Testing involved two benchmarks: highest throughput and lowest latency. Informatica achieved these test results with the Cisco Unified Computing System:
o Latency as low as 660 nanoseconds
o Maximum fan-out throughput of 21.1 million messages per second.
• The test results are outlined in a new report, Informatica IPC Shared-Memory Transport Performance Benchmark.
• Based on an industry standard x86 platform, the Cisco Unified Computing System portfolio supports ultra-low message latency and reliability in automated trading conditions, which can help businesses to meet changing market demands and establish a competitive advantage.
• Impressive results measuring in at an order of magnitude faster than the fastest network-based transports previously tested.
• Enables full trading systems to be run on a single server as opposed to a rack of servers, potentially eliminating hundreds of microseconds from trade latencies by replacing a network transport with an shared memory-based transport called Interprocess Communication (IPC).
• Current Informatica customers already in production have taken advantage of this performance with no code changes across all messaging modes including streaming, persistent, and low-latency load balancing modes.
• The IPC transport replaces the role of the network in routing data messages among applications all running on the same physical machine. For example this allows for market data feed handlers to feed automatic trading strategy applications that in turn generate stock orders all within about a millionth of a second.
• The Ultra Messaging products tested are generally available and in production at several leading financial trading firms.
• The Cisco Unified Computing System unites compute, network, storage access, and virtualization into a scalable, modular architecture that can be managed as a single system. Based on industry standards, the Unified Computing System is a new computing model that uses integrated managemenent and combines a "wire once" unified fabric with an industry standard computing platform to optimize virtualization, reduce data center total overall cost, and provide dynamic resource provisioning for increased business agility.

Additional Statistics
As a result of the tests Informatica was additionally able to show:

• Near perfect linear horizontal scaling to multiple CPU cores implying that as Moore's law continues to provide more computing power by adding additional cores to processors the additional power will be fully utilized.
• Observed latencies were significantly under one microsecond all the way through 512-byte messages; showing predictable and stable increases in transport latency as message size increased.