64bit.biz - Sun Microsystems servers, storage and workstations, VMware virtualization, Juniper firewalls, Oracle database

Newsletter by e-mail

Sales department

ICQ: 553656844
Skype: arman.ks
Jabber: akirakosyan@abclinuxu.cz

ICQ: 34043399
Skype: saki-fk

Tel.: +420 267 108 384

Discount products

Favourite servers

Product bargain

News

Sun Microsystems and GigaSpaces break 1.3 million messages per second benchmark on SPARC T5240 server / 11^thOctober 2008

Sun announced new benchmark performance results that address the scalability needs of capital market customers. In rigorous performance tests, the energy-efficient Sun SPARC Enterprise T5240 Server processed 1.3 million transactions per second with 60 concurrent clients to achieve an improvement of up to 40 percent over competing multi-threaded platforms. The average throughput per client thread was sustained at about 21,000 transactions per second. When matched against comparable x86 systems, users get up to five times higher performance in half the space and 2.5 times better performance per watt.

Bookmark this page

"This is great news for our customers in the financial services sector that need high-performance computing platforms to meet severe scalability, throughput and latency demands. Increased trading volumes require enterprise-class systems and software that can process enormous amounts of data-reliably and efficiently-while also maintaining the ability to scale for future growth," said Ambreesh Khanna, Head of the Global Financial Services Industry Group, Sun Microsystems. "Through our collaboration with GigaSpaces, we are delivering proven, record-breaking solutions to meet and exceed the exacting demands of these customers."

As a result of Direct Market Access (DMA) and electronic and algorithmic trading, parallel processing and latency have become major issues to consider when receiving quotes and executing orders. This need is even more critical today as customers look for ways to quickly and cost-effectively scale their highly concurrent applications. Unfortunately, a large number of applications cannot be easily re-architected to scale out on multiple servers.

The SPARC Enterprise T5240 Server with third-generation CoolThreads technology running the GigaSpaces eXtreme Application Platform (XAP) addresses this problem with its scalable in-memory data grid. The joint solution provides the ability for applications to easily scale up to extreme and virtually unprecedented levels of throughput and latency, as demonstrated by the new Sun-GigaSpaces scalability benchmark.

"We have added more and more clients performing business transactions with the system coping with the demand for additional throughput with linear scalability," said Nati Shalom, GigaSpaces Founder and CTO. "The SPARC Enterprise T5240 managed to process 1.3 million transactions per second with 60 concurrent clients whereas the average throughput-per-client-thread is sustained at about 21,000 transactions per second. We are yet to see such linear scalability at these levels of application threads."

Sun's SPARC Enterprise T5140 and T5240 servers are one of the industry's first dual-socket CMT platforms, delivering the highest compute density ever, with up to 128 threads in as little as 1 RU of physical space, while maintaining maximum power efficiency to achieve outstanding levels of scalability and performance per watt. Threads are inexpensive in Sun's processor design, providing extra headroom for free.

The SPARC Enterprise SPARC T5120 and SPARC T5240 system architecture incorporates third-generation CoolThreads Chip Multithreading (CMT) technology that provides users increased and extremely scalable computational density while staying within variously constrained envelopes of power and cooling. The very high levels of integration found in the SPARC Enterprise T5140 and T5240 servers helps reduce latency, lower costs, and improve security and reliability. In addition, balanced system design provides support for a wide range of application types-from Web services to high performance computing (HPC).

News archive

Login