Performance and Scalability

RTI Product Tour

You have to meet demanding performance requirements. And most likely, they'll be more demanding next year. You also need to leave room for new requirements, unforeseen problems, and other things. You don't design your system up front to use 100% of your machines' CPU, memory, and storage resources. Nor should you base your system on a communications infrastructure that's just "good enough" to get by.

A communications product that not only meets but exceeds current requirements for performance and scalability reduces risk in key areas:

• System components will be less likely to fail as a result of not receiving the data they require in a timely way.
• The system will be less likely to outgrow its infrastructure and need to be redesigned.

It also saves you money: you don't need to purchase hardware acceleration, and fewer machines can handle the same data volume. That adds up to lower administrative costs and decreased demands for power and cooling.

RTI Data Distribution Service provides top performance, whether measured in terms of latency, throughput, or real-time determinism.

• Efficient architecture - Eliminate performance bottlenecks and single points of failure.
• Performance-optimized implementation - Minimize overhead on your network and computing resources.

The result is an information backbone that not only delivers low latency at high throughput but also scales from small embedded boards to global enterprises.

Efficient Architecture for Demanding Workloads

One reason for RTI's performance and scalability is its elegant peer-to-peer architecture.

Other middleware products require servers to broker messages (whether running on dedicated machines or as daemon processes on each machine). These introduce throughput and latency bottlenecks as well as timing variations. Brokers also increase system administration costs and represent single points of failure within a distributed system, putting data reliability and availability at risk. Products that depend on them can break down or suffer latency spikes when traffic volume surges—which is typically when you need them the most.

RTI doesn't use brokers. Messages flow directly from publishers to subscribers with minimal overhead. Without a broker to impede performance, RTI Data Distribution Service is able to deliver over a million data updates per second and to provide extremely low latency, even when network conditions are suboptimal. Only RTI delivers consistently high performance.

Performance-Optimized Implementation

RTI software is architected for high, deterministic performance.

With RTI Data Distribution Service, individual single-threaded applications can stream up to 950 megabits of data per second with inter-node latencies as low as 43 microseconds (as measured over Gigabit Ethernet between 2.0 GHz Opteron PCs). These rates are 10 to 25 times faster than can be achieved with most other messaging solutions, including JMS.

Zero-Copy Interface

RTI Data Distribution Service allows messages to be sent and received without any copying of data between the application and middleware. This reduces latency and processor overhead, particularly for large messages, such as are often used to distribute high-rate streaming data.

Highly Configurable Quality of Service

RTI Data Distribution Service provides over 20 configurable Quality of Service (QoS) options that allow its performance to be optimized based on the requirements of individual applications or message topics. For example, performance can be optimized to maximize throughput or to minimize latency or based on whether best effort or reliable messaging is required.

Large System? No Problem

As your system scales up, so does RTI Data Distribution Service. In fact, in RTI's testing, scaling from one subscriber per topic to almost 900 led to a decrease in throughput per subscriber of less than 15%.

Multi-Core

RTI Data Distribution Service is ready for multi-core systems. The programming interface is designed for concurrent use, and even single-threaded applications will take advantage of additional cores as the middleware efficiently processes incoming data and periodic events in the background. Smart concurrency control, based on patented lock-less data structures, maximizes concurrency across data streams while minimizing unnecessary context switches and system calls within each stream.

Publish-Subscribe over Multicast

RTI Data Distribution provides a publish-subscribe messaging model and takes advantage of multicast when supported by the underlying transport. When using multicast, messages are simultaneously delivered to multiple subscribers with minimum latency and minimum overhead on the underlying network. Both reliable and best effort delivery modes are supported.

RTI allows a single topic to be partitioned across multiple multicast groups. This allows per-topic throughput to exceed the capacity of any single consumer or network link. IGMP snooping allows your network switches to keep unneeded messages off of your network.

Content-Based Filtering

RTI Data Distribution Service is message-content aware and allows subscribers to specify filter expressions that are applied to each message before it is sent. By filtering unneeded messages in the middleware layer on the sending side, overhead is reduced on both the network and receiving node. In addition, application logic may be simplified if message recipients no longer have to analyze each message for relevance.

Messaging systems that require data to be sent to a central server for filtering cannot offer this performance optimization.

Site-to-Site Routing

Multicast is great on a local network, but when you need to connect systems across a wide-area network or the Internet, you need a tool that’s optimized for that environment. RTI Routing Service can bridge DDS-based networks to one another across a WAN. At the same time, it can cleanse and transform your data to adapt to the different external interfaces provided by independently developed subsystems.

Scaling Down

Unlike traditional messaging middleware, RTI Data Distribution Service, with its low-overhead architecture, is also able to support resource-constrained devices. In addition to embedded versions of Windows and Linux, it also runs under most major real-time operating systems.

For very small systems, for example, with less than a megabyte of memory, especially those with high assurance requirements, RTI offers RTI Data Distribution Service, Safety-Critical Edition. Safety-Critical Edition is the first commercial, standards-based middleware solution for developing distributed high-assurance applications.

See the data and details about performance and scalability or continue the tour to learn more about how RTI adds value with built-in data recording and analysis tools…