The inaugural meeting of The Open Group Real-time and Embedded Systems Forum was attended by 70 individuals from 50 different organizations.
It was recognized that this forum can be key in accelerating the proliferation of embedded systems through development of test and certification programs.
Key speakers from the Real-time industry attended the meeting outlining current trends in the industry and providing input to the Forum workplan. These included representatives from the following organizations: Boeing (Richard Paine), Wind River (John Fogelin), QNX (Steve Furr) , FSMLabs (Victor Yodaiken), LynuxWorks (Mitch Bunnell), Venturcom (Myron Zimmerman) , NCITS-R1 (Russ Richards), US Army (Bill Protzman), Mitre Corporation (Doug Jensen), OS JTF (Lt Col Glenn Logan, Curtis Royster,Jr.), The Aerospace Corp (Sam Bowser), TimeSys (Mark Gerhardt), UDI Project/Interphase (Kevin Quick) , Mitre/OMG (Dave Emery) and The Open Group (Dave Lounsbury).
The concept of the Real-time and Embedded System Forum as a meeting place for coordination between suppliers, practitioners and existing industry standard groups has been firmly established at this meeting. Industry groups identified to work with, and represented at the inaugural meeting included IEEE PASC, The Society of Automotive Engineers, The Object Management Group, NCITS R1 and the UDI Project.
Prior to the meeting, and included with the meeting materials was a prospectus for the Forum. During the meeting, sub-groups were formed to validate the proposals in the prospectus and positions put forward by the champions and to propose further action by The Open Group Real-time and Embedded Systems Forum.
The break-out groups have validated that the following key areas are clearly problems where work needs to be done:
The next steps:
The forum will be officially launched on July 31.
The next steps will be the formation of a steering committee and workgroups for each activity. There needs to be agreement on specific working groups to address key topics .
The Open Group will analyze the output from the meeting and working with the Founder members of the Real-time and Embedded Systems Forum will propose a forward work plan. Statements of work for each area will be drafted and subgroups should meet either face to face, or by teleconference prior to the next plenary.
The next plenary meeting is in Washington during w/c October 23rd. This will take the form of a one-day open plenary, and up to one and a half days closed working group meetings.
Allen welcomed the attendees to the day. Andrew introduced the sessions and the objectives for the Forum. He explained that the mission of the forum was to grow the market for real-time and embedded systems through deployment of standardized product, and that the Forum was going to work with existing industry groups as a coordinator and integrator. Groups identified to work with, and represented at the meeting included IEEE PASC, The Society of Automotive Engineers, The Object Management Group, NCITS R1 and the UDI Project.
Andrew explained that the objectives of the day are to focus on a work program and initial schedule.
The membership of the forum is targeted at all of the people involved in the definition and implementation of real-time system.
Andrew mentioned that The Open Group would be working to sign agreements with more founder members over the next 30 days (the cutoff is July 31) with announcements being made when the initial work program is decided with the founders.
The initial speakers for the day will be presenting on current trends in the industry and give their perspectives on the work required. There will be some work groups at the end of the day to look at some plans for the work items.
John introduced his presentation as a rally call to the efforts that are being undertaken in the embedded space. In relating the present day status to that of the development of the electric motor, John foresees that computers will become embedded in everything and we will no longer think about using a computer.
His quotes show'd how far the industry has come in the last 30-50 years. Particularly emphasising that the Internet is the world's largest embedded system.
John forecast that the space for embedded systems is set to grow exponentially in the next few years, due to the increasing power and lower cost of the computers.
Reliability will be a key driver of embedded systems. Airline control systems, car control systems, etc. Must be failsafe.
The challenges faced by the group are in the increasing embedded content, and increasing diversity of the technology involved in embedded systems. Multiprocessor and Heterogeneous advancements also add to the complexity and the programming dilemma.
John asserted to the group that there is no Embedded platform for this environment.There is a kit of technology to help build platforms. Any attempt to restrict this flexibility could be counter productive to the development of embedded systems.
However, John sees a convergence of the Desktop and Embedded Devices into one of a Smart device environment that is purpose built, with multiple platforms. There will be a strong built in redundancy in embedded systems as technology continues to move forward rapidly.
John then addressed the Linux and it's applicability to embedded systems. The ability to use the source code in developing the embedded platform has been appreciated by the community. The issue of the source being free is of secondary importance. The availability of applications is the highest value appreciation of Linux. However, John did comment that the divergence of Linux solutions is already causing problems as the industry moves forward.
In conclusion, there are many software challenges ahead, and the community must be prepared to address the diversity and complexity of systems. John commented that there are many social challenges ahead
Victor started by giving his definition of Real-time. He sees three definitions: Hard (times must be met), Soft (times should be met), and hand writing (which is usually interrupt driven). He show'd some tough real-time examples - particularly 1Gb ethernet which has to send a packet every 1.5 microseconds.
RT Linux follows the POSIX PSE51 standard. It is a minimal real-time system that looks like a single thread. Linux runs as a sub-thread of the Real-time system.
Victor then proceeded to describe the architecture and design principles through a series of slides. RT Linux does not try to do anything with GUI's - let the standard Linux system handle X-Window, and equally TCP/IP is not real-time.
RT Linux extends the object model, uses the POSIX-threads "lite", follows PSE54 in the main but not exclusively.
Victor sees RT Linux being used in many devices - particularly in data acquisition devices.
The basic model of POSIX real-time fits the RT Linux development well. Extensions to SMP are also easily handled. Linux was needed to be extended to allow floating point to be turned on or off. Added a facility to run a specific thread on a particular cpu. Periodic execution has also be added. They are looking toward the development of killing threads, joining threads
Some things left out of the POSIX definition as the CONFIGURATION options, priority inheritance, no scheduling parameters. Signals not implemented efficiently.
QNX is the leader in Real-Time systems based on POSIX. Steve is the QNX representative to the Embedded Linux Forum.
Steve's premise is that Linux will emerge as the market share leader for embedded markets, because it is freely available and supported by may platforms. However, Steve's presentation is designed to address some of the questions posed by his premise. These are: what is Linux? Will dejure or defacto standards prevail? will development recognise the role of standards organisations.
Steve feels that the forking of Linux has been overstated, he sees the Linux community trying things over the years, and suitably merging at defined points in time. He used the example of a flock of birds and how they moved in general harmony together with some differences in the particular birds.
Steve moved on to Real-time and commented that the standard Linux kernel is not POSIX compatible. Other areas that need to be considered are Scalability of Real-Time solutions and the Reliability of the system.
Steve moved on to EL/IX (Embeded Linux/Posix API). This is s 1003.13 implementation with networking and BSD and GNU extension. EL/IX does require kernel modifications, and Steve indicated that none currently exists.
The Linux community, referred to as the Linux Community Process by Steve, indicated that this was an ad-hoc project oriented group. The implementation is effectively the specification. But the market is very fluid in the development of new technology. The acceptance criteria for this Linux community is that the technology is available and that there are no IP incumberences. there is a very low barrier to entry. the community is very networked and rarely meets to agree specifications, etc.
Steve did comment that the Ad-hoc process does not necessarily yield faster results. Open participation should not impact focused efforts. usually the people with most time to contribute do the majority of the work. There is no real voting, because success of a piece of technology is that others will use it.
Finally, Steve suggested that the role of the Forum should be to:
- Promote adherence to standards
- Develop profiles for POSIX, GNU, etc.
- Foster community involvement
- Fast track community developed specifications.
Mitch is the CTO at LynuxWorks.
His view of the market and standards Real-time. Mitch worked in the ".1b" standards. He provided some insight into the development of Real-time POSIX over the years.
What criteria should someone use when deciding to use a Real-time system. Typically most decisions are a compromise of the available technology.
The important problems that Mitch sees in the using embedded systems are:
- The 3rd Party vendors cannot make money in the area, because systems are too divergent.
- The source code being free is irrelevant
RT Linux is showing a resurgence of the UNIX market because of its popularity. It is most stable in i86 platforms. There is a possibility of having a standard binary implementation for the system. This should include - POSIX.1, some of POSIX.1b, Single Unix Specification sockets, BSD and System V calls as well as X Window support. he would like to see a test suite to check for this ABI. It will be good for the development of embedded Linux
This should promote the development of the 3rd Party software market for Linux.
Myron initially showed a demonstration of the latency for the real-time system.
His agenda for this morning was to talk about the VenturCom implementation of a Real-time software in their company. In particular their work on real-time has been based around the Win32 API.
Myron reviewed the requirements of new embedded systems that are being developed before then discussing those requirements against the Windows platforms that are available. Windows CE requires more development effort, but is cheaper to deploy. Microsoft is targeting Windows NT Embedded to the high end of the market.
Myron then described the capability and features of Windows CE, and NT Embedded for specific devices. The latter offers a mix and match of components, and a tool for creating the image on a ROM.
In talking about the real-time extensions to Windows Myron presented the features and architecture of the Windows Real-time extensions. The details of these features are provided in the slide presentation.
Myron then described the VenturCom DCX product to allow Real-time applications to be developed for Windows CE and NT Embedded systems.
The theme of Russ's presentation is about the Good, the Sad and The Ugly. POSIX is doing well, but the Java Specification wars are the real Ugly aspect of the current.
DISA believes that they are not the only ones that has requirements of standards, and they are trying to work with the private sector community to identify common areas of standards work. The government recognises the need to participate in standards fora.
Russ provided some of the background and politics to the Java Specification for Real-time. Today there still exists two initiatives in the area of Java Standardisation - J Consortium and the Java Community Process from SUN.
Russ indicated that 4 projects related to real-time have been started by the DoD to improve standardisation. Russ can provide details to anyone who wants to look at background pages related to the successful and unsuccessful projects as detailed in his slides.
Interoperability is still a major issue for the DoD. Rugged development and other specifically DoD developments are no longer the focus. Now the DoD is trying to move toward the market and the latest innovations.
Russ continued to list standards being endorsed by the government, and the processes to improve the current project implementations. They are willing to work with the industry to help fix standards when they are seen as not being sufficient.
Finally Russ called the Forum to action in the some of the areas of Real-time interest within the DoD. The DoD has 500 people working in the area of standards (down from 2000 a few years ago). Collaborative working with groups like the Real-time Forum will help the DoD leverage its investments.
Russ presented a picture of the proposed modifications to the DIICOE that the Open Group could help facilitate through the work of the Real-time Forum. He encouraged the group to work in the forum.
Richard talked about the real-time requirement within Mobility, his main area of expertise at Boeing. Boeing are looking to deploy radio networks within organisations, move staff to Voice over IP and in airplanes, introduce broadband communications from the satellite to support mobile networks.
Focusing on the Real-time requirement for wireless, Russ describe the 4 areas of network that Boeing have defined and want to achieve:
- Personal Wireless
- Campus - High Bandwidth
- Regional Low Bandwidth
- National/International Low Bandwidth
Richard continued to define the elements of the Boeing architecture. The real-time requirements they see are Service Profile, Policy Management,Class of Service, Address Conversion, QoS Management, E911 Support and Security. These are the area where response has to be predictable so that devices can operate efficiently.
Richard highlighted that QoS is the highest requirement of real-time behaviour for Wireless Devices and LANs
Bill introduced the WSTAWG - An Army working group related to the development of standards across weapons systems. The group has been formed for a couple of years - trying to define a set of extensions to the JTA that would be specific to the Weapons Systems.
It's mission, detailed on the slide, is to define the minimum set of additional features necessary. Re-use and compatibility are high on the priority list.
The architecture defines many interfaces between pieces of the system, such that particular components of the system could be replaced without having to replace the whole system.
WSTAWG has seen that many systems being developed with common components, but slightly different semantics. Real-time CORBA is an example of a useful technology, but one that is not yet mature. No vendors have bought into these standards. The WSTAWG have decided not to endorse specific vendors. The slides detail the goals of the group, and the service areas that the group focus on for the weapons system. Details of the plans and the prototypes developed should the different technologies and API's that are implemented to be used in the Operating Environment.
Bill listed the specific API that are required. The Design was presented in UML. The OE is written in CORBA IDL. Bindings for Ada and C languages are defined. The project is nearing completion of the first version, and is now looking to see what the lessons learned have been. There is a need to update where the various technology have reached in the last year or two - particularly a re-sync with POSIX, and CORBA.
In summary, Bill requested the forum to address some of the issues of standards development. There own desires are to be able to validate there requirements again standards and emerging technology, mature technologies such as RT CORBA, and then identify new RT technologies that can be incorporated into the OE.
Kevin gave an overview of the Uniform Driver Interface (UDI). UDI was created by Project UDI -- an open industry initiative -- to allow for 100% portability of device driver source across a wide range of Operating Systems and platforms. UDI isolates drivers from OS policies and platform details, allowing a driver to be moved to a new OS with at most a recompile.Kevin reported that the UDI project has worked with NCITS to look at the Real-time impact on UDI. As a result an attribute field had been added to the information held within the driver that could be used to counteract priority inversion. Project UDI is very interested to hear in other changes that might be needed to better support real-time systems. It was later noted in the panel session that Project UDI is considering standardization through NCITS once they consider the work stable and mature, at the moment there is much concentration on proof of concept by implementation.
Questions during the session ranged from:
What are the plans for standardization of UDI? and although UDI does not block, is it predictable in its response?
Which is the better implementation model for real-time Linux, is it better to add full real-time capability to Linux, or is it better to run Linux as a thread under a real-time kernel? This generated some discussions.
Another question was on why the vendor support for the POSIX 1003.13 profiles was not forthcoming? There was a lot of discussion about the merits of the various POSIX 1003.13 profiles, with a question from the audience on why no-one had implementation profile 52? The response from the real-time operating system vendors was that most of them implemented the complete profile 54 and then let the user decide which if any options to deselect. The downside of this was that applications may not be portable as you cannot guarantee that the option selections will all be the same.
A question was asked whether profile testing of the operating systems would be useful, it was felt that it would be, and perhaps this would lead to the availability of systems that do support the profiles. It was also recognized that application side tools need also to be available so that developers can know when they are using facilities outside of a profile.
Col. Glen Logan OSJTF
Colonel Logan presented a set of slides outlining the OSJTF supporti for the forum.
Dave Emery Mitre (a view from the OMG)
Dave presented a set of slides on behalf of Dock Allen, who was unable to attend the meeting. He outlined areas where he recommended the group not work, and areas where it should.
E. Doug Jensen - Real-time Java
Sam Bowser - Security
Mark Gerhardt - New APIs
Dave Lounsbury - Quality of Service
At the end of the Afternoon the 6 Panel leaders reported back to the Group:
Dave Emery - Performance Testing
Dave reported that their discussions were very wide ranging and mainly related to component performance testing. He presented some slides of the points covered. In conclusion, though, Dave indicated that it was not clear to the group if it would make any difference to undertake performance testing.
E. Doug Jensen - Real-time Java
Doug Reported that the group was very concerned about the political situation relating to Java, and that their unanimous recommendation was that work by a third group between SUN and the J-Consortium was not recommended.
Dave Lounsbury - Quality of Service
Dave reported that all of the group discussion revolved around single system QoS. There was definitely a need for more discussion on and exposure to QoS technologies such as HiperD. Dave commented that there was no immediate need for any specifications work in this area.
Curtis Royster - Conformance Testing
Curtis reported that there was a need for continued POSIX testing. There were many areas of testing that the group felt could make a significant impact on the acceptance of Real-time. His flip-charts listed many areas to be considered. They also discussed looking into the policy of 3rd party testing, self certification issues, and the overlapping of services in different areas - to maximise potential.
Sam Bowser - Security
Sam apologised that his group did not actually meet during the break-out due to other commitments. However, Sam did report that he had held informal discussions and that there was need for security and the development of a business case for security. Security is seen as an overhead, and there is continual push back on the need and cost of implementing good security. The group wanted more initiatives to generalize security.
Mark Gerhardt - New APIs
The forum will be officially launched on July 31.
The next steps will be the formation of a steering committee and workgroups for each activity. There needs to be agreement on specific working groups to address key topics The structure of the Forum needs to allow each group to "meet" at least once per month, either face-to-face or telephonically.
The Open Group will analyze the output from the meeting and working with the Founder members of the Real-time and Embedded Systems Forum will propose a forward work plan. Statements of work for each area will be drafted and subgroups should meet either face to face, or by teleconference prior to the next plenary.
The next plenary meeting is in Washington during w/c October 23rd.
©
1995-2000
Last Modified: Wed Jul 5 08:42:39 BST 2000
