29West's UME Shatters 100 Mircosecond Persistent Latency Benchmark

Chicago - 11 May 2009

29West, Inc., the market leader in high performance low-latency messaging solutions, today releases its report, '29West Ultra Messaging(r) for the Enterprise (UME) Persistent Performance on Commodity Hardware with Gigabit Ethernet.' The report focuses on the performance of UME for persistent operation on low-end commodity hardware with untuned out-of-the-box Linux operating systems.

Highlights of the report include: delivery latency as low as 56 microseconds, throughput as high as 1.2 million messages per second, and a demonstration that delivery latency of long running receivers is unaffected by failure and recovery scenarios of other receivers. The results highlight the low total cost of ownership using the UME solution, which since its launch in 2006 has been implemented in production at many firms worldwide. This report will be one of the subjects at 29West's Zero Latency Tour events, beginning later this month.

Current market conditions have caused messaging rates to vastly increase. Mission-critical components inside a trading system demand not only low latency but also persistent operation in order to ensure all trades are executed.

"We tested UME's delivery latency in various ways," says Greg Lorence, UME product manager and author of the report. "One-way latencies averaged 56 microseconds for low-volume data. We tested how much data could be pushed through our commodity systems, and this was close to 1.2 million messages per second. Even increasing message sizes 15 times had no significant effect on latency with UME. Nor did recovery of a failed system component have any impact on other participating receivers."

Mark Mahowald, president and founder of 29West, says, "29West is winning head-to-head product evaluations consistently, and our customers report 50% reductions in the hardware costs needed to support messaging volumes when using UME compared to other designs."

"In today's climate firms with limited resources, to run full pilots, are doing paper evaluations," says Mahowald. "We wanted to publish a range of UME performance results to meet this need. Our unique Parallel Persistence design has no scalability concerns so we can add receivers with virtually no impact on the senders or the persistent store. As you add receivers, the aggregate throughput growth will be nearly linear. We will be demonstrating this throughput scalability in future performance reports."



Become a bobsguide member to access the following

1. Unrestricted access to bobsguide
2. Send a proposal request
3. Insights delivered daily to your inbox
4. Career development