Search For:

Displaying 1-37 out of 37 total
Looking at data
Found in: Parallel and Distributed Processing Symposium, International
By Dror G. Feitelson
Issue Date:April 2008
pp. 1-9
Collecting and analyzing data lies at the basis of the scientific method: findings about nature usher new ideas, and experimental results support or refute theories. All this is not very prevalent in computer science, possibly due to the fact that computer...
   
Paired Gang Scheduling
Found in: IEEE Transactions on Parallel and Distributed Systems
By Yair Wiseman, Dror G. Feitelson
Issue Date:June 2003
pp. 581-592
<p><b>Abstract</b>—Conventional gang scheduling has the disadvantage that when processes perform I/O or blocking communication, their processors remain idle because alternative processes cannot be run independently of their own gangs. To ...
 
The Supercomputer Industry in Light of the Top500 Data
Found in: Computing in Science and Engineering
By Dror G. Feitelson
Issue Date:January 2005
pp. 42-47
The Top500 list, which has been updated semiannually for the past decade, ranks the 500 most powerful computers installed worldwide. Analyzing this data gives an impartial look at the supercomputer industry's current state and development trends, and sheds...
 
On the Scalability of Centralized Control
Found in: Parallel and Distributed Processing Symposium, International
By Dror G. Feitelson
Issue Date:April 2005
pp. 298a
Scalability of clusters and MPPs is typically discussed in terms of limits on growth: something which grows at a rate of O(log p) (where p is the number of processors) is said to be more scalable than something whose growth rate is O(p). But in practice p ...
 
Hierarchical Indexing and Document Matching in BoW
Found in: Digital Libraries, Joint Conference on
By Maayan Geffet, Dror G. Feitelson
Issue Date:June 2001
pp. 259-267
BoW is an on-line bibliographical repository based on a hierarchical concept index to which entries are linked. Searching in the repository should therefore return matching topics from the hierarchy, rather than just a list of entries. Likewise, when new e...
 
Supporting Priorities and Improving Utilization of the IBM SP Scheduler Using Slack-Based Backfilling
Found in: Parallel Processing Symposium, International
By David Talby, Dror G. Feitelson
Issue Date:April 1999
pp. 513
Distributed memory parallel systems such as the IBM SP2 execute jobs using variable partitioning. Scheduling jobs in FCFS order leads to severe fragmentation and utilization loss, which lead to the development of backfilling scheudlers such as EASY. This p...
 
Distributed Hierarchical Control for Parallel Processing
Found in: Computer
By Dror G. Feitelson, Larry Rudolph
Issue Date:May 1990
pp. 65-77
<p>A description is given of a novel design, using a hierarchy of controllers, that effectively controls a multiuser, multiprogrammed parallel system. Such a structure allows dynamic repartitioning according to changing job requirements. The design g...
 
Development and Deployment at Facebook
Found in: IEEE Internet Computing
By Dror G. Feitelson,Eitan Frachtenberg,Kent L. Beck
Issue Date:July 2013
pp. 8-17
Internet companies such as Facebook operate in a "perpetual development" mindset. This means that the website continues to undergo development with no predefined final objective, and that new developments are deployed so that users can enjoy them...
 
Exploiting Core Working Sets to Filter the L1 Cache with Random Sampling
Found in: IEEE Transactions on Computers
By Yoav Etsion,Dror G. Feitelson
Issue Date:November 2012
pp. 1535-1550
Locality is often characterized by working sets, defined by Denning as the set of distinct addresses referenced within a certain window of time. This definition ignores the fact that dramatic differences exist between the usage patterns of frequently used ...
 
A global scheduling framework for virtualization environments
Found in: Parallel and Distributed Processing Symposium, International
By Yoav Etsion,Tal Ben-Nun,Dror G. Feitelson
Issue Date:May 2009
pp. 1-8
A premier goal of resource allocators in virtualization environments is to control the relative resource consumption of the different virtual machines, and moreover, to be able to change the relative allocations at will. However, it is not clear what it me...
 
Success of Open Source Projects: Patterns of Downloads and Releases with Time
Found in: Software Science, Technology and Engineering, IEEE International Conference on
By Ayelet Israeli, Dror G. Feitelson
Issue Date:October 2007
pp. 87-94
<p>The success of software projects has many different facets: meeting user requirements, being developed within given budget and time constraints, and actually being used. We focus on one of them: the pattern of adoption by users, and its possible r...
 
L1 Cache Filtering Through Random Selection of Memory References
Found in: Parallel Architectures and Compilation Techniques, International Conference on
By Yoav Etsion, Dror G. Feitelson
Issue Date:September 2007
pp. 235-244
Distinguishing transient blocks from frequently used blocks enables servicing references to transient blocks from a small fully-associative auxiliary cache structure. By inserting only frequently used blocks into the main cache structure, we can reduce the...
 
Backfilling Using System-Generated Predictions Rather than User Runtime Estimates
Found in: IEEE Transactions on Parallel and Distributed Systems
By Dan Tsafrir, Yoav Etsion, Dror G. Feitelson
Issue Date:June 2007
pp. 789-803
<p><b>Abstract</b>—The most commonly used scheduling algorithm for parallel supercomputers is FCFS with backfilling, as originally introduced in the EASY scheduler. Backfilling means that short jobs are allowed to run ahead of their time ...
 
Asimov's Laws of Robotics Applied to Software
Found in: IEEE Software
By Dror G. Feitelson
Issue Date:July 2007
pp. 112, 111
Asimov's Laws of Robotics constrain robots to serve their human masters. Minor rewording shows that similar principles are very relevant to software too. These laws of software encompass a host of desiderata and trade-offs that software developers need to ...
 
Probabilistic Prediction of Temporal Locality
Found in: IEEE Computer Architecture Letters
By Yoav Etsion, Dror G. Feitelson
Issue Date:January 2007
pp. 17-20
The increasing gap between processor and memory speeds, as well as the introduction of multi-core CPUs, have exacerbated the dependency of CPU performance on the memory subsystem. This trend motivates the search for more efficient caching mechanisms, enabl...
 
Using Site-Level Modeling to Evaluate the Performance of Parallel System Schedulers
Found in: Modeling, Analysis, and Simulation of Computer Systems, International Symposium on
By Edi Shmueli, Dror G. Feitelson
Issue Date:September 2006
pp. 167-178
The conventional performance evaluation methodology for parallel system schedulers uses an open model to generate the workloads used in simulations. In many cases recorded workload traces are simply played back, assuming that they are reliable representati...
 
Metrics for Mass-Count Disparity
Found in: Modeling, Analysis, and Simulation of Computer Systems, International Symposium on
By Dror G. Feitelson
Issue Date:September 2006
pp. 61-68
Mass-count disparity is the technical underpinning of the
 
An Empirically-Based Criterion for Determining the Success of an Open-Source Project
Found in: Software Engineering Conference, Australian
By Dror G. Feitelson, Gillian Z. Heller, Stephen R. Schach
Issue Date:April 2006
pp. 363-368
In order to determine a success criterion for open-source software projects, we analyzed 122,205 projects in the SourceForge database. There were 80,597 projects with no downloads at all. We restricted our analysis to the 41,608 projects that together were...
 
Adaptive Parallel Job Scheduling with Flexible Coscheduling
Found in: IEEE Transactions on Parallel and Distributed Systems
By Eitan Frachtenberg, Dror G. Feitelson, Fabrizio Petrini, Juan Fernández
Issue Date:November 2005
pp. 1066-1077
<p><b>Abstract</b>—Many scientific and high-performance computing applications consist of multiple processes running on different processors that communicate frequently. Because of their synchronization needs, these applications can suffe...
 
Improving and Stabilizing Parallel Computer Performance Using Adaptive Backfilling
Found in: Parallel and Distributed Processing Symposium, International
By David Talby, Dror G. Feitelson
Issue Date:April 2005
pp. 84a
The scheduler is a key component in determining the overall performance of a parallel computer, and as we show here, the schedulers in wide use today exhibit large unexplained gaps in performance during their operation. Also, different scheduling algorithm...
 
Experimental Analysis of the Root Causes of Performance Evaluation Results: A Backfilling Case Study
Found in: IEEE Transactions on Parallel and Distributed Systems
By Dror G. Feitelson
Issue Date:February 2005
pp. 175-182
<p><b>Abstract</b>—The complexity of modern computer systems may enable minor variations in performance evaluation procedures to actually determine the outcome. Our case study concerns the comparison of two parallel job schedulers, using ...
 
Flexible CoScheduling: Mitigating Load Imbalance and Improving Utilization of Heterogeneous Resources
Found in: Parallel and Distributed Processing Symposium, International
By Eitan Frachtenberg, Dror G. Feitelson, Fabrizio Petrini, Juan Fernandez
Issue Date:April 2003
pp. 85b
Fine-grained parallel applications require all their processes to run simultaneously on distinct processors to achieve good efficiency. This is typically accomplished by space slicing, wherein nodes are dedicated for the duration of the run, or by gang sch...
 
Barrier Synchronization on a Loaded SMP Using Two-Phase Waiting Algorithms
Found in: Parallel and Distributed Processing Symposium, International
By Dan Tsafrir, Dror G. Feitelson
Issue Date:April 2002
pp. 0080
Little work has been done on the performance of barrier synchronization using two-phase blocking, as the common wisdom is that it is useless to spin if the total number of threads in the system exceeds the number of processors. We challenge this and show t...
 
User-Level Communication in a System with Gang Scheduling
Found in: Parallel and Distributed Processing Symposium, International
By Yoav Etsion, Dror G. Feitelson
Issue Date:April 2001
pp. 10058a
One of the scarce resources that limits communication performance is buffer space on the network interface card. This becomes even worse when it is partitioned among several time-sliced processes. However, if gang scheduling is used, it is possible to swap...
 
Gang Scheduling with Memory Considerations
Found in: Parallel and Distributed Processing Symposium, International
By Anat Batat, Dror G. Feitelson
Issue Date:May 2000
pp. 109
A major problem with time slicing on parallel machines is memory pressure, as the resulting paging activity damages the synchronism among a job's processes. An alternative is to impose admission controls, and only admit jobs that fit into the available mem...
 
Self-Tuning Systems
Found in: IEEE Software
By Dror G. Feitelson, Michael Naaman
Issue Date:March 1999
pp. 52-60
<p>Tuning operating systems helps system administrators improve system performance and efficiency, but it can be time-consuming. These authors propose a mechanism, based on genetic algorithms, to automate this process by running simulations of system...
 
The Parallel Break Construct, or How to Kill an Activity Tree
Found in: Parallel Processing Symposium, International
By Yair I. Friedman, Dror G. Feitelson, Iaakov Exman
Issue Date:April 1996
pp. 230
Most parallel languages provide means to express parallelism, e.g. a parallel-do construct, but no means to terminate the parallel activities spawned by such constructs. We propose three high-level primitives for this purpose, which are defined by analogie...
 
Parallel I/O Subsystems in Massively Parallel Supercomputers
Found in: IEEE Concurrency
By Dror G. Feitelson, Peter F. Corbett, Sandra Johnson Baylor, Yarsun Hsu
Issue Date:September 1995
pp. 33-47
Applications on MPPs often require a high aggregate bandwidth of low-latency I/O to secondary storage. This requirement can met by internal parallel I/O subsystems that comprise dedicated I/O nodes, each with processor, memory, and disks. Massively paralle...
 
Characterization and assessment of the linux configuration complexity
Found in: 2013 IEEE 13th International Working Conference on Source Code Analysis and Manipulation (SCAM)
By Ahmad Jbara,Dror G. Feitelson
Issue Date:September 2013
pp. 11-20
The Linux kernel is configured for specific uses by manipulations of the source code during the compilation process. These manipulations are performed by the C pre-processor (CPP), based on in-line directives. Such directives, and the interleaving of multi...
   
On extracting session data from activity logs
Found in: Proceedings of the 5th Annual International Systems and Storage Conference (SYSTOR '12)
By David Mehrzadi, Dror G. Feitelson
Issue Date:June 2012
pp. 1-7
Activity logs from large-scale systems facilitate the study of user behavior, which can be used to improve and tune the user experience. However, the available data often lacks important elements such as the identification of user sessions. Previous work t...
     
Trading off quality for throughput using content adaptation in web servers
Found in: Proceedings of the 4th Annual International Conference on Systems and Storage (SYSTOR '11)
By Dror G. Feitelson, Michael Gopshtein
Issue Date:May 2011
pp. 1-14
A basic problem in managing web servers is capacity planning. A partial solution is to use content adaptation, where the system automatically trades off quality for throughput, e.g. by eliminating graphical decorations and adjusting page layout. We evaluat...
     
Distinguishing humans from robots in web search logs: preliminary results using query rates and intervals
Found in: Proceedings of the 2009 workshop on Web Search Click Data (WSCD '09)
By Dror G. Feitelson, Omer Duskin
Issue Date:February 2009
pp. 15-19
The workload on web search engines is actually multiclass, being derived from the activities of both human users and automated robots. It is important to distinguish between these two classes in order to reliably characterize human web search behavior, and...
     
Process prioritization using output production: Scheduling for multimedia
Found in: ACM Transactions on Multimedia Computing, Communications, and Applications (TOMCCAP)
By Dan Tsafrir, Dror G. Feitelson, Yoav Etsion
Issue Date:November 2006
pp. 318-342
Desktop operating systems such as Windows and Linux base scheduling decisions on CPU consumption; processes that consume fewer CPU cycles are prioritized, assuming that interactive processes gain from this since they spend most of their time waiting for us...
     
Desktop scheduling: how can we know what the user wants?
Found in: Proceedings of the 14th international workshop on Network and operating systems support for digital audio and video (NOSSDAV '04)
By Dan Tsafrir, Dror G. Feitelson, Yoav Etsion
Issue Date:June 2004
pp. 110-115
Current desktop operating systems use CPU utilization (or lack thereof) to prioritize processes for scheduling. This was thought to be beneficial for interactive processes, under the assumption that they spend much of their time waiting for user input. Thi...
     
Effects of clock resolution on the scheduling of interactive and soft real-time processes
Found in: Proceedings of the 2003 ACM SIGMETRICS international conference on Measurement and modeling of computer systems (SIGMETRICS '03)
By Dan Tsafrir, Dror G. Feitelson, Yoav Etsion
Issue Date:June 2003
pp. 172-183
It is commonly agreed that scheduling mechanisms in general purpose operating systems do not provide adequate support for modern interactive applications, notably multimedia applications. The common solution to this problem is to devise specialized schedul...
     
Hierarchical indexing and document matching in BoW
Found in: Proceedings of the first ACM/IEEE-CS joint conference on Digital libraries (JCDL '01)
By Dror G. Feitelson, Maayan Geffet
Issue Date:January 2001
pp. 259-267
BoW is an on-line bibliographical repository based on a hierarchical c oncept index to which entries are linked. Searching in the repository should therefore return matching topics from the hierarchy, rather than just a list of entries. Likewise, when new ...
     
The Vesta parallel file system
Found in: ACM Transactions on Computer Systems (TOCS)
By Dror G. Feitelson, Peter F. Corbett
Issue Date:February 1992
pp. 225-264
The Vesta parallel file system is designed to provide parallel file access to application programs running on multicomputers with parallel I/O subsystems. Vesta uses a new abstraction of files: a file is not a sequence of bytes, but rather it can be partit...
     
 1