Joe Bergmann, The Open Group
The Real-time and Embedded Systems Forum met from Wednesday January 23th through to Thursday January 24th co-located with The Open Group quarterly conference in Anaheim.
The Real-time and Embedded Systems Forum continues to gather momentum. Attendance at this meeting was 108, making this one of the largest, and certainly the fastest growing group within The Open Group. Sessions held at Anaheim included the second annual Real-time Linux interest group. Represenatives from OMG, SAE, DoD RTAG, IEEE POSIX SSWG-RT and NCITS R-1 gave status reports. The sessions on Safety Critical, Hard Real-time Java, Security for Limited Resource Environments, POSIX Real-time Profiles and Real-time Access to Data were an overwhelming success. The Real-time and Embedded Forum concluded with a BOF on "Open Systems and Military Applications.
The second annual meeting of the Linux Interest Group included a series of presentations ranging from implementing Real-time on native Linux, through real-time applications on Linux, through to implementation issues such as high resolution timers, resource allocation and POSIX threads . The Profiles group looked at the proposed evolution for the POSIX.13 profiles and a new tracing capability in detail.
The future meeting schedule for the forum is as follows:
Joe Bergmann, The Open Group
Evolution of Real- time Standards in NCITS TC R1, Real-time Computing
Systems - Russ Richards Chief, Technical Architecture Branch DISA
Interoperability Directorate
RTSIG - Dock Allen, The MITRE Corporation
Real- time Advisory Group (RTAG) - John Maurer
Timeliness
meets QoS -What should we do? Dock Allen The MITRE Corporation
Real-Time Capability for Native Linux - Kevin Morgan, MontaVista Software
Applications That Require a Real-time Linux Variant - Steve Brosky,
Concurrent Computer Corp
High Resolution POSIX Timers for Embedded Linux - John Mehaffey, MontaVista
Software
Application Design Impact of CPU and Network Reservations in Linux - Doug
Locke, TimeSys Corporation
Moving Linux to the Enterprise -- Next Generation POSIX threads, Bill
Abt et al, IBM
Real-time and Embedded Profiles
Michael Gonzalez a presentation on the current status
of the POSIX profiling work.
Real-time and Embedded Profiles Working Group session - POSIX Tracing -- Francois Riche, IBM
TETware professional Version 1.4 Realtime and Embedded Systems Extension
- Alan Haffenden, Product Manager
Software Development under DO- 178B John Joseph Chilenski, Associate
Technical Fellow, Airborne Software Engineering, Boeing Commercial Airplanes
Real Time and Quality of Service Dave Emery, Co- Chair, RT/ Embedded
Forum
Military Transition - Understanding the environment & differences
Leslie A. Johnson, The Boeing Company
Supporting Secure,Distributed, Real- time Systems:A Middleware Perspective
Bill Beckwith
Computing Security for RT and Embedded Systems - Kernel module approach
Mitch Bunnell, CTO LynuxWorks
The Common Criteria for Information Technology Security Evaluation (CC)
Michael McEvilley
RTLinux and process control security Victor Yodaiken, Finite State
Machine Labs Inc. (FSMLabs)
See QoS Task Force
Hotspot: experiences and lessons - Cliff Click, Senior Staff Engineer,
Sun Microsystems, Inc.
Requirements For Real- time Extensions For the Java™ Platform Doug
Jensen, Mitre
Real-Time Java for Embedded Systems (RTJES) 1Lt Jason Lawson, AFRL/IFTA,
Information Directorate, Air Force Research Laboratory
Thoughts Regarding Sun's Real- Time Specification for Java - Kelvin
Nilsen, J Consortium
Real- time Data Access Requirements Charlotte Wales, RTAG RT DAS Subgroup
Co- chair
USAF Avionics Upgrade Process
Real-Time and Embedded Systems Issues for Military Systems Acquisition
Real-Time
Aerospace Application Programming Interface Support (RAAPIS) Camilo
Segura, Minerva Rodriguez
Real- Time Specification for Java™
MOST
The first session on the Wednesday included a series of briefings from liasions to the forum.
P1003.5g (ADA Bindings)
A complete, merged document is being generated with the following changes:
Most of the 1003.1j and 1003.1d changes to 1003.5c are acomplete. There is a mapping of 1003.1q to an ADA Binding. The associated text is still needed for the document. The balloting committee needs to be formed, it is planned to begin the first ballot before April.
P1003.13a (Profiles)
For the POSIX profiles revision, it has been reported that a first cut of "Embedded" Profiles is complete. It has been decided to use the existing profiles (PSE51, PSE52, PSE53, PSE54) and remove unnecessary functionality. A current dependency being waited upon is the P1003.13b updates. There have been email discussions to merge the P1003.13a and P1003.13b PARs.
GOA Status
Preferred Standards for Avionics Domain is approximately 75% complete. We are working with the R1 group to complete the Resource Access or Driver level. We have a document outline for the Guidance Document for the Avionics Domain and now we need to populate it with text. The GOA Standard is up for reaffirmation. This will be complete next week.
The SAE AS-5 Committee is meeting next week in El Segundo. GOA and AADL are meeting during this week. OS API is completing the document updates and getting ready for ballot. If you wish to be on the balloting committee, contact Diane Schleicher. (You must be an IEEE member)
NCITS TC R1
Russ Richards presented status. He showed the projects approved and underway. He then discussed project UDI and the need for real-time device drivers and protocol modules. He then scoped it and presented the R1 draft standard. He then explained whom ought to be involved and invited delegates to join the cause.
RTAG
John Maurer presented status. He started with a toaster chart of DII COE taxonomy and went on to explain the RTAG vision. He explained that RTAG was formerly RT TWG and went through the participation, key efforts to date, charter and subgroup. He concluded by giving contact details.
Real-time capability for native Linux -- Kevin Morgan, MontaVista Software
Kevin began by presenting updates from MontaVista's view. He went on to say what they saw Linux as and the scope for Linux customisation.
He tightened things up by giving a definition of terms. He explained that primary Linux issues included responsiveness and shared MontaVista's general approaches in this regard. He went through specific MontaVista efforts to date and reported on half some dozen or so points within their technology status.
He zeroed in on latency measurements and performance testing before talking about interrupt latency for 5 or 6 different chips. He gave an insight into pre-emption latency. He shared benchmarks from a well-known Japanese customer outlining the measurement methods used, the benchmark results and the response curves.
He concluded with Linux's key filter is all about performance and understandability.
Applications for real-time Linux -- Steve Brosky, Concurent Computer Corporation
Steve began with there being two approaches to RT Linux, RT/Linux or RTAI and real-time version of Linux. He gave information on both approaches saying both approaches have merit but some applications are better suited to one or the other real- time Linux approach. To support the broadest number of applications, both approaches should be utilized.
Next he presented applications requiring a real-time variant of Linux. He suggested three examples:
He concluded by talking about RT forks and RT Linux variants.
High Resolution POSIX timers for Linux John Mehaffey, MontaVista
John started off by explaining what high-resolution timers are and why they are needed. He dipped into x86 hardware time sources then giving some comparisons. He detailed current timer algorithms and high-resolution timer algorithms.
He explained POSIX clock types and spent some time on their programming. He concluded with the current issues, this is only currently on the Intel x86 architecture, it is beta code and has not been accepted into the standard kernel.
Using Reservation Technology in Linux -- Doug Locke
Doug started with embedded applications and looked at priorities and ways of improving on priorities. He then explained non-priority performance analysis along with CPU reservations. He went through a pictorial of temporal reservation parameter and then gave an example of CPU reservation. He explained ways to use CPU reservation as well as network reservations. At this point he looked at standard Linux networking architecture with network reservation architecture. Before summarising he presented a potential application.
Bill started with the project overview and explained the key design points. He
showed the current status, looked at compliance testing and touched on performance
and compatibility. Next he delved into thread related kernel issues and asked what
was the state being talked about.
He went through NGPT kernel patches and gave a case history within the telecom
industry. He felt that NGPT is a better solution for threading on the Linux platform.
He looked at NGPT integration into Intel Telecom Linux technology and scanned the
success criteria for NGPT. He reviewed the new features of NGPT 2.0, considered
GDB support and IBM exploitation of NGPT 2.0
Q
Have you discussed the design with developers?
A
We are in discussion with kernel developers right now.
Q
What about embedded Linux consortium?
A
We have not opened a dialogue with them but we want to.
Q
Can you use both threading models in different processors that use both models?
A
Yes. But they would not be able to share semaphores. One must keep the firewall up
between them.
A proposal for POSIX .13 revision
Michael Gonzalez, Universidad de Cantabria
Francois Riche, IBM
Michael began with a table of contents of his presentation. He introduced RT systems
by considering the elements of a RT system. He looked at RTOSs and then explained
what POSIX was. He considered the motivations of RT POSIX and looked at RT
extensions, threads extensions and additional RT extensions before introducing
tracing.
At this point Francois Riche of IBM took over to look at tracing in more detail.
Francois explained what tracing was and plunged into POSIX tracing API. He
followed this with tracing API inputs and the state of the art of tracing. He stressed
the benefits of a tracing standard before going through three tracing models, one of
which was without a disk. He considered tracing roles and then introduced POSIX
tracing API features and tracing functionality grouping. He ended with some
interpretations are already available with there also being one ADA implementation
existing.
Michael came back now where he gave a history of some of the POSIX standards and
talked about the motivation of POSIX .13.
Moving onto the second part he gave an overview of the current profiles and
displayed a good pictorial of profile characteristics. He concluded this part by
considering the structure of a profile.
In part three he went through the new POSIX.13 requirements.
In part four he started with future work and went through the minimal options
required in all profiles. Next he went through the controller and multi file system-
related options after which he looked at the dedicated and multi multiple process-
related options. He then looked at options required only in the multi-purpose RTS
before looking at miscellaneous options. He concluded with options not required in
any profile.
He felt that the group had made good progress and that a new profile should be
situated somewhere between the dedicated and multi. There was need to reach a
small number of profiles.