JavaBooster by bullion

financial_graphs_on_the_computer_screen_offre_0

Break Java limits and go real-time!

Powered by bullion and Zing® JVM

The JavaBooster platform is based on bullion, high-end enterprise x86 server from Bull (the Atos brand for technology products and software) and the Zing Java Virtual Machine (JVM) from Azul Systems®.

Bullion servers have a unique modular and scalable architecture, from 2 to 16 processors (up to 288 cores) and from 48GB to 24TB of memory. Bullion’s scalability means future needs can be addressed by simply adding or reallocating capacity which fully protects your investments. In addition, its highest standard of availability and quality of service coupled with the recent announcement that bullion beat the gross performance world record make it the enterprise server of choice for in-memory applications.

Another key component to address Java in-memory computing, is the Zing Java Virtual Machine (JVM) from Azul. Zing is designed to break the classical Java 150/300 GB practical memory limit per JVM instance (induced by Java Garbage Collector pauses) and expand it to 2 TB without performance impact. Zing enables essential applications to meet even the most demanding service level agreements (SLA) whatever the data volume and without depending on JVM tuning experts. This bullion–Zing combination named JavaBooster by bullion is a best of breed platform that addresses two different use-cases:

  • Solving classical Java performance or capacity problems
  • Enabling very large in-memory Java applications

Waive Java performance limits

Java continues to be a preferred choice for enterprises developing mission-critical platforms such as e-Commerce, web portals and electronic trading systems. Indeed Java can claim multiple competitive advantages:

  • A rich high-level language
  • A mature and widely adopted technology
  • Multi-platform: by design, Java aims at application portability. Developers can with a same high-level language create applications that may run on multiple environments to reach a broader market
  • Skill availability: there are more than 9 million Java developers worldwide, and many future talents will arise as Java is widely taught
  • Easy expandability: adding more memory to a Java environment is easy making the upscaling efficient and predictable

Though when reaching a few hundred of Gibabytes of memory, Java’s Garbage Collector (GC) processes tend to last longer and impact application response-time and the user experience. When facing a GC-based or latency-sensitive performance problem the JavaBooster platform is the ideal solution.

Large-scale in-memory Java computing

In-memory computing provides extreme velocity thanks to a direct access in a few microseconds to data stored “close” to the processor. A classical server hosts up to 1,5 or 3TB of RAM on which up to 90% maximum may be used for in-memory computing (the Operating System and a few side applications are using the remaining part). This caps the maximum amount of data that may be hosted in a single in-memory system.

One way to address this issue is to scale-out the application. This might be relevant in some use-cases, but commonly network latency (counted in milliseconds) collides with memory access time (in microseconds). This negates partially the in-memory performance advantage.

The ability to scale memory in a single server is key to leveraging the benefits of a big in-memory data set. This is possible with bullion. Specifically for applications within the financial sector, Atos has recently introduced the Real-time analytics appliance (powered by Java technology and JavaBooster by bullion platform).