E. Douglas Jensen | Virginia Tech (original) (raw)

Papers by E. Douglas Jensen

Public reporting burden for the collection of information is estimated to average 1 hour per resp... more Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number.

Springer eBooks, 2008

No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form ... more No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

This paper compares several parametric and adaptive failure detection schemes in terms of their r... more This paper compares several parametric and adaptive failure detection schemes in terms of their respective QoS. We introduce an improvement over existing methods, and evaluate their benefits. First, we propose an optimization to enhance the adaptation of Chen&amp;amp;#x27;s FD, which significantly improves QoS, especially in the aggressive range and when the network is unstable. Second, we address the problem of

The Distributed Real-Time Specification for Java (DRTSJ) is under development within Sun's Java C... more The Distributed Real-Time Specification for Java (DRTSJ) is under development within Sun's Java Community Process (JCP) as Java Specification Request 50 (JSR-50), lead by the MITRE Corporation. We present the engineering considerations and design decisions settled by the Expert Group, the current and proposed form of the Reference Implementation, and a summary of open issues. In particular, we present an approach to integrating the distributable threads programming model with the Real-Time Specification for Java and discuss the ramifications for composing distributed, real-time systems in Java. The Expert Group plans to release an initial Early Draft Review (EDR) for previewing the distributable threads abstraction in the coming months, which we describe in detail. Along with that EDR, we will make available a demonstration application from Virginia Tech, and a DRTSJ-compatible RTSJ VM from Apogee.

We present an optimal real-time scheduling algorithm for multiprocessors-one that satisfies all t... more We present an optimal real-time scheduling algorithm for multiprocessors-one that satisfies all task deadlines, when the total utilization demand does not exceed the utilization capacity of the processors. The algorithm called LLREF, is designed based on a novel abstraction for reasoning about task execution behavior on multiprocessors: the Time and Local Execution Time Domain Plane (or T-L plane). LLREF is based on the fluid scheduling model and the fairness notion, and uses the T-L plane to describe fluid schedules without using time quanta, unlike the optimal Pfair algorithm (which uses time quanta). We show that scheduling for multiprocessors can be viewed as repeatedly occurring T-L planes, and feasibly scheduling on a single T-L plane results in the optimal schedule. We analytically establish the optimality of LLREF. Further, we establish that the algorithm has bounded overhead, and this bound is independent of time quanta (unlike Pfair). Our simulation results validate our analysis on the algorithm overhead.

International Conference on Distributed Computing Systems, 1982

We argue that the key underpinning of the current stateof-the real-time practice-the priority art... more We argue that the key underpinning of the current stateof-the real-time practice-the priority artifact-and that of the current state-of-the real-time art-deadline-based timeliness optimality-are entirely inadequate for specifying timeliness objectives, for reasoning about timeliness behavior, and for performing resource management that can dependably satisfy timeliness objectives in many dynamic real-time systems. We argue that time/utility functions and the utility accrual scheduling paradigm provide a more generalized, adaptive, and flexible approach. Recent research in the utility accrual paradigm have significantly advanced the state-of-the-art of that paradigm. We survey these advances.

Heuristic algorithms have enjoyed increasing interests and success in the context of Utility Accr... more Heuristic algorithms have enjoyed increasing interests and success in the context of Utility Accrual (UA) scheduling. However, few analytical results, such as bounds on task-level and system-level accrued utilities are known. In this paper, we propose the S-UA algorithm that can provide probabilistic bounds on tasklevel accrued utilities. Lower bound on system-level accrued utility ratio (AUR) is also derived and maximized by S-UA.

Journal of Parallel and Distributed Computing, Mar 1, 2010

We consider optimal real-time scheduling of periodic tasks on multiprocessors-i.e., satisfying al... more We consider optimal real-time scheduling of periodic tasks on multiprocessors-i.e., satisfying all task deadlines, when the total utilization demand does not exceed the utilization capacity of the processors. We introduce a novel abstraction for reasoning about task execution behavior on multiprocessors, called T-L plane and present T-L plane-based real-time scheduling algorithms. We show that scheduling for multiprocessors can be viewed as scheduling on repeatedly occurring T-L planes, and feasibly scheduling on a single T-L plane results in an optimal schedule. Within a single T-L plane, we analytically show a sufficient condition to provide a feasible schedule. Based on these, we provide two examples of T-L plane-based real-time scheduling algorithms, including non-work-conserving and work-conserving approaches. Further, we establish that the algorithms have bounded overhead. Our simulation results validate our analysis of the algorithm overhead. In addition, we experimentally show that our approaches have a reduced number of task migrations among processors, compared to a previous algorithm.

Springer eBooks, Aug 23, 2007

ABSTRACT In a decentralized network system, an authenticated node is referred to as a Byzantine n... more ABSTRACT In a decentralized network system, an authenticated node is referred to as a Byzantine node, if it is fully controlled by a traitor or an adversary, and can perform destructive behavior to disrupt the system. Typically, Byzantine nodes together or individually attack point-to-point information propagation by denying or faking messages. In this paper, we assume that Byzantine nodes can protect themselves from being identified by authentication mechanisms. We present an authentication-free, gossip-based application-level propagation mechanism called LASIRC, in which "healthy" nodes utilize Byzantine features to defend against Byzantine attacks. We show that LASIRC is robust against message-denying and message-faking attacks. Our experimental studies verify LASIRC's effectiveness.

Journal of Parallel and Distributed Computing, Feb 1, 2010

We present the first Utility Accrual (or UA) real-time scheduling algorithm for multiprocessors, ... more We present the first Utility Accrual (or UA) real-time scheduling algorithm for multiprocessors, called global Multiprocessor Utility Accrual scheduling algorithm (or gMUA). The algorithm considers an application model where real-time activities are subject to time/utility function time constraints, variable execution time demands, and resource overloads where the total activity utilization demand exceeds the total capacity of all processors. We consider the scheduling objective of (1) probabilistically satisfying lower bounds on each activity's maximum utility, and (2) maximizing the system-wide, total accrued utility. We establish several properties of gMUA including optimal total utility (for a special case), conditions under which individual activity utility lower bounds are satisfied, a lower bound on systemwide total accrued utility, and bounded sensitivity for assurances to variations in execution time demand estimates. Finally, our simulation experiments validate our analytical results and confirm the algorithm's effectiveness.

Time/utility function time constraints (or TUFs) and utility accrual (UA) scheduling optimality c... more Time/utility function time constraints (or TUFs) and utility accrual (UA) scheduling optimality criteria, constitute, arguably, the most effective and broadest approach for adaptive, dynamic time-critical resource management. A TUF, which is a generalization of the classical deadline constraint, specifies the utility of completing an application activity as an applicationor situation-specific function of that activity's completion time. With TUF time constraints, timeliness optimality criteria can be specified in terms of accrued (e.g., summed) activity utilities. This paper overviews past and recent advances on adaptive resource management for dynamic time-critical systems using UA algorithms. Emerging challenges and new research directions are also identified.

24th International Conference of the …, 2006

The various purposes for which a dynamic tasks might be constructed, such as to test for knowledg... more The various purposes for which a dynamic tasks might be constructed, such as to test for knowledge, teach, or to assist professionals or the lay public in understanding the systems they are dealing with (or part of), are discussed. The idea analysis method is suggested as a means to fit a task to its purpose. Idea analysis entails analysing the task in terms of what basic ideas need to be familiar if one is to be able solve the task. It is just as important to know what knowledge a task does not require as to know what it does require, and if the requirements corresponds to the goal(s) motivating the construction of the task. To provide an example, the Computer Security Incident Response Team (CSIRT) task, a close analogue to the one-stock reindeer management task by Moxnes, is analysed, and several issues of general importance are revealed.

We consider scheduling distributed real-time tasks in unreliable (e.g., those with arbitrary node... more We consider scheduling distributed real-time tasks in unreliable (e.g., those with arbitrary node and network failures) and untrustworthy systems (e.g., those with Byzantine node behaviors). We present a distributed real-time scheduling algorithm called LRTG. The algorithm makes two novel contributions. First, LRTG uses gossip for reliably propagating task scheduling parameters and for discovering task execution nodes. Second, the algorithm guards against potential disruption of message propagation due to Byzantine attacks using a mechanism called LASIRC. By doing so, the algorithm provides assurances on task timeliness behaviors, despite system unreliability and untrustworthiness. Our performance evaluation shows LRTG's effectiveness.

We consider scheduling real-time distributable threads in the presence of node/link failures, mes... more We consider scheduling real-time distributable threads in the presence of node/link failures, message losses, and dynamic node joins and departures. We present a distributed scheduling algorithm called RTMG. The algorithm uses gossip-based communication for discovering eligible nodes. Traditionally, gossip protocols incur high message overhead. We explain that this problem is not that serious. We present a hybrid message propagation protocol

We consider scheduling real-time tasks in the presence of message loss and Byzantine node failure... more We consider scheduling real-time tasks in the presence of message loss and Byzantine node failures in unreliable networks. We present scheduling algorithms called RTQG and RTQG-B. The algorithms use quorum-based gossip communication strategies for dynamically and dependably discovering eligible nodes. Compared with its predecessors,our protocol exhibits better performance. RTQG utilizes quorum systems to limit the range of each gossip round. Using the intersection property of quorum systems, RTQG has advantages in message propagation and robustness to Byzantine node failures. Our simulation studies verify our analytical results.

Springer eBooks, Nov 24, 2007

We demonstrate a consensus utility accrual scheduling algorithm for distributable threads with ru... more We demonstrate a consensus utility accrual scheduling algorithm for distributable threads with run-time uncertainties in execution time, arrival models, and node crash failures. The DUA-CLA algorithm's message complexity (O(fn)), lower time complexity bound (O(D + fd + nk)), and failure-free execution time (O(D + nk)) are established, where D is the worst-case communication delay, d is the failure detection bound, n is the number of nodes, and f is the number of failures. The "lazy-abort" property is shown-abortion of currentlyinfeasible tasks is deferred until timely task completion is impossible. DUA-CLA also exhibits "schedule-safety"-threads proposed as feasible for execution by a node which fails during the decision process will not cause an otherwise-feasible thread to be excluded. These properties mark improvements over earlier strategies in common-and worst-case performance. Quantitative results obtained from our Distributed Real-Time Java implementation validate properties of the algorithm.

ABSTRACT First Page of the Article

We present RT-P2P, a real-time peer-to-peer (P2P) system that allows application-level end-to-end... more We present RT-P2P, a real-time peer-to-peer (P2P) system that allows application-level end-to-end timing requirements to be satisfied in P2P systems. P2P systems are fundamentally characterized by: a large number of geographically distributed nodes that require little infrastructural support from the underlying network; an unbounded number of nodes (a permanently evolving set of nodes); and consequently no process/node with global knowledge of the system structure. Interesting features of such networks include the fact that they allow a rich set of nodes to act as relay points for other nodes, and a rich set of overlay paths (selected by peers) to be constructed. These features have traditionally made overlay routing-where end hosts have the flexibility of routing traffic to their destinations through the desired choice of intermediate overlay nodes (unlike in IP routing)-a very attractive approach for end-to-end performance optimizations in P2P networks. Key aspects of our RT-P2P infrastructure include a real-time P2P protocol, real-time communication algorithm, and analytical performance models. We analytically establish the timing properties of RT-P2P. Our simulation studies validate our analytical results.

Springer eBooks, 2008

International Conference on Distributed Computing Systems, 1982

Journal of Parallel and Distributed Computing, Mar 1, 2010

Springer eBooks, Aug 23, 2007

Journal of Parallel and Distributed Computing, Feb 1, 2010

24th International Conference of the …, 2006

Springer eBooks, Nov 24, 2007

ABSTRACT First Page of the Article

Introduction to the Time/Utility (/Value) Function Paradigm: Abstract, 2020

Updated May 24, 2020 A Time/Utility (née Time/Value) Function (TUF) specifies an action's (e.g.... more Updated May 24, 2020

A Time/Utility (née Time/Value) Function (TUF) specifies an action's (e.g., task's) application-specific utility depending on its completion time (C). By convention, a TUF is concave. It has a critical time (even if it is linear) after which its utility does not increase. A conventional deadline (d) is a simple special case, a downward step TUF having utility values {1,0}. More generally, a TUF permits downward (and upward) step functions to have any appropriate utilities {u1, u2}. Tardiness is a simple special case whose non-zero utility is the linear function C-d. More generally, a TUF allows non-zero earliness and tardiness to be non-linear. Thus, one useful interpretation of utility can be timeliness, providing a rich generalization of traditional action completion time constraints in real-time systems. TUF utility may include negative values. TUFs and their utility scales and values are derived from domain-specific subject matter knowledge. The optimality criteria for scheduling TUFs are maximal utility accrual (UA)-which can be interpreted as actions' collective timeliness-and predictability of that accrued utility (while respecting dependencies and resource constraints). The scheduler performs application-specific trade-offs between accrued utility and its predictability. The TUF/UA paradigm is intended for (but not limited to) open-world systems, so imperfections in the scheduling parameters are inevitable. Some of these imperfections may be amenable to stochastic scheduling. Others are too major and complex for orthodox (e.g., additive) probability theory. Imprecision may call for using fuzzy set theory. Epistemic uncertainties--e.g., ignorance of, or conflicts among, scheduling parameters--may be present and require an appropriate kind and degree of resilience in the UA algorithmic techniques. Thus, UA schedulers may base utility accrual and its predictability on more general uncertainty models, such as a (potentially "fuzzified" version of a) belief-based theory (e.g., the Transferable Belief Model, etc.). Online UA scheduling efficiency is often enhanced by implementing the scheduler in hardware (e.g., custom RISC-Vs, GPUs, FPGAs, ASICs). The TUF/UA paradigm has been particularly successful in military combat systems, because of the extreme uncertainties in those environments.

Digest and its References of Introduction to Utility (Value) Based Scheduling (6 Feb 20), 2020

This Introduction describes the TUF/UA paradigm and its default system model in more detail than ... more This Introduction describes the TUF/UA paradigm and its default system model in more detail than previously documented. The default system model is based on the author's extensive experience with applications which use this paradigm, and on generality. Any instantiation of the paradigm is application-specific and normally has a system model different from (a subset of) the default one. The default system model also serves to illuminate research opportunities in this area. This paradigm has the side effect of highlighting the greater generality and applicability of "real-time" than is conventionally perceived, especially by the computing community. A Time/Utility Function (TUF)-originally [Jensen 77] [Jensen+ 85], and still often by others, called Time/Value Function-expresses the timeliness (both urgency and worth) of completing an action (such as a computational task or a device physical motion) in an application-specific utility ratio scale [Prasad+ 03], as an application-specific function of when that action's operation completes. TUFs and their utility scales and values are representations derived from system-and application-specific subject matter knowledge (e.g., Clark+ 99] [Theys+ 91]).

EXTENDED ABSTRACT: An Introduction to Time/Utility (or Time/Value) Functions and Utility Accrual Real-Time Scheduling, 2019

Extended Abstract (with References) This Introduction describes the TUF/UA paradigm and its def... more Extended Abstract (with References)

This Introduction describes the TUF/UA paradigm and its default system model in more detail than previously documented. The default system model is based on application experience and generality. Any instantiation of the paradigm is application-specific and normally has a system model different from (a subset of) the default one. This paradigm has the side effect of illuminating the greater generality and applicability of "real-time" than is conventionally perceived. ¶ A Time/Utility Function (TUF)-originally [Jensen 77] [Jensen+ 85], and still often by others, called Time/Value Function-expresses the timeliness (both urgency and worth) of completing an action (such as a computational task or a device physical motion) in an application-specific utility ratio scale [Prasad+ 03], as an application-specific function of when (usually* in physical time) that action's operation completes. TUFs and their utility scales and values are derived from system-and application-specific subject matter knowledge (e.g., Clark+ 99]). A framework for assigning TUF utility values has been proposed for a class of very limited cases [Burns+ 2000]; but most cases necessitate specialized tools and techniques (e.g., [Lee+ 07]) *. ¶ Utility functions are normally non-convex-concave or linear (cf. Figure 2 in [Jensen 04]). Constant ones can either represent priority or be one way to represent relative importance. A (non-constant) TUF may have a "critical time" (a deadline is a special case), after which its utility does not increase [Locke 86]. A utility function's range may include negative values (penalties). A TUF's shape may be dynamically adapted-e.g., at an action or application mode change (such as for ballistic missile flight phases) [Maynard+ 88], cf. Figures 8 and 9 in [Jensen 04]. ¶ Actions which are being scheduled collectively may be any mix of aperiodic, sporadic, and periodic, but periodic actions do not receive traditional preferential treatment. Collectively scheduled (perhaps a subset of) actions are made to have consonant scales. At each of one or more scheduling instants, the scheduling algorithm considers the TUFs of all ready actions (and other system model information about the current situation), and schedules a set (from one to all) of those actions for operation. Specific UA scheduling algorithms are devised to provide specific acceptable timeliness and predictability-real-time QoS-assurances for specific system models and application situations. ¶ The scheduling is primarily according to two application-specific real-time QoS criteria-Utility Accrual (UA), which is the (commonly, expected) polynomial sum of their collective and individual utilities; and the (usually non-deterministic) predictability of that sum (i.e., of system* timeliness). Thus, UA scheduling is generally not greedy (and intentionally not fair), so actions may be either preemptible or not. A schedule normally also considers actions' constraints such as precedence and resource dependencies [Clark 90] [Li+ 06], and properties such as energy [Wu 05] and relative importance. ¶ In this paradigm, importance is application-specific (e.g., in terms of track quality, or weapon spherical error probable, etc.) and dynamic *. It is distinct from, and may be orthogonal to, an action's timeliness (urgency)-both

ABSTRACT Informally, a system is a “real-time” one if its core properties of timeliness and pred... more ABSTRACT

Informally, a system is a “real-time” one if its core properties of timeliness and predictability of timeliness are integral to its logic, not only performance measures. In general, those properties are dynamic due primarily to intrinsic epistemic uncertainties—e.g., ignorance, inaccuracy, non-determinism—in the system and its application environment. Despite such uncertainties, dynamically real-time systems have multiple application-specific kinds and degrees of criticality—including even the most extreme safety-critical systems (e.g., for warfare). Traditional real-time computing systems are a special case whose core properties are predominately static and presumed to be known á priori, thus greatly limiting those systems’ applicability. Many dynamically real-time systems exist, built by application domain experts outside the real-time field, but without the benefits of a coherent foundation for real-time per se. The design, implementation, and application of real-time systems can be extended and strengthened by creating such a foundation based on principles for those core properties. This book introduces one approach to that. Timeliness is dynamically expressive using time/utility functions. Uncertainty of timeliness predictability can be reasoned about using various mathematical theories of evidence (belief functions). The foundation has been successfully employed in different real-time contexts having wickedly dynamic core properties, where traditional static real-time perspectives and techniques were insufficient.

This is the Introduction page in my Work-in-progress preview of a work-in-progress book "Introduc... more This is the Introduction page in my Work-in-progress preview of a work-in-progress book "Introduction to Fundamental Principles of Dynamically Real-Time Systems," [Jensen 2018]. The Abstract page is in another of my sessions here.