We consider the problem of OS resource management for real-time and multimedia systems where multiple activities with different timing constraints must be scheduled concurrently. Time on a particular resource is shared among its users and must be globally managed in real-time and multimedia systems. A resource kernel is meant for use in such

Traditional real-time systems have largely avoided the use of disks due to their relative slow speeds and their unpredictability. However, many real-time upplications including multimedia systems and real-time database ap. plications benefit significantly from the use of disks to store and access real-time data. In this paper, we investigate the problem of obtaining guaranreed timely access to files on a disk in a real-time system. Our study focuses on several aspects of this problem of providing a real-time filesystem.First, we consider the use of two real-time disk scheduling algorithms: earliest deadline scheduling and just-in-time scheduling, a variation of aperiodic servers for ?he disk. The latter algorithm is designed to improve disk throughput that can be hurt when a real-time scheduling algorithm such as EDF is applied directly. Admission control policies with practically acceptable properties of performance and usabilio are provided. Next, we design and implement a real-time jilesystem on the RT-Mach microkernel-based system running a real-time shell. The new inteqace we develop is based on RT-Mach 's resource reservation paradigm and provides guaranteed and timelv access for multiple concurrent applications requiring disk bandwidth with direrent timing and volume requirements. Finally, we perform a detailed per$ormance evaluation of the real-time filesystem including its raw petjormance. We show the following positive but rather surprising result: our realrime scheduling filesystem not only provides guaranreed and timely access but also does so at relatively high levels of throughput. Traditional disk scheduling algorithms ofer completely unacceptable file access latencies for real-time applications and do so only at slightly higher throughput.

We Motivation for Resource KernelsExample real-time systems include aircraft fighters such as F-22 and the Joint Strike fighter [19], beverage bottling plants, autonomous vehicles, live monitoring systems, etc. These systems are typically built using timeline based approaches, production/consumption rates [9] or priority-based schemes, where the resource demands are mapped to specific time slots or priority levels, often in ad hoc fashion. This mapping of resources to currently available scheduling mechanisms introduces many problems. Assumptions go undocumented, and violations go undetected with the end result that the system can become fragile and fail in unexpected ways. We argue for a resource-centric approach where the scheduling policies are completed subsumed by the kernel, and applications need only specify their resource and timing requirements. The kernel will then make internal scheduling decisions such that these requirements are guaranteed to be satisfied.Various timing constraints also arise in desktop and networked multimedia systems. Multi-party video conferencing, mute but live news windows, recording of live video/audio feeds, playback of local audio/video streams to remote participants etc. can go on concurrently with normal computing activities such as compilation, editing and browsing. A range of implicit timeliness constraints need to be satisfied in this scenario. For example, audio has stringent jitter requirements, and video has high bandwidth requirements [8]. Disk accesses for compilation should take lower precedence over disk accesses for recording a live telecast.Two points argue in favor of resource-centric kernels we call "resource kernels": • Firstly, operating system kernels (including microkernels) are intended to manage resources such that application programs can assume in practice that system resources are made available to them as they need them. In real-time systems, system resources such as the disk, the network, communication buffers, the protocol stack and most obviously the processor are shared. If one applica-

Pubfic reporting burden {or this collection of information is AGENCY USE mil (Leave blank)2. REPORT DATE SepOl REPORT TYPE AND DATES COVEREDFinal Mar 96 -Mar 99 TITLE AND SUBTITLE END-TO-END RESERVATION SERVICES IN REAL-TIME MACH AUTHOR(S)Raj Rajkumar, K. Juvva, A. Molano, S. Oikawa, C. Lee, J.P. Lehoczky, D.P. Siewiorek, C.W. Mercer and K. Yoshida FUNDING NUMBERSC -F30602-96-1-0160 PE -62301E PR -D649 TA -00 WU -01 PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)Carnegie Mellon University Department of Computer Science 5000 Forbes Avenue Pittsburg. PA 15213 Approved for public release; distribution unlimited. PERFORMING ORGANIZATION REPORT NUMBER SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES)Defense 12b. DISTRIBUTION CODE ABSTRACT IMaximum 200 words)The RT-Mach microkernel supports a processor reserve abstraction which permits threads to specify their CPU resource requirements. If admitted by the kernel, it guarantees that the requested CPU demand is available to the requestor. This kernel-supported mechanism is designed to be relatively simple based on the microkernel notion that user-level policies can use this simple mechanism to build more complex and powerful schemes. In this report, the focus is on the needs of such user-level policies in the form of a dynamic Quality of Service (QoS) server. Three goals are sought: 1) explore the necessity, sufficiency, power and flexibility of the kernel-supported reserve mechanism, 2) dynamic management of application quality in real-time and multimedia applications, and 3) investigate our ability to predict and achieve end-to-end application delays in realistic distributed real-time and multimedia applications. A two-pronged approach to accomplish these goals is used. First, the processor reserve abstraction in a user-level dynamic quality of service server is applied. A QoS server can allow applications to dynamically adapt in real-time based on system load, user input or application requirements. Second, the dynamic QoS control capabilities to a distributed multimedia application whose threads have to interact and coordinate with each other within and across processor boundaries are applied. A new notion called continuous thread of control is introduced to assist in bundling processor reserves. The experiments show that we can indeed predict and achieve end-to-end delays in a distributed multimedia application. Tables for Thesis 69 LIST OF AUTHOR'S AND PAPER'S SUBJECT TERMS1 AbstractWe consider the problem of OS resource management for real-time and multimedia systems where multiple activities with different timing constraints must be scheduled concurrently. Time on a particular resource is shared among its users and must be globally managed in real-time and multimedia systems. A resource kernel is meant for use in such systems and is defined to be one which provides timely, guaranteed and protected access to system resources. The resource kernel allows applications to specify only their resource demands leaving the kernel to satisfy those demands using hi...

In this paper we focus on two practical considerations that arise in the design of a real-time filesystem. Firstly, disk bandwidth management should be dynamic, which in turn would allow a QoS manager to dynamically reallocate disk bandwidth to running applications based on their changing needs. Secondly, real-time access to file system data structures should be deterministic, in order to avoid unexpected latencies when accessing files from disk. These issues have implications to the design of the filesystem and to its schedulability analysis. We address both these problems and present an implementation in RTFS (Real-Time Filesystem Server), a real-time filesystem supporting disk bandwidth reservation running on top of the Real-Time Mach microkernel. Finally, quantitative comparisons of actual achieved filesystem bandwidth and response times are used to validate our approach.