29West makes product upgrades; introduces memory mapped transport

29West today announced the general availability of LBM release 3.4 and UME release 2.1.

  0 Be the first to comment

External

This content is provided by an external author without editing by Finextra. It expresses the views and opinions of the author.

These releases will greatly expand monitoring and support for added control to tune performance in a high loss environment.

Monitoring to ensure there is no loss and that receiving applications are not falling behind is the next key to optimizing system level performance, and ultimately, profitability for high frequency traders in banks, exchanges and hedge funds. LBM 3.4 provides additional monitoring information that allows application developers to monitor the performance and health of their messaging applications in real time, using graphical SNMP management stations or other monitoring tools. The added ability to monitor receiver queue depth ensures that the 29West ultra low-latency transport is being fully leveraged by the receiving application. LBM 3.4 and UME 2.1 include monitoring tools that are unique in exposing how far behind an application might be running due to this latency.

"There is a lot of attention focused on messaging latency, and rightly so," said Mike Garwood, director of software development at 29West, "as low-latency messaging is critical to optimal system performance".

"One thing we have noticed is that in many cases, our customers have a rock solid messaging layer, but there are 'silent killers' of latency buried in the broader system design," Garwood explained. "For example, if a receiving application is four messages behind, the transport messaging latency may no longer be the weak link in the chain. Until the receiving application is able to keep up, latency will increase as messages wait for the receiver to be ready to process them. By exposing queue depth information and other key parameters via SNMP, we can provide the application developer a real-time dash board on the true end-to-end performance of their system," he added. "This allows customers to see exactly how much buffering latency is being added on message receivers, a previously silent source of end-to-end system latency."

The improved stattated statistics are included as part of LBM 3.4/UME 2.1 and later, and can be used easily with 29West's SNMP Monitoring Agent 1.1 and later, or incorporated into other monitoring tools. All these products are in general availability today.

From customers building and running smart order routers and exchange order execution platforms to small proprietary trading firms, electronic market makers and foreign exchange systems engineers, 29West addresses ultra low-latency messaging needs for the financial industry. In all cases, users report that aggressive monitoring enables significantly better performance.

Separately, 29West today announced it will provide an Inter-Process Communication (IPC) transport using shared memory for its LBM and UME messaging solutions. The feature currently provides one-way messaging latency of roughly 3 to 5 microseconds with throughput of approximately 13 gigabits/second to applications running on a single multi-core commodity system. The 29West IPC Transport will be part of LBM 3.5, which will offer early access availability at the end of March, 2009, and general availability slated for May 1, 2009.

29West has developed this technology in response to the growing multi-core trend in the microprocessor industry, as well as the growth in multi-chip server design which is enabling truly unprecedented parallelization opportunities for financial application developers.

"Intel and AMD have made it clear that future development in microprocessor architecture is going to be horizontal," said Mark Mahowald, founder and CEO of 29West. "In this environment, where it is feasible to run multiple financial applications needing pub/sub messaging on the same machine, we can offer our clients untouchably low latency. We have always felt the best way to optimize messaging performance is to remove everything possible between the sending and receiving applications. We removed the servers and daemons in 2004; now, using shared memory removes the kernel and network stack as well."

Remove the network; remove the network latency. In high frequency trading, where time is counted in microseconds, the race is on to remove every possible source of latency in the system. In this race to zero, users of the IPC transport will have a definite edge.

29West's IPC transport will be available at no charge to all current 29West customers, and will come as a standard feature in future releases of LBM and UME. The new transport makes use of the same API calls, and only simple configuration changes are required to activate its use. The 29West IPC transport will be supported on Windows, Linux, Solaris and AIX, with other platforms being added as customer needs dictate.

From customers building and running smart order routers and exchange order execution platforms to small proprietary trading firms, electronic market makers and foreign exchange systems engineers, 29West provides the ultra low-latency messaging that financial industry users demand.

Leveraging the IPC transport on the same machine for multiple processes will give users yet another level of high performance.

Sponsored [Webinar] Weathering Macroeconomic Headwinds: How should CFOs invest in Tech?

Comments: (0)

[Webinar] The Next Fintech Frontier: A path to enhanced cross-border solutions amidst economic uncerFinextra Promoted[Webinar] The Next Fintech Frontier: A path to enhanced cross-border solutions amidst economic uncertainty