Archive for the ‘Benchmarks’ Category

Intel, QuIC Financial Technologies Partner to Test Faster, Lower Latency Solution

Wednesday, December 19th, 2007

QuIC Financial Technologies, provider of risk management, pricing and financial analytics solutions, has partnered with Intel for testing of Intel’s new Harpertown quad core 45Nm processors. QuIC’s testing of Intel’s new Harpertown processors showed promising double digit performance improvement results in client risk management and simulation trials, the partners say. QuIC’s testing also found that in using its latest generation of 45Nm quad-core IntelXeon processors, Intel has managed to deliver ultra low latency and significant performance increases in algorithmic trading, direct market access and market data delivery.

QuIC News Release External Link

Technorati Tags: , , , , , , , ,

STAC Tests Latency on HP Blade Servers with Gigabit Ethernet, Voltaire InfiniBand

Monday, December 17th, 2007

STAC has released a study of latency on HP c-Class Blade Servers with Gigabit Ethernet (GbE) and Voltaire InfiniBand. The system used Reuters Market Data System (RMDS) 6, running on HP BL460c Blades with 2 x Dual Core Intel Xeon 5160 3.0 GHz processors and Red Hat Enterprise Linux 4.4. STAC investigated latency on Gigabit Ethernet versus Voltaire InfiniBand using a traditional RMDS topology in an identical system environment. It also tested a stacked RMDS topology that maximised the utilisation of cores while also taking into consideration factors in real customer deployments.

Read More »

Intel, StreamBase, DSAL Do Further Real-Time Event Processing Tests

Wednesday, December 12th, 2007

Data Stream Analysis Limited (DSAL), provider of real-time complex data and event processing solutions, has released results from the second in a series of performance tests carried out at Intel facilities using the StreamBase complex event processing (CEP) engine running on Intel processors. The tests were conducted at Intel’s Low Latency Laboratory at Winnersh in Berkshire, UK, and were carried out on calculations of aggregated values used in algorithmic trading strategies. The results achieved were: 48 per cent - increase in updates processed per second; 53 per cent - reduction in the standard deviation between updates processed per second.

Technorati Tags: , , , , , , ,

Wombat Enters Data Fabric Space, Delivers All of OPRA on a 1U Server

Wednesday, December 12th, 2007

Low latency technology provider Wombat is running a fully subscribed pilot of a new middleware solution which it claims enables ultra high performance messaging with minimal data centre resource consumption.

The vendor reckons it has proven the power of Wombat Data Fabric and its ability to provide market-leading density on open, standard x86 hardware by running a full data from OPRA, the US options market, on a single Intel-based server available in a single rack unit (1U) form factor. In the test using a full day of OPRA data, the OPRA feed handler, running on a two socket eight core Intel 5400 Xeon server, published all updates on all symbols to MAMA clients across a single Voltaire network switch. Average CPU utilisation was just 10 per cent, escalating to less than 35 per cent at peaks, with a mean latency of less than 100 microseconds. No speciality or proprietary hardware devices, such as FPGA cards, are required by the solution, reducing cost and complexity when compared with alternative “appliance” approaches, Wombat reckons.

Wombat News Release External Link

Technorati Tags: , , , , , , ,

BEA Benchmarks Weblogic Event Server on Intel’s Caneland

Tuesday, December 4th, 2007

BEA Systems has published a benchmark of its CEP offering - Weblogic Event Server - demonstrating microsecond-level latencies when processing one million events per second. The benchmark was run on Intel’s Caneland platform, compring four quad core processors.
Read More »

IBM Supports STAC Council and STAC Lab

Friday, November 30th, 2007

IBM has become an official innovation sponsor for the Manhattan-based Securities Technology Analysis Centre’s (STAC’s) STAC Lab and has joined the STAC Benchmark Council. As an innovation sponsor, IBM will provide the Lab with its latest BladeCenter systems and IBM WebSphere Front Office for Financial Markets Version 2.0. IBM will, in turn, use the Lab as a foundation for testing workloads such as market data, analytics and execution for financial markets. STAC will add several IBM products to its “Racked-and-STAC’d” list, the inventory of hardware and software products with which STAC is ready to support customer projects. As a member of the STAC Benchmark Council, IBM will contribute to specifications under consideration by the Council’s participating trading firms.

STAC News Release External Link

Technorati Tags: , , , , , ,

Aspeed Doubles Performance With Intel’s Xeon 5400

Thursday, November 29th, 2007

Aspeed Software is the first of Intel’s financial markets software partners to publish performance figures leveraging Intel’s recently introduced Xeon 5400 chip. Benchmarking a single-threaded options pricing application, Aspeed reports that a 5400-based system delivers a 2.2-fold run time reduction while consuming just 13 percent more power, when compared to a Xeon 5100-based system.
Read More »

Intel Joins STAC Benchmark Council, Becomes Innovation Sponsor

Thursday, November 22nd, 2007

Intel Corporation has joined The Securities Technology Analysis Center (STAC) Benchmark Council and has entered into a multi-year lab agreement with STAC as an innovation sponsor. The agreement builds upon a benchmarking relationship that Intel established with STAC over a year ago. Intel has also made a multi-year commitment to provide the Manhattan-based STAC Lab with the latest Intel technologies. This includes its just-announced 45-nanometer Intel Xeon processor 5400 series (Harpertown), which preliminary STAC testing has found to decrease power consumption by 20 per cent over a previous Xeon processor 5300 series (Clovertown) when supporting the same market data workload.

STAC News Release External Link

Technorati Tags: , , , , , , , ,

New STAC Study Measures RMDS 6 on IBM System x3650 with Novell SLES 9

Wednesday, November 14th, 2007

STAC has released a study of the Reuters Market Data System (RMDS 6) on the IBM System x3650 server with 2 Dual-Core Intel Xeon 5160 3.0 GHz processors, running Novell SuSE Linux Enterprise Server 9.3 and communicating over Gigabit Ethernet.
In the project, STAC optimised and benchmarked the Reuters Market Data System in a traditional topology on the IBM servers. The results are the highest Point-to-Point Server fanout and highest MarketFeed throughput for a Source Distributor achieved to date on a two socket server for this topology:

• 2.59 million outbound user updates per second in the Point-to-Point Server fanout test
• 420,000 updates per second through a Source Distributor machine with MarketFeed

Additionally, this stack was able to achieve the highest throughput, 550Kups, in a low latency configuration of RMDS in a traditional topology with Ethernet.

The report is available to download at http://www.STACresearch.com.

STAC News Release External Link

Technorati Tags: , , , , ,

BEA Benchmarks WebLogic Event Server Performance, Proves Low Latency at High Data Rates

Wednesday, November 14th, 2007

BEA has released a new benchmark study on BEA WebLogic Event Server performance, which it says demonstrates its ability to provide low latency at very high data rates. BEA entered the event driven architecture market in May, with the launch of WebLogic Event Server – pitched as the first and only Java application server designed for event-driven applications.

The benchmark application conducted in the study implemented a capital markets front office scenario in which an application monitors an incoming stream of market data watching for the occurrence of certain conditions that will then trigger an action. The study revealed that WebLogic Event Server was able to sustain an event injection rate of up to 1 million events per second while maintaining microsecond latencies on a Quad-Core Intel Xeon processor 7350 series-based server. At this injection rate, the average event latency for the full processing path within the server was 67.3 microseconds, with 99.4 per cent of the events processed in less than 200 microseconds and 99.99 per cent processed in less than five milliseconds.

The report is available for download at www.bea.com/eventserver

BEA News Release External Link

Technorati Tags: , , , , , ,

You are not logged-in