Real-Time Non-Stop Linux Project Activities

Activities

2001:

Linux real-time and high-availability study.

The project studied Linux latency and response times in the case of failing node in a high-availability cluster. The cluster had primaries and secondaries that took over in case of failure, but they didn't perform very well from the low-latency point of view. Overall results subsequently measured real-time Linux response time to network events under processor load.

2002:

Embedded Linux and CORBA study.

The previous project was extended to use CORBA on Linux and simultaneously also in an embedded system. CORBA requests were used to access a simple database, and the time spent on processing the requests was measured. We found severe performance problems with PostgreSQL. Some of the embedded systems did not have sufficient processing power to use CORBA and achieve a low enough latency.

Carrier Grade Linux specification work.

OSDL (Open Source Development Labs) is doing specification called Carrier Grade Linux (CGL). The project studied one small requirement named Hot Device Identity, or later called Persistant Device Naming. After a study of the problem, a solution for ethernet devices was made, which is available online at HDI sourceforge.net page.

Linux and CPU benchmarking.

The purpose of the project was to create a Linux benchmarking suite designed for telecom purposes. Main focus was on benchmarking CPU, memory and operating systems. To get a more accurate results, the benchmark used so called macro benchmark approach i.e. created an application that simulates some typical telecom environment and benchmarked how long it takes to perform a set of operations. The idea of the benchmark was to create many threads that would compete from the same resources. Other aspect was a high degree of context switches, since often the systems receive a large number of short messages and the time to process a single message is short. Therefore the system does many context swithces since the processing of messages is divided evenly to the threads.

2003:

Carrier Grade Linux specification work.

The project started to study the different open source clustering file systems for Linux, but the requirement was dropped from CGL and study was abandoned. Instead scalability problems with thousands of connections and the solutions were qualitatively studied in the remaining short time.

Linux and DSP interoperability study.

The purpose is to study what kind of mechanisms can be used when a real-time operating system (RTOS) and Linux need to communicate with each other. In the future it is possible that Linux will be used in the third generation mobile phone networks in control plane. In order to be usable in there it is vital that there exists an efficient mechanism for a Linux system to communicate with RTOS systems in the user plane. The requirements for this communication system come from both the Linux side i.e. small computing overhead since there may be tens of thousands of connections to the Linux system and also from the RTOS side i.e. the memory usage should be small preferably in the order of 100 kB. After a more general availability study the project tries to create an SCTP implementation for the RTOS environment. The approach is to port an existing SCTP implementation to the RTOS environment.

2005:

RTAI/Fusion and Adeos porting to PPC64.

This project will port the Adeos nanokernel and RTAI/Fusion real-time enhancements to PPC64 architecture. They are Linux kernel modifications that already have been ported to 32 bit PPC architectures. After porting their relevant characteristics will be studied shortly. Some patches are available.