Efficient monitoring of Web resources Avigdor Gal (joint work with Haggai Roitman and Louiqa Raschid) IFIP 2.6 meeting 24/6/2009, Nicosia, Cyprus.

Slides:



Advertisements
Similar presentations
Feedback Control Real-Time Scheduling: Framework, Modeling, and Algorithms Chenyang Lu, John A. Stankovic, Gang Tao, Sang H. Son Presented by Josh Carl.
Advertisements

QoS-based Management of Multiple Shared Resources in Dynamic Real-Time Systems Klaus Ecker, Frank Drews School of EECS, Ohio University, Athens, OH {ecker,
Tight Bounds for Online Class- constrained Packing Hadas Shachnai Bell Labs and The Technion IIT Tami Tamir The Technion IIT.
Analysis of : Operator Scheduling in a Data Stream Manager CS561 – Advanced Database Systems By Eric Bloom.
ECE 667 Synthesis and Verification of Digital Circuits
Hadi Goudarzi and Massoud Pedram
Best-Effort Top-k Query Processing Under Budgetary Constraints
Playback delay in p2p streaming systems with random packet forwarding Viktoria Fodor and Ilias Chatzidrossos Laboratory for Communication Networks School.
Online Scheduling with Known Arrival Times Nicholas G Hall (Ohio State University) Marc E Posner (Ohio State University) Chris N Potts (University of Southampton)
Cloud Computing Resource provisioning Keke Chen. Outline  For Web applications statistical Learning and automatic control for datacenters  For data.
A system Performance Model Instructor: Dr. Yanqing Zhang Presented by: Rajapaksage Jayampthi S.
All Hands Meeting, 2006 Title: Grid Workflow Scheduling in WOSE (Workflow Optimisation Services for e- Science Applications) Authors: Yash Patel, Andrew.
1 Sensor Relocation in Mobile Sensor Networks Guiling Wang, Guohong Cao, Tom La Porta, and Wensheng Zhang Department of Computer Science & Engineering.
Variability Oriented Programming – A programming abstraction for adaptive service orientation Prof. Umesh Bellur Dept. of Computer Science & Engg, IIT.
Towards Feasibility Region Calculus: An End-to-end Schedulability Analysis of Real- Time Multistage Execution William Hawkins and Tarek Abdelzaher Presented.
UNIVERSITY OF JYVÄSKYLÄ Yevgeniy Ivanchenko Yevgeniy Ivanchenko University of Jyväskylä
An architecture for Privacy Preserving Mining of Client Information Jaideep Vaidya Purdue University This is joint work with Murat.
1 On Handling QoS Traffic in Wireless Sensor Networks 吳勇慶.
Scheduling for Embedded Real-Time Systems Amit Mahajan and Haibo.
1 GOAL PROGRAMMING. 2 We will now address problems that involve multiple, conflicting objectives that can be tackled by linear programming techniques.
A general approximation technique for constrained forest problems Michael X. Goemans & David P. Williamson Presented by: Yonatan Elhanani & Yuval Cohen.
EE 685 presentation Optimization Flow Control, I: Basic Algorithm and Convergence By Steven Low and David Lapsley Asynchronous Distributed Algorithm Proof.
System-Wide Energy Minimization for Real-Time Tasks: Lower Bound and Approximation Xiliang Zhong and Cheng-Zhong Xu Dept. of Electrical & Computer Engg.
Differentiated Multimedia Web Services Using Quality Aware Transcoding S. Chandra, C.Schlatter Ellis and A.Vahdat InfoCom 2000, IEEE Journal on Selected.
A Hybrid Caching Strategy for Streaming Media Files Jussara M. Almeida Derek L. Eager Mary K. Vernon University of Wisconsin-Madison University of Saskatchewan.
SWIM 1/9/20031 QoS in Data Stream Systems Rajeev Motwani Stanford University.
Penn ESE535 Spring DeHon 1 ESE535: Electronic Design Automation Day 5: February 2, 2009 Architecture Synthesis (Provisioning, Allocation)
New Challenges in Cloud Datacenter Monitoring and Management
1/24 Algorithms for Generalized Caching Nikhil Bansal IBM Research Niv Buchbinder Open Univ. Israel Seffi Naor Technion.
IBM Research – Thomas J Watson Research Center | March 2006 © 2006 IBM Corporation Events and workflow – BPM Systems Event Application symposium Parallel.
Balancing energy demand and supply without forecasts: online approaches and algorithms Giorgos Georgiadis.
Efficient Scheduling of Heterogeneous Continuous Queries Mohamed A. Sharaf Panos K. Chrysanthis Alexandros Labrinidis Kirk Pruhs Advanced Data Management.
Embedded System Design Framework for Minimizing Code Size and Guaranteeing Real-Time Requirements Insik Shin, Insup Lee, & Sang Lyul Min CIS, Penn, USACSE,
1 An SLA-Oriented Capacity Planning Tool for Streaming Media Services Lucy Cherkasova, Wenting Tang, and Sharad Singhal HPLabs,USA.
Multi-Unit Auctions with Budget Limits Shahar Dobzinski, Ron Lavi, and Noam Nisan.
Network Aware Resource Allocation in Distributed Clouds.
Approximation Algorithms for NP-hard Combinatorial Problems Magnús M. Halldórsson Reykjavik University
An Integration Framework for Sensor Networks and Data Stream Management Systems.
Efficient and Scalable Computation of the Energy and Makespan Pareto Front for Heterogeneous Computing Systems Kyle M. Tarplee 1, Ryan Friese 1, Anthony.
1 Evaluating top-k Queries over Web-Accessible Databases Paper By: Amelie Marian, Nicolas Bruno, Luis Gravano Presented By Bhushan Chaudhari University.
© 2009 IBM Corporation 1 Improving Consolidation of Virtual Machines with Risk-aware Bandwidth Oversubscription in Compute Clouds Amir Epstein Joint work.
Scheduling policies for real- time embedded systems.
Chapter 8 PD-Method and Local Ratio (4) Local ratio This ppt is editored from a ppt of Reuven Bar-Yehuda. Reuven Bar-Yehuda.
ECO-DNS: Expected Consistency Optimization for DNS Chen Stephanos Matsumoto Adrian Perrig © 2013 Stephanos Matsumoto1.
August 16, 2010 MPREF’10 Dynamic House Allocation Sujit Gujar 1, James Zou 2 and David C. Parkes 2 5 th Multidisciplinary Workshop on Advances in Preference.
Competitive Queue Policies for Differentiated Services Seminar in Packet Networks1 Competitive Queue Policies for Differentiated Services William.
EE 685 presentation Optimization Flow Control, I: Basic Algorithm and Convergence By Steven Low and David Lapsley.
1 Iterative Integer Programming Formulation for Robust Resource Allocation in Dynamic Real-Time Systems Sethavidh Gertphol and Viktor K. Prasanna University.
Information Theory for Mobile Ad-Hoc Networks (ITMANET): The FLoWS Project Competitive Scheduling in Wireless Networks with Correlated Channel State Ozan.
Memory Hierarchy Adaptivity An Architectural Perspective Alex Veidenbaum AMRM Project sponsored by DARPA/ITO.
A Optimal On-line Algorithm for k Servers on Trees Author : Marek Chrobak Lawrence L. Larmore 報告人:羅正偉.
Trust and Security for Next Generation Grids, Securing Grid-Based Supply Chains Marco Di Girolamo HP Italy Innovation Center, Italy On.
Loss-Bounded Analysis for Differentiated Services. By Alexander Kesselman and Yishay Mansour Presented By Sharon Lubasz
CSCI1600: Embedded and Real Time Software Lecture 24: Real Time Scheduling II Steven Reiss, Fall 2015.
Rounding scheme if r * j  1 then r j := 1  When the number of processors assigned in the continuous solution is between 0 and 1 for each task, the speed.
Chapter 9: Web Services and Databases Title: NiagaraCQ: A Scalable Continuous Query System for Internet Databases Authors: Jianjun Chen, David J. DeWitt,
Sunpyo Hong, Hyesoon Kim
An overlay for latency gradated multicasting Anwitaman Datta SCE, NTU Singapore Ion Stoica, Mike Franklin EECS, UC Berkeley
On the Placement of Web Server Replicas Yu Cai. Paper On the Placement of Web Server Replicas Lili Qiu, Venkata N. Padmanabhan, Geoffrey M. Voelker Infocom.
Determining Optimal Processor Speeds for Periodic Real-Time Tasks with Different Power Characteristics H. Aydın, R. Melhem, D. Mossé, P.M. Alvarez University.
Resource Optimization for Publisher/Subscriber-based Avionics Systems Institute for Software Integrated Systems Vanderbilt University Nashville, Tennessee.
Spark on Entropy : A Reliable & Efficient Scheduler for Low-latency Parallel Jobs in Heterogeneous Cloud Huankai Chen PhD Student at University of Kent.
Chapter 8 PD-Method and Local Ratio (5) Equivalence This ppt is editored from a ppt of Reuven Bar-Yehuda. Reuven Bar-Yehuda.
PA an Coordinated Memory Caching for Parallel Jobs
Distributed Submodular Maximization in Massive Datasets
Ontology Evolution: A Methodological Overview
Cluster Resource Management: A Scalable Approach
Coverage Approximation Algorithms
Utility-Function based Resource Allocation for Adaptable Applications in Dynamic, Distributed Real-Time Systems Presenter: David Fleeman {
Data Placement Problems in Database Applications
Presentation transcript:

Efficient monitoring of Web resources Avigdor Gal (joint work with Haggai Roitman and Louiqa Raschid) IFIP 2.6 meeting 24/6/2009, Nicosia, Cyprus

Profile-Based Online Data Delivery Data delivery: the delivery of data of interest (specified in profiles) from servers (data providers) to clients (data consumers).  Push vs. Pull  Server capabilities vs. Client requirements Profiles: specify what, when, how data should be delivered, and its delivery value. Online: the decision making of what and when to deliver is usually done without a complete knowledge of the all “stream” of future requirements or capabilities in the system, while considering the sources’ dynamic behavior.

Example: Monitoring RSS Feeds pull push Other example applications: E-Commerce & E-Markets Grid Mashups & Portals Continuous Queries (CQ) Cache Management …

Research Goals Proposed a generic model for profile based online data delivery.  Allows to negotiate over the dynamic nature of resources and use time based constraints. Considered both server capabilities and user requirements.  allow the generation of a hybrid push-pull solution. Handle various data delivery aspects:  Dual approach for targeted data delivery.  Hybrid push-pull framework and data delivery solution.  Capturing data delivery tradeoffs  Complex data delivery under bandwidth constraints

Related Work  Push  Systems: BlackBerry, JMS, Google Alerts  Web Caching & Synchronization  Publish/Subscribe  Stream processing/CQ & CEP  Broadcast systems  CDNs (e.g., RSS aggregation)  Pull  Update Models  Web Crawling/Monitoring (WIC)  Sensor-nets  Grid  Web Services  Mashups  Web Caching (LR-Profiles, Prefetching)  PDCM  Hybrid: Pop-Pap, Data Gerrymandering, Ajax, RSS

Data delivery model: Data and Architecture Data delivery model: Data and Architecture

ProMo Proxy - Overview

Data delivery model: Profiles We propose a novel profile model based on execution intervals. Execution Interval: an association of a time interval with some resource.  Complex execution intervals can be also specified.  Can be specified either explicitly or implicitly (using EI-patterns that are further derived using an update model).  Have some unique properties that effect scheduling. Profiles: a set of execution intervals, and include:  Notification rules that associate utility values to execution intervals.  Profile owner role (either client or a server).

Execution intervals - Example

Example Client Profile profile owner role notification rules complex-EI pattern local and global utilities

Schedules, Constraints, and Data Delivery Metrics Schedule: A mapping Constrained Schedules: limited budget for different data delivery tasks (e.g., “politeness” constraints or upper bound on parallel monitoring/listening tasks). Data delivery metrics:  Completeness (max)  Data latency (min)  System resource utilization (Probes) (min)  Execution time (min)  Gained Utility (Satisfiability) (strict) Data delivery objectives and performance evaluation are based on those metrics.

A Dual Approach for Targeted Data Delivery Instead of maximizing utility under (strict) system resource constraint, minimize system resource utilization while (strictly) satisfying (all) user profiles. Main motivation: dynamic allocation according to user profiles may produce benefit for both objectives.  We propose an optimal static algorithm SUP for the dual problem.  Under some conditions, SUP is even optimal for both objectives!  We further present adaptive versions of SUP, fbSUP and fbSUP(λ), that handle non-static situations using feedback.  Overall, results show that the dual approach is capable to dominate the traditional approach and has good utility/budget performance in the non-static case.

ProMo: Hybrid Framework for Online Data Delivery Idea: mediate between clients and servers while considering both client requirements and server capabilities. Solution: use the same profile structure both for servers and clients, as a result:  Matching clients and servers becomes easy.  Easy to generate hybrid schedules. We provide a taxonomy of server capabilities and data delivery patterns. The algorithm supports various capability patterns (e.g., pull-only, push-only, hybrid, push-filter, and conditional- pull)

The Proxy Capturing Approximate Data Delivery Tradeoffs (“The Proxy Dilemma”) Completeness Completeness  more alternatives for drivers Delay Delay  More time to react High completeness may result in delayed delivery  less time to react.  Low delay may results in missing updates  less alternatives to consider. 

Bandwidth Constrained Complex Profile Satisfaction Example 1: Arbitrage MonitoringExample 2: Mashups

Future Work Dual approach:  Consider more constrained settings (e.g., lower bound of gained utility).  Adaptive switch between OptMon1 and OptMon2 solutions.  More general probabilistic adaptive framework.  None-uniform probing costs. ProMo hybrid push-pull  Consider a constrained setting with ProMo (e.g., minimization push-pull costs, politeness constraints).  Develop a more refined server commitment model and server selection and ranking techniques.

Future Work (cont.) Tradeoffs:  Find offline approximation for the general case.  Find online policies with competitive guarantees for the general case.  Usage of Pareto sets as design tool for online policies.  Private profiles based tradeoffs.  Consider complex profiles. Complex:  General cost-benefit model (e.g., consider utility gain vs. monitoring costs).  Use other complex profile semantics (e.g., OR, SUBSET).  Develop update models for complex monitoring.

Backup Slides

Schedules (cont.) delay TjTj riri

Model: Feasible Schedules

Execution Intervals – Properties and Effects on Scheduling Inter-resource overlap  Directly affects the probing congestion Intra-resource overlap  Allows more then a single EI to be captured by a single probe. Rank  Effects the difficulty in satisfying a single client requirement or finding a suitable server capability (thus, some pull will be required).  Can cause a skew in resource access patterns. Explicit vs. Implicit  Implicit may require to use update models to derive explicit EIs and therefore, introduce noise in to the model. Utility  Effect the relative importance of capturing.

SUP – Dual Optimality max clique

fbSUP (SUP with feedback)

fbSUP(λ)

SUP vs. TTL & WIC Static case - FPN(1.0)Dynamic case - Poison #probes(SUP) = 2,462 max #probes(WIC) = >65,000 max #probes(TTL) = >65,000 #probes(SUP) = 3,904 max #probes(WIC) = >20,000 max #probes(TTL) = >7,000

fbSUP vs. fbSUP(λ) Both adaptive versions improve on SUP (with moderate probe budget increase) fbSUP(λ) improve even for X=1 fbSUP(λ) is the dominant Up to X<4 fbSUP(λ) requires slightly more budget then fbSUP For X≥4 fbSUP(λ) completely dominates fbSUP.

ProMo – Server Capabilities vs. Data Delivery Patterns

ProMo Middleware - Example

Pareto Sets and Approximation

Efficient Offline Optimal Solution Efficient Offline Optimal Solution (case with no intra-resource overlaps)

Offline Optimal Algorithm Correctness From the algorithm construction: Pareto optimality: By induction, let S j be the j th schedule that is added to S: SPSP S SjSj S’ S

Efficient Online Policies Efficient Online Policies (case with no intra- resource overlaps) Look Ahead: Look Back: TradeoffLatencyCompletenessOrderingPolicy ??Optimal LA 4-approx. 2-approx. LB ?Less then LA Optimal LAB 2-approx.Optimal2-approx. LBA

LA (EDF) Optimal completeness LA (EDF) Optimal completeness (no intra- resource overlap) r i’ riri case 1: TjTj r i’ riri case 2: TjTj (preserve)  gain.  other resource r i’ was selected (but not by LA)  apply Lemma 42 (preserve or gain). case 1: (gain) case 2:  preserve.  other resource r i’ was selected by LA (but not by S ^ )  apply Lemma 42 (preserve or gain).

LAB Dominates LA  The proof follows from Lemma 42 (completeness preservation) and Lemma 47  each local change from LA to LAB would result in less delay.Lemma 42  The proof follows from the definition of LAB potential: and ordering operator:.

LB Tradeoff 4-Approximation LB Tradeoff 4-Approximation (no intra- resource overlap) Completeness 2-approximation Completeness 2-approximation: Basic idea: given any schedule S and LA, if we change S into LA, each change might improve the performance by at most 1  S has no more then 2 times less completeness then LA. Latency 4-approximation Latency 4-approximation: T k s : k-th first time that OPT didn’t probe, but LB did (and T k f be the last time)…and T k s’, T k f ’ when both did. Best case: LB and OPT act the same  black circles. Worst case (triangles): OPT has: while LB has at most: Thus we get approximation ratio = 4 Whenever the EIs have uniform width W  LB is 2-approximation

LBW Tradeoff 2-Approximation LBW Tradeoff 2-Approximation (no intra- resource overlap) Completeness 2-approximation Completeness 2-approximation: Same as in LB. Optimal Latency Optimal Latency: LB’s greedy “delay traps”: Delay trap Case 1: S S’ j Case 2: S S’ j S

Online policies vs. Optimal Pareto Set

Online policies vs. Optimal Pareto Set: Runtime Scalability Analysis

Online policies: Budget Impact

Online policies: Workload Impact (no intra-resource overlap)

Online policies: Workload Impact (with intra-resource overlap)

Proposed Offline Approximation As A we use Bar-Yehuda et al. algorithm for scheduling split-intervals. C=1  A provides 2k-approx.  we get (2k+2)-approx. C>1  A provides (2k+1)-approx.  we get (2k+3)-approx. Drawbacks: the transformation may be quite expensive. A doesn’t scale (requires LP solution for fractional version of the problem).

Greedy Online Policies Property (no-intra resource overlaps) OrderingPolicy Optimal for simple profilesS-EDF l-competitive whereMRSF Similar to MRSF for problem instances with profilesM-EDF

MRSF: l -Compatitive (case with no intra-resource overlap) “Good guys”“Bad guys” Pick good guys  gain (length( I )) Pick bad guys  gain 1 Ratio = length( I ))  at the worst case every CEI has equal length  comp. ratio:

Online Policies vs. Offline approx. For rank(P)=1 both WIC and EDF are optimal. For any rank(P) the worst case optimal upper bound is OPT rank(1) /rank(k). Simple policies (i.e., WIC,EDF) do not fit into problems with complex profiles here COMP MRSF ≥ COMP off ≥ OPT/2k offline policy doesn’t scale online policies scale quite well