotSpot and CMS GC policy. You should be able to walk away with the ability to commence a decent GC tuning exercise on your own. The session does summarize the techniques and the necessary JVM options needed to accomplish this task. Naturally when tuning large scale JVM platforms, the underlying hardware tuning cannot be ignored, hence the session will take detour from the traditional GC tuning talks out there and dive into how you optimally size a platform for enhanced memory consumption.

Virtualizing and Tuning Large Scale Java Applications

3:15 PM MDT

This session shares many of the production proven methods of running Java on vSphere. Covering how to size JVMs, and VMs for large scale deployments. The session will have a special section on GC tuning and show how a wide range of JVMs can be tuned using a GC recipe developed over the past 15 years of actual field experience in tuning JVMs.

Three key trends and associated tuning techniques are discussed in this session. The key trends are: Consolidation, Elasticity and Flexibility, and Performance

Consolidation
Many of our customers find that their middleware deployments have proliferated and are becoming an administrative challenge associated with higher costs. We see a trend across customers who look to virtualization as a way of reducing the number of server instances. At the same time, customers are taking the consolidation opportunity to rationalize the number of middleware components needed to service a particular load. Middleware components most commonly run within a Java Virtual Machine (JVM) with an observed scale of 100 to 1000s of JVM instances and provide many opportunities for JVM instance consolidation. Hence, middleware virtualization provides an opportunity to consolidate twice – once to consolidate server instances, and, secondly, to consolidate JVM instances. This trend is far-reaching, because every IT shop on the planet is considering the cost savings of consolidation.
One customer in the hospitality sector went through the process of consolidating their server footprint and at the same time consolidated many smaller JVMs that were less than 1GB heap. They consolidated many of these smaller 1GB JVMs into 2 categories, those that were 4GB, and others that were 6GB. They performed the consolidation in such manner that the net total amount of RAM available to the application was equal to the original amount of RAM, but with fewer JVM instances. They did all of this while improving performance and maintaining good SLAs. They also reduced the cost of administration considerably due to the reduced number of JVM instances they had to manage, and refined environment that helped easily achieve SLA.
Another customer, in the insurance industry, was able to achieve the same as the above customer, but additionally was able to over-commit CPU in development and QA environments in order to save on third party software license costs.
On the other hand, sometimes we come across customers that have a legitimate business requirement to maintain one JVM for an application, and/or one JVM per a line of business. In these cases, you cannot really consolidate the JVM instances, as that would cause intermixing of the lifecycle of one application from one line of business with another. However, while such customers don’t benefit from eliminating additional JVM instances through JVM consolidation, they do benefit from more fully utilizing the available compute resource on the server hardware, that otherwise would have been underutilized in a non virtualized environment

Elasticity and Flexibility
It is increasingly common to find applications with seasonal demands. For example, many of our customers run various marketing campaigns that drive seasonal traffic towards their application. With VMware, you can handle this kind of traffic burst, by automatically provisioning new virtual machines and middleware components when needed, and then automatically tear down these VMs when the load subsides.
In addition, the ability to change updating/patching hardware without causing outage is paramount for middleware that supports the cloud era scale and uptime. VMware VMotion gives you the ability to move VMs around without needing to stop applicators and or the VM. This flexibility alone makes virtualization of middleware worthwhile when managing large-scale middleware deployments. One customer in the financial space, handling millions of transactions per day, used VMotion quite often to schedule their hardware upgrades without any time downtime. What otherwise would be a costly scheduled downtime to their business.

Performance
Customers often report improved middleware platform performance when virtualizing. Performance improvements are partly due to the updated hardware that customers will typically refresh during a virtualization project. There is also some performance improvement due to the robust VMware hypervisor. A recent customer that reported a great level of performance provided the following testimony

“With our OrderExpress project we upgraded our Middleware Services, Commerce, Portal, WCM, Service Layer, DB2 Database; migrated from AIX to Linux; virtualized on VMware; moved the application into a three-tier DMZ; increased our transactions by over 150 percent; and added significant new capabilities that greatly improved the customer experience. Changing such a wide range of technology components at once was a huge challenge. However using VMware vSphere and additional architectural changes we were successful in improving performance by over 300 percent; lowered costs in the millions; improved security, availability, and scalability; and how we plan to continue evolving this application to maintain greater than 30 percent yearly growth.”

– Jeff Battisti, Senior Enterprise Architect at Cardinal Health

In this session, I will show some actual JVM and VM sizes for middleware components both small and large JVMs. Will also detail out GC tuning recipe that I have developed over the years,that has been shown to handle JVM heap sizes form 4GB to 88GB+, and higher. Of course the introduction of in-memory databases has driven the trend to have these larger JVMs and hence why we will discuss what is the best way to tune the JVM, VM, and the hardware platform they are deployed on.

I see the sizing question as the most commonly asked question with our customer base,and as a result I plan to focus on it during the session.