Java has been the industry standard for all kinds of business applications. The reason of success had been the platform portability, perfect OO programming, Evolution of 3rd party frameworks & tools, and ofcourse Open source nature.
There had been one concern though. the performance. When Java released 1.3 way back in time a decade ago, it was featurefull but sluggish. As Java eveolved over all these years to current release Java 6, It has gained lots of Virtual machine optimizations. There were improved memory management using intelligent GC. We have seen compile time and compiler optimizations which enabled improvement in run time performance.
Comparing the performnce of 1.3 vs. 1.6, you will see differences of Big Leap. There are certain computational aspects where VM is optimised to the level of C performance. There had been certain benchmarks which showed that, but lets not get into that.
What we are here to discuss will change the way Business works. Today, time/ mission critical Realtime application cannot dream of using Java at their core. But if this post is followed, its quiet possible.
Long ago, it was thought that what if we could design a chip that could accelerate VM’s computations to significant numbers. It was a long from being practical. But it had been made possible by a Product vendor named Azul systems.
Azul has designed a hardware infrastructure that promises to boost Java by 5x times. Thats a lot!
The Azul Vega™ 3 7300 and 3300 Series compute appliances offer highly scalable compute capacity for Java-based applications, with up to 864 processor cores and 768 GB of memory in a flat SMP configuration. The unique software and hardware-assisted features of the Azul solution enable applications to transparently tap into this capacity over the network and reach unprecedented levels of scalability, throughput, and performance consistency. With this production-proven architecture the Azul Vega 3 Series has set the performance record on industry standard SPECjbb2005 and XMLMark tests.
The Azul technology really is astoundingly cool (if a bit heavy on your budget.) The Vega 2 series maxes out with an amazing 768 cores and supports 768 Gb of ram in a flat SMP configuration. This thing really is the closest thing in the Java world to infinite muscle.
It isn’t unusual to see Azul beat competitors by a large margin in industry-standard benchmarks. (Azul’s model 7280 provided 26X to 45X the throughput of the Sun T2000, and 18X to 33X the throughput of the dual-socket, dual-core AMD Opteron.) Not only does Azul provide amazing hardware, they also provide the Azul VM, a Java VM dedicated to this kind of HTC environments.
The Vega 3 Series appliances advance the unique business benefits delivered by the Azul solution, and are the fastest path to maximizing the ROI of business-critical applications. Without changing application code or architecture, applications can deliver 5X to 50X greater scalability and throughput, while providing consistent performance even under unpredictable load. The Azul solution also reduces server sprawl by an order of magnitude, simplifying existing deployments and reducing datacenter costs by 30% or more.
The new Azul Compute Appliances are built around the Vega 3 processor, the first 54-core chip designed and optimized for Java workloads. the company claims that the highly scalable Azul Virtual Machine allows individual application instances to scale to 670 GB of memory heaps and hundreds of processor cores in parallel without changing the application code. This breakthrough scalability is achieved with unique capabilities such as hardware-assisted Pauseless Garbage Collection, which eliminates the impact of garbage collection-related application pauses, and Optimistic Thread Concurrency, which minimizes the impact of scalability bottlenecks caused by memory lock contentions.
The Hardware:
The Azul Compute Appliance is available in two hardware configurations, depending on the number of processors and memory required.
- 3300 Series is provided in a standard 5U chassis allowing between 108 and 216 Vega processor cores, and between 48 and 192 GB of memory
- 7300 Series is provided in a standard 14U chassis allowing between 432 and 864 Vega processor cores, and between 192 and 768 GB of memory
Management tools
The software management capabilities of the Azul Compute Appliance provide configuration and monitoring of appliance and application resource use and health, with standard integration into common network management systems.
- Compute Pool Manager (CPM) provides a complete array of capabilities to configure, manage and monitor status for all hardware elements of the compute pool. CPM lets users snapshot and track appliance capacity utilization and application resource usage.
- REALTime Performance Monitor (RTPM) is a diagnostics and tuning tool instrumented into AVM. RTPM offers a uniquely detailed view of the real-time activity of both the application and the underlying VM. RTPM is the perfect tool for rapidly diagnosing scalability, response time and throughput problems in record time.
- Status and Alert Interfaces. CPM supports SNMP monitoring and alerts through standard NMS platforms. Generated CPM events can be directed to syslog output, and forwarded via SMTP to a user-defined email alias, enabling administrative personnel to easily stay abreast of compute appliance status.
- In-Band and Out-of-Band. Management features are accessible either as part of the data network or through a separate management network interface.
Here is the performance improvement that the company claims:
The Vega 3 7300 Series is uniquely positioned to address many of the highest priority requirements for Java-based applications:
- Breakthrough Scalability for Extreme Transaction Processing
– Offers the fastest path to delivering Extreme Transaction Processing performance, without changing the application’s code or architecture
– Meets the unique scalability, reliability and memory needs of Java applications
- Engine for Service Oriented Architectures
– Provides industry leading Enterprise Service Bus scalability
– Serves as a consolidation platform for multiple SOA services, providing dynamic resource allocation between services to support highly unpredictable workloads, and eliminating IO bottlenecks between services
- Power Stations for Data Grids
– Increases the maximum amount of data cached on the grid by 50X
– Increases data throughput by 20X
– Allows applications to co-reside with the cached data on a single system, increasing application performance with faster data access
- Platform for Shared Services
– Constitutes an ideal platform for shared services, providing shared compute capacity to multiple applications over the network
– Preserves OS, resource and security isolation between applications
– Reduces number of conventional servers by 10X or more
– Reduces TCO by 50%
- Industry’s Most Efficient Delivery of Compute Capacity
– Reduces server sprawl with up to 33x more throughput than conventional x86
– Reduces power, cooling and space requirements by 50% – 95%
Architecture:
Vega3-based Azul Compute Appliances transparently provide massive amounts of shared processing and memory resources to Java™ applications on the network. Each rack-mounted appliance simultaneously delivers up to 864 processor cores and 768 GB of memory to multiple Java applications that are launched from their original servers through the network and execute on the appliances using Azul Virtual Machine (AVM) software, a fully compliant Java Virtual Machine™ (JVM) implementation. The integrated Compute Pool Manager (CPM) software manages appliances, compute pools, and resource allocation for applications using powerful administrator-defined policies. REALTime Performance Monitor (RTPM) provides a detailed view of the real-time activity of both the application and the underlying VM.
Hardware specs can be found here.
There is alot company has revealed about its products, do check them out.
The Era of RealTime mission critical Java based application has already arrived. I know of certain Telecom, Banking companies that have employed such solutions in their applications and their experience with the appliance is breathtaking.
Azul has no plans of supporting the project via Open source license. We would look forward for a project may be by Sun to make an equivalent platform using SPARC systems scaling upto hundreds of Processor cores and Terabytes of RAM. The game is endless.
loading...
loading...
nice, i liked Azul systems!
loading...
loading...
hey
ive been doing recently research and found out this blog post regarding tunning applications using different JVMs
http://bigdatamatters.com/bigdatamatters/2009/08/jvm-performance.html
loading...
loading...