RTI's messaging infrastructure provides your trading applications with the capacity to accommodate exploding message volumes and the speed they need to keep ahead of increasing volatility.
Application-to-application latency is as low as 30 microseconds—even over Gigabit Ethernet. Latency remains consistently low under load, providing reliable and timely delivery when volumes spike and you need it most.
Low latency is preserved even when persistent messaging is required. RTI's persistence service can receive messages in parallel with other consumers—including over reliable multicast—rather than functioning as a broker that only forwards messages after they are stored. (A brokered configuration is also supported for transactional applications and those that require guaranteed consistency across subscribers.)
Time-critical data can be cached locally, in-memory, for instantaneous access when needed and to offload backend data stores. Message data can also be automatically cached, freeing applications from having to asynchronously process and manage incoming data that is not needed immediately.
Individual producer and consumer threads can sustain over one million messages per second, maximizing the amount of data that can be analyzed by a single consumer and providing plenty of headroom to accommodate exploding message volumes.
System wide, throughput is limited only by the network switching fabric: aggregate throughput of hundreds of millions of messages per second can be achieved. This supports the most demanding high-frequency trading applications and data feeds, such as the Options Price Reporting Authority (OPRA).
This level of performance is possible because message produces and consumers communicate in a true peer-to-peer manner, eliminating the message brokers, servers and daemon processes that constitute choke points in traditional messaging implementations. With RTI, the network switch is the only point through which all messages flow.
Use of reliable multicast allows dozens of consumers to subscribe to the same message stream with virtually no impact on latency or maximum sustainable throughput.
Streams can be efficiently balanced across thousands of multicast groups. Internet Group Management Protocol (IGMP) support enables message filtering at wire speed by IGMP snooping switches and routers.
Subscribers that cannot keep up with message rates can be automatically and flexibly mitigated. This prevents slow subscribers from impacting the performance of producers and other subscribers.
RTI makes efficient use of network and processor resources, allowing CPU cycles to be dedicated to deeper analysis, supporting higher throughput, or reducing the number of required servers.
Content awareness allows messages to be filtered based on field values, not just their topic or subject. This improves efficiency by freeing applications from having to process unneeded messages. When used with Complex Event Processing (CEP), content filtering offloads the CEP engine to reduce CPU overhead and maximize CEP throughput.
Message contents are sent in a compact binary format, separately from metadata, minimizing network overhead for improved latency and increased capacity. Message definitions (i.e., field identifiers and types) can be pre-configured or exchanged at startup time, rather than with each message. This provides the functionality of self-describing messages, including heterogeneous interoperability, but without their traditional overhead.
Quality of Service (QoS)—such as configuration of the reliability protocol, durability and delivery order—can be tuned per-stream and per-consumer, even for consumers of the same stream. This allows tradeoffs between latency, throughput and resource consumption to be made for each application based on its unique requirements.
Full control is provided over the messaging-layer memory and network bandwidth that can be used by a given application. This protects against resource exhaustion, which could otherwise significantly degrade system performance or deny service to another critical application.
Read about other benefits of RTI Messaging.
© Copyright Real-Time Innovations. 2007-2012. All rights reserved.
