Web switch support for differentiated services

Slides:



Advertisements
Similar presentations
Scheduling in Web Server Clusters CS 260 LECTURE 3 From: IBM Technical Report.
Advertisements

© 2006 Cisco Systems, Inc. All rights reserved. MPLS v2.2—8-1 MPLS TE Overview Introducing the TE Concept.
Consistency and Replication Chapter 7 Part II Replica Management & Consistency Protocols.
CS 443 Advanced OS Fabián E. Bustamante, Spring 2005 Resource Containers: A new Facility for Resource Management in Server Systems G. Banga, P. Druschel,
1 Storage-Aware Caching: Revisiting Caching for Heterogeneous Systems Brian Forney Andrea Arpaci-Dusseau Remzi Arpaci-Dusseau Wisconsin Network Disks University.
CLOUD COMPUTING AN OVERVIEW & QUALITY OF SERVICE Hamzeh Khazaei University of Manitoba Department of Computer Science Jan 28, 2010.
1 Routing and Scheduling in Web Server Clusters. 2 Reference The State of the Art in Locally Distributed Web-server Systems Valeria Cardellini, Emiliano.
A Case for Relative Differentiated Services and the Proportional Differentiation Model Constantinos Dovrolis Parameswaran Ramanathan University of Wisconsin-Madison.
An Adaptable Benchmark for MPFS Performance Testing A Master Thesis Presentation Yubing Wang Advisor: Prof. Mark Claypool.
Fair Scheduling in Web Servers CS 213 Lecture 17 L.N. Bhuyan.
1 Introduction to Load Balancing: l Definition of Distributed systems. Collection of independent loosely coupled computing resources. l Load Balancing.
Differentiated Multimedia Web Services Using Quality Aware Transcoding S. Chandra, C.Schlatter Ellis and A.Vahdat InfoCom 2000, IEEE Journal on Selected.
Adaptive Content Delivery for Scalable Web Servers Authors: Rahul Pradhan and Mark Claypool Presented by: David Finkel Computer Science Department Worcester.
LDU Parametrized Discrete-Time Multivariable MRAC and Application to A Web Cache System Ying Lu, Gang Tao and Tarek Abdelzaher University of Virginia.
12006/9/26 Load Balancing in Dynamic Structured P2P Systems Brighten Godfrey, Karthik Lakshminarayanan, Sonesh Surana, Richard Karp, Ion Stoica INFOCOM.
1 Exploring Data Reliability Tradeoffs in Replicated Storage Systems NetSysLab The University of British Columbia Abdullah Gharaibeh Matei Ripeanu.
Dynamic Load Balancing on Web-server Systems Valeria Cardellini, Michele Colajanni, and Philip S. Yu Presented by Sui-Yu Wang.
Web Server Load Balancing/Scheduling Asima Silva Tim Sutherland.
Computer Science Cataclysm: Policing Extreme Overloads in Internet Applications Bhuvan Urgaonkar and Prashant Shenoy University of Massachusetts.
OPTIMAL SERVER PROVISIONING AND FREQUENCY ADJUSTMENT IN SERVER CLUSTERS Presented by: Xinying Zheng 09/13/ XINYING ZHENG, YU CAI MICHIGAN TECHNOLOGICAL.
Adaptive Overload Control for Busy Internet Servers Matt Welsh and David Culler USENIX Symposium on Internet Technologies and Systems (USITS) 2003 Alex.
Budget-based Control for Interactive Services with Partial Execution 1 Yuxiong He, Zihao Ye, Qiang Fu, Sameh Elnikety Microsoft Research.
Mechanisms for Quality of Service in Web Clusters V. Cardellini, E. Casalicchio, S.Tucci M. Colajanni University of Roma “Tor Vergata” University of Modena.
© 2006 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Injecting Realistic Burstiness to.
DYNAMIC LOAD BALANCING ON WEB-SERVER SYSTEMS by Valeria Cardellini Michele Colajanni Philip S. Yu.
Network Computing Laboratory Load Balancing and Stability Issues in Algorithms for Service Composition Bhaskaran Raman & Randy H.Katz U.C Berkeley INFOCOM.
1 / 21 Providing Differentiated Services from an Internet Server Xiangping Chen and Prasant Mohapatra Dept. of Computer Science and Engineering Michigan.
1 Evaluation of Cooperative Web Caching with Web Polygraph Ping Du and Jaspal Subhlok Department of Computer Science University of Houston presented at.
Architecture for Resource Allocation Services Supporting Interactive Remote Desktop Sessions in Utility Grids Vanish Talwar, HP Labs Bikash Agarwalla,
Network Processing Systems Design
Towards a High Performance Extensible Grid Architecture Klaus Krauter Muthucumaru Maheswaran {krauter,
Md Baitul Al Sadi, Isaac J. Cushman, Lei Chen, Rami J. Haddad
GreenCloud: A Packet-level Simulator of Energy-aware Cloud Computing Data Centers Dzmitry Kliazovich, Pascal Bouvry, Yury Audzevich, and Samee Ullah Khan.
Instructor Materials Chapter 6: Quality of Service
Web Server Load Balancing/Scheduling
CIIT-Human Computer Interaction-CSC456-Fall-2015-Mr
Web Server Load Balancing/Scheduling
Introduction to Load Balancing:
SCTP Handoff for Cluster Servers
Hybrid Cloud Architecture for Software-as-a-Service Provider to Achieve Higher Privacy and Decrease Securiity Concerns about Cloud Computing P. Reinhold.
Federated IdM Across Heterogeneous Clouding Environment
Establishing End-to-End Guaranteed Bandwidth Network Paths Across Multiple Administrative Domains The DOE-funded TeraPaths project at Brookhaven National.
Analyzing Security and Energy Tradeoffs in Autonomic Capacity Management Wei Wu.
Distributed Multimedia Systems
July 2009 doc.: IEEE /xxxxr0 July 2009
Measuring Service in Multi-Class Networks
Auburn University COMP7500 Advanced Operating Systems I/O-Aware Load Balancing Techniques (2) Dr. Xiao Qin Auburn University.
Implementing a Load-balancing Web Server Using Red Hat Cluster Suite
Transparent Adaptive Resource Management for Middleware Systems
© 2008 Cisco Systems, Inc. All rights reserved.Cisco ConfidentialPresentation_ID 1 Chapter 6: Quality of Service Connecting Networks.
Quality of Service in the Internet
EEC-484/584 Computer Networks
Integrated Resource Management for Cluster-based Internet Services
TimeTrader: Exploiting Latency Tail to Save Datacenter Energy for Online Search Balajee Vamanan, Hamza Bin Sohail, Jahangir Hasan, and T. N. Vijaykumar.
Ying Qiao Carleton University Project Presentation at the class:
Admission Control and Request Scheduling in E-Commerce Web Sites
Load Balancing in Distributed Systems
QuaSAQ: Enabling End-to-End QoS for Distributed Multimedia Databases
Multiple-resource Request Scheduling. for Differentiated QoS
Buffer Management for Shared-Memory ATM Switches
AAA: A Survey and a Policy- Based Architecture and Framework
Hawk: Hybrid Datacenter Scheduling
Route Metric Proposal Date: Authors: July 2007 Month Year
EE 122: Lecture 22 (Overlay Networks)
Host and Small Network Relaying Howard C. Berkowitz
QoS routing Finding a path that can satisfy the QoS requirement of a connection. Achieving high resource utilization.
Modeling and Evaluating Variable Bit rate Video Steaming for ax
Requirements of Computing in Network
L. Glimcher, R. Jin, G. Agrawal Presented by: Leo Glimcher
Presentation transcript:

Web switch support for differentiated services PAWS 2001 Web switch support for differentiated services V. Cardellini, E. Casalicchio M. Colajanni, M. Mambelli University of Roma “Tor Vergata” University of Modena Speaker: Michele Colajanni colajanni@unimo.it http://www.ce.uniroma2.it/people/colajanni Additional Info: http://www.ce.uniroma2.it

Outline Motivations Cluster-based Web server systems V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Outline Motivations Quality of Service & Quality of Web Services Cluster-based Web server systems Policies for the Quality of Web Services Results System and workload model Performance metrics Simulation results Summary

Why QoS in Web services? The second generation of Web sites V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Why QoS in Web services? The second generation of Web sites communication channel for critical information fundamental technology for information systems of the most advanced companies and organizations business-oriented media The new Web requires differentiation of users and services supports to heterogeneous applications and user expectation differentiated pricing for content hosting and service providing

Server-side proposals for QoWS V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Server-side proposals for QoWS QoS Network side Server side Single server Multiple servers Operating system Web server Single Web site Web hosting

A Web cluster architecture V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli A Web cluster architecture WAN Application servers . Client requests Layer-7 Web switch . LAN Web servers One-way vs. two-way architectures Layer-4 vs. Layer-7 Web switches (content-blind vs. content-aware switches)

Quality of Web Services (QoWS) V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Quality of Web Services (QoWS) High performance systems  Systems for Quality of Service QoS principles and mechanisms have been deeply investigated in the computer network area, but QoS principles are not immediately applicable to the server side of the Web system Network QoS and server QoWS principles must be combined to provide a peer-to-peer QoS for Web services The focus of this talk will be on QoWS for Web clusters

From QoS to QoWS in Web clusters V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli From QoS to QoWS in Web clusters To define QoWS principles we need to find out feasible mechanisms to achieve QoWS Web cluster components able to implement QoWS principles and mechanisms Our idea: start from main QoS principles classification of services performance isolation high resource utilization request admission declaration access control QoWS

QoWS principles Classification (at Web switch) Performance isolation V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli QoWS principles Classification (at Web switch) users and services identification users and services classification Performance isolation queuing scheduling policies (at Web server) resource partitioning (at the Web server for fine-grained level, at the Web Switch for coarse-grained level) High resource utilization (at Web switch/server) dynamic resource partitioning Request admission (at Web switch/server) estimation of resource demand (at Web switch) access control mechanism (at Web switch/server)

Web switch support for QoWS (Classification and request admission) V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Web switch support for QoWS (Classification and request admission) Application servers WAN Dropped requests Admitted requests Client requests Layer-7 Web switch . . LAN Web servers

V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Web switch support for QoWS (Performance isolation and high resource utilization) High users WAN Application servers Dropped requests Admitted requests Client requests Layer-7 Web switch . LAN Low users Web servers

V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Policies for QoWS (1) Policies classified on the basis of the increasing number of QoWS principles they satisfy classification and request admission plus performance isolation plus high resource utilization Classification and request admission SwitchAdm service denied to the lower class of users by the Web switch rejection mechanism based on cluster load (load threshold)

Policies for QoWS (2) Plus performance isolation V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Policies for QoWS (2) Plus performance isolation StaticPart server nodes are statically partitioned in High Set (HS) and Low Set (LS) different classes of requests are assigned to different server sets Plus high resource utilization Demand-Driven Service Differentiation (DDSD) [Zhu01] dynamic partitioning based on the resolution of a constrained optimization problem the resolution provides the number of servers assigned to each service class and the admission rate for each service class

Policies for QoWS (3) Plus high resource utilization DynamicPart V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Policies for QoWS (3) Plus high resource utilization DynamicPart

Policies and QoWS principles V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Policies and QoWS principles SwitchAdm StaticPart DynamicPart DSSD Yes Yes No No Yes Yes Yes No Yes Yes Yes Yes Classification Request Performance High resource admission isolation utilization

System and workload model V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli System and workload model Parameter Value (default) Number of servers Disk Memory transfer rate Intra-server bandwidth 10-20 (10) 20 MBps, 7200 RPM, 0.05 msec c.d., 9 msec s.t. 100 MBps 100 Mbps switched LAN Session arrival rate Requests per session User think time Objects per request HTML size - body - tail Embedded object size 100 - 300 clients per second Inverse Gaussian ( = 3.86,  = 9.46 ) Pareto ( = 1.4, k = 1) Pareto ( = 1.33, k = 2) Lognormal ( = 7.630,  = 1.001 ) Pareto ( = 1, k = 10240) Lognormal ( = 8.215,  = 1.46 ) Types of service Dynamic requests (80% of static requests, 20% of dynamic requests) Hyper-exponential (high, medium, low intensive)

Performance metrics 95-percentile of latency time V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Performance metrics 95-percentile of latency time latency time: completion time of a Web page request at the Web cluster side X-percentile is used to define the Service Level Agreement (SLA) for predictive services SLA on dynamic requests for high class requests set to 4 seconds Percentage of dropped requests percentage of users that perceive a deny of service Stretch factor ratio between latency time and service time

Sensitivity of latency time to client arrival rate V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Simulation results (1) Sensitivity of latency time to client arrival rate High class Low class

Sensitivity of stretch factor to client arrival rate V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Simulation results (2) Sensitivity of stretch factor to client arrival rate High class Low class

Sensitivity to the variable workload composition V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Simulation results (3) Sensitivity to the variable workload composition Latency time Dropped requests

Simulation results (4) Latency time Dropped requests V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Simulation results (4) Latency time Dropped requests Sensitivity to the number of Web servers (constant ratio between offered load and number of Web servers)

V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Conclusions Basic requirements for QoWS are satisfied by policies that integrate main mechanisms for network-QoS into the Web switch: request classification and admission control high resource utilization dynamic server partitioning Policies that satisfy all QoWS principles are able to meet SLA for different load and system conditions The violation of even a single QoWS principle prevents to meet the SLA

Work in progress Address scalability issue for the Web switch V. Cardellini,E. Casalicchio, M. Colajanni, M. Mambelli Work in progress Address scalability issue for the Web switch New policies for dynamic request partitioning Prototype implementation Additional information: http://www.ce.uniroma2.it