Declaration of Rachel J. Watters on Authentication of Publication
`
`I, Rachel J. Watters, am a librarian, and the Head of Resource Sharing for the
`
`General Library System, Memorial Library, located at 728 State Street, Madison,
`
`Wisconsin, 53706. Part of my job responsibilities include oversight of Wisconsin
`
`TechSearch (“WTS”), an interlibrary loan departmentat the University of Wisconsin-
`
`Madison. I have workedasalibrarian at the University of Wisconsin library system
`
`since 1998,starting as a graduate student employee in the Kurt F. Wendt Engineering
`
`Library and WTS,then asa librarian in Interlibrary Loan at Memorial Library. I began
`
`professional employment at WTSin 2002 and became WTSDirector in 2011. In 2019,
`
`I became Head of Resource Sharing for UW-Madison’s General Library System. I have
`
`a master’s degree in Library and Information Studies from the University of Wisconsin-
`
`Madison. Through the course of my studies and employment, I have becomewell
`
`informed aboutthe operations of the University of Wisconsin library system, which
`
`follows standard library practices.
`
`This Declaration relates to the dates of receipt and availability of the following:
`
`Osman,S., Subhraveti, D., Su, G., and Nieh, J. (2000). The
`design and implementation of Zap: a system for migrating
`computing environments. Proceedings ofthe Fifth Symposium on
`Operating Systems Design and Implementation (OSDI ’02),
`December 9-11, 2002, Boston, Massachusetts, USA (Operating
`Systems Review, Special Issue, Winter 2002), 361-376.
`
`Standard operating procedures for materials at the University of Wisconsin-
`
`Madison Libraries. When an issue wasreceived by the Library, it would be checked in,
`
`Amazon Ex. 1015
`IPR Petition - USP 7,519,814
`
`Amazon Ex. 1015
`IPR Petition - USP 7,519,814
`
`

`

`Declaration of Rachel J. Watters on Authentication of Publication
`
`stamped with the date of receipt, added to library holdings records, and made available
`
`to readers as soonafterits arrival as possible. The procedure normally took a few days
`
`or at most 2 to 3 weeks.
`
`Exhibit A to this Declaration is a true and accurate copyofthetitle page with
`
`library date stamp of Proceedings ofthe Fifth Symposium on Operating Systems Design
`
`and Implementation (OSDI ’02), December 9-11, 2002, Boston, Massachusetts, USA
`
`(Operating Systems Review, Special Issue, Winter 2002) (2002), from the University of
`
`‘ Wisconsin-Madison Library collection. Exhibit A also includes an excerpt of pages 361
`
`to 376 ofthat issue, showingthearticle entitled The design and implementation ofZap:
`
`a systemfor migrating computing environments (2002). Based onthis information, the
`
`date stamp’on the journalissuetitle page indicates The design and implementation of
`Zap: a systemfor migrating computing environments (2002) was received by the Kurt
`F. Wendt Library, University of Wisconsin-Madison on April 14, 2003.
`
`Based on the information in Exhibit, A, it is clear that the issue was received by
`
`the library on or before April 14, 2003, catalogued and available to library patrons
`
`within a few daysor at most 2 to 3 weeksafter April 14, 2003.
`
`Membersofthe interested public could locate the Proceedings ofthe Fifth
`
`Symposium on Operating Systems Design and Implementation (OSDI ’02), December 9-
`
`11, 2002, Boston, Massachusetts, USA(Operating Systems Review, Special Issue,
`
`Winter 2002) (2002) publication after it was cataloged by searching the public library
`
`2
`
`

`

`Declaration of Rachel J. Watters on Authentication of Publication
`
`catalog or requesting a search through WTS. Thesearch could be donebytitle and/or
`
`subject key words. Membersofthe interested public could access the publication by
`
`locating it on the library’s shelves or requesting it from WTS.
`
`I declare that all statements made herein of my own knowledgearetrue andthat
`
`all statements made on information and belief are believed to be true; and further that
`
`these statements were made with the knowledgethatwillful false statements and the like
`
`so made are punishable by fine or imprisonment, or both, under Section 1001 of Title 18
`
`of the United States Code.
`
`Date: August 16, 2024
`
`Memorial Library
`728 State Street
`Madison, Wisconsin 53706
`
`Rachel J. Watters
`Head of Resource Sharing
`
`

`

`
`
`
`
`
`EXHIBIT A
`EXHIBIT A
`
`

`

`A Publication of the
`Association for Computing Machinery
`Special Interest Group on Operating Systems
`
`OSRSpecial Issue — Winter 2002
`
`OPERATING
`SYSTEMS
`REVIEW
`
`and Implementation (OSDI'02)
`
`Proceedings of the
`
`Fifth ACM Symposium on
`Operating Systems Design
`
`

`

`OPERATING SYSTEMS REVIEW
`A Publication of the ACM Special Interest Group on Operating Systems
`
`
`V. Chairperson
`Valerie Issarny
`INRIA Ur Rocquencourt
`Domaine de Voluceau
`Rocquencourt - BP 105 Cedex
`78153 Le Chesnay France
`+33-1-39-63-57-17
`(valerie.issarny @inria.fr)
`
`Chairperson
`David Kotz
`Dept. of Computer Science
`Dartmouth College
`6211 Sudikoff Laboratory
`Hanover, NH 03755-3510 USA
`+1-603-646-1439
`(dfk@cs.dartmouth.edu)
`Newsletter Editor
`William M. Waite
`Dept. of Electrical and
`Computer Engineering
`University of Colorado
`Boulder, CO 80309-0425 USA
`+1-303-492-7204
`(William.Waite@Colorado.edu)
`
`Secretary-Treasurer
`Michael Dahlin
`Dept. of Computer Science
`Campus Mail Code: C0500
`University of Texas
`Austin, TX 78712 USA
`+1-512-471-9549
`(dahlin@cs.utexas.edu)
`Information Director
`Amin Vahdat
`D308 LSRC, Box 90129
`Department of Computer Science
`Duke University
`Durham, NC 27708 USA
`+1-919-660-6566
`(vahdat@cs.duke.edu)
`
`dents cost is included in $8.00); the non-member annualsubscription is $30.00.
`
`OPERATING SYSTEMS REVIEWis an informal publication of the ACM Special Interest Group on
`Operating Systems (SIGOPS), whose scope ofinterest includes: Computer operating systems and archi-
`tecture for multiprogramming, multiprocessing, and time sharing; resource management; evaluation and
`simulation; reliability, integrity, and security of data, communications among computing processors; and
`computer system modeling analysis.
`Membership in SIGOPS(at $15 per annum) is open to ACM members and associate members, and stu-
`dent members(at $8 per annum). Non-members of SIGOPS may subscribe to OPERATING SYSTEMS
`REVIEWat $30 per year. All SIGOPS membersreceive OPERATING SYSTEMS REVIEW.
`SIGOPS membership application formsare available from ACM Headquarters (or see back coverof this
`issue), 1515 Broadway, 17th Floor, New York, New York 10036, telephone (212) 869-7440. Changes of
`address and other matters pertaining to the SIGOPSmailing list should be directed to ACM Headquarters,
`not to any of the SIGOPSofficers.
`Contributions to OSRare sent to the editor and should be single spaced andprinted on oneside ofthe
`paper only. Text and diagrams must be contained within a 6.5 inch wide by 8.5 inch high (165mmby
`216mm) area with at least one inch (25mm)top and side margins, 1.5 inch (38mm) bottom margin, and
`be camera-ready(i.e., be an original having sufficient contrast and density for reproduction). Allletters to
`the editor will be considered for publication unless accompanied by a requestto the contrary. Except for
`editorial items, all sources of material appearing inOPERATING SYSTEMS REVIEW will be clearly
`identified.
`Items attributed to individuals will ordinarily be interpreted as personalrather than organiza-
`tional opinions. The technical material in this newsletter is not refereed. Deadlines for submitting
`material are the 15th of February, May, August, and November. The SIGOPSconference proceed-
`ings will be mailed as the December issue in odd numbered calendar years, and the ASPLOS
`conference proceedingswill be mailed in even numbered calendaryears.
`OPERATING SYSTEMS REVIEW (ISSN 0163-5980) is published five times a year (January, April,
`July, October and December) by the Association for Computing Machinery Inc., 1515 Broadway, New
`York, NY 10036. Periodicals postage paid at New York, NY 10001, and at additional mailing offices.
`POSTMASTER:Send address changes to OPERATING SYSTEMS REVIEW, ACM,1515 Broadway,
`New York, NY 10036.
`Subscriptions: Annual subscription cost of $12.53 is included in the member dues of $15.00 (for stu-
`
`

`

`e
`e
`USENIXAssociation
`
`"
`
`on Wen,
`oeURay
`
`Proceedingsof the
`
`Fifth Symposium on Operating Systems
`
`Design and Implementation
`
`(OSDI 702)
`
`Co-sponsored by ACM SIGOPS
`
`In Cooperation with IEEE TCOS
`
`December9-11, 2002
`Boston, Massachusetts, USA
`
`
`
`
`

`

`
`
`Symposium Organizers
`Program Chair
`David Culler, University of California, Berkeley
`Peter Druschel, Rice University
`Program Committee
`Eric Brewer, University of California, Berkeley
`Miguel Castro, Microsoft Research
`Carla Ellis, Duke University
`Dawson Engler, Stanford University
`Deborah Estrin, University of California, Los
`Angeles/ISI
`Greg Ganger, Carnegie Mellon University
`
`Jim Gray, Microsoft Research
`Jay Lepreau, University of Utah
`Robert Morris, Massachusetts Institute of Technology
`Timothy Roscoe, Intel Research Lab at Berkeley
`Chandu Thekkath, Microsoft Research
`David Wetherall, University of Washington
`Steering Committee
`Michael B. Jones, Microsoft Research
`Frans Kaashoek, Massachusetts Institute of Technology:
`
`The USENIX Association Staff
`
`External Reviewers
`Atul Adya
`Sameer Ajmani
`David Andersen
`Darrell Anderson
`Andrea Arpaci-Dusseau
`Remzi Arpaci-Dusseau
`Tuomas Aura
`Godmar Back
`Paul Barham
`Bobby Bhattacharjee
`David Black
`Richard Black
`Trevor Blackwell
`Nikita Borisov
`Thomas Bressoud
`Nevil Brownlee
`Philip Buonadonna
`Andrew Campbell
`John Carter
`Alberto Cerpa
`Surendar Chandra
`Jeff Chase
`Benjie Chen
`Mike Chen
`Peter Chen
`Andy Chou
`Brent Chun
`Patrick Crowley
`Mike Dahlin
`David Dill
`John Douceur
`Eric Eide
`Kevin Eustice
`Ted Faber
`Mike Feeley
`Jason Flinn
`Rick Floyd
`Cédric Fournet
`
`Michael Franklin
`Ayalvadi Ganesh
`Johannes Gehrke
`Garth Gibson
`Daniel Giffin
`Thomer M. Gil
`Steve Gribble
`Dirk Grunwald
`Richard Guy
`Steven Hand
`Timothy Harris
`Chris Hawblitzel
`Mike Hibler
`Michael Hicks
`Wilson Hsieh
`Galen Hunt
`Sitaram Iyer
`Trent Jaeger
`Paul Jardetzky
`Brad Karp
`Kimberly Keeton
`Anne-Marie Kermarrec
`Eddie Kohler
`David Kotz
`Michael Kozuch
`Orran Krieger
`Lakshman Krishnamurthy
`John Kubiatowicz
`Eren Kursun
`Alvin Lebeck
`Philip Levis
`Hank Levy
`Jinyang Li
`Kai Li
`Darrell Long
`Ewing Lusk
`Mesaac Makpangou
`Rob Malan
`
`Laurent Massoulié
`Roy Maxion
`David Maziéres
`Ethan Miller
`Jeffrey Mogul
`Richard Mortier
`David Mosberger
`Todd Mowry
`Gilles Muller
`Andrew Myers
`David Nagle
`Erich Nahum
`Juan Navarro
`Silvia Nittel
`Brian Noble
`Chris Olston
`Vivek Pai
`Parveen Patel
`Larry Peterson
`David Petrou
`Tan Pratt
`Calton Pu
`Vijay Raghunathan
`Sylvia Ratnasamy
`John Regehr
`Alastair Reid
`Peter Reiher
`Mike Reiter
`Erik Riedel
`Rodrigo Rodrigues
`Michael Roe
`Mendel Rosenblum
`Antony Rowstron
`Constantine Sapuntzakis
`Mahadev Satyanarayanan
`Dan Scales
`MargoSeltzer
`Srinivasan Seshan
`
`Jonathan Shapiro
`Marc Shapiro
`Dan Simon
`Emin Gun Sirer
`Stephen Smalley
`Alex Snoeren
`Ankur Srivastava
`Dave Stewart
`Doug Terry
`Amin Vahdat
`Alistair Veitch
`Geoff Voelker
`Werner Vogels
`James von Behren
`David Wagner
`Carl Waldspurger
`Jonathan Walpole
`Randolph Wang
`Matt Welsh
`Brian White
`Ted Wobber
`Alec Wolman
`Fan Ye
`Erez Zadok
`Ellen Zegura
`Beichuan Zhang
`Ben Zhao
`Lidong Zhou
`Yuanyuan Zhou
`Ben Zorn
`
`
`
`

`

`OSDI ’02: 5th Symposium on
`Operating Systems Design and Implementation
`
`December 9-11, 2002
`Boston, Massachusetts, USA
`
`Message from the Program Chair ............0..00c cece ence cece eteueneeuueesuereuveteennnes vii
`
`Index of Authors 2.0.0... icc cece ce eee bebe eee eaeeebetetneeetbbeebbvecneennnbs ix
`
`Monday, December 9
`
`Decentralized Storage Systems
`Session Chair: Greg Ganger, Carnegie Mellon University
`
`FARSITE:Federated, Available, and Reliable Storage for an Incompletely Trusted Environment .............. 1
`Atul Adya, William J. Bolosky, Miguel Castro, Gerald Cermak, Ronnie Chaiken, John R. Douceur, Jon Howell,
`Jacob R. Lorch, Marvin Theimer, and Roger P. Wattenhofer, Microsoft Research
`
`Taming Aggressive Replication in the Pangaea Wide-Area File System ............0.0. 000 ccc ecueueeesees 15
`Yasushi Saito, Christos Karamanolis, Magnus Karlsson, and Mallik Mahalingam, HP Labs
`
`Ivy: A Read/Write Peer-to-Peer File System ..... 2.0... ccc ccc cece cence cece ee neeeteeennnennaees 31
`Athicha Muthitacharoen, Robert Morris, Thomer M. Gil, and Benjie Chen, Massachusetts Institute of Technology
`
`Robustness
`Session Chair: Miguel Castro, Microsoft Research
`
`Defensive Programming: Using an Annotation Toolkit to Build DoS-Resistant Software ............6-eee005 45
`Xiaohu Qie, Ruoming Pang, and Larry Peterson, Princeton University
`
`Using Model Checking to Debug Device Firmware ........ 00.20 cece cee eee ence tu eeeeuceerueeenunees 61
`Sanjeev Kumarand KaiLi, Princeton University
`
`CMC: A Pragmatic Approach to Model Checking Real Code ..........00..0 0.0 ccc cce eee eeceeeeueeeees 75
`Madanlal Musuvathi, David Y. W. Park, Andy Chou, Dawson R. Engler, and David L. Dill, Stanford University
`
`Kernels
`Session Chair: Carla Ellis, Duke University
`
`Practical, Transparent Operating System Support for Superpages ..........0..0 0.00 cece ee cee eueeeuaes 89
`Juan Navarro, Rice University and Universidad Catélica de Chile; Sitaram Iyer, Peter Druschel, and Alan Cox,
`Rice University
`
`Vertigo: Automatic Performance-Setting for Linux... 0. cence eee teens eee e eee eee 105
`Krisztidn Flautner, ARM Limited; Trevor Mudge, University ofMichigan
`
`Cooperative I/O: A Novel I/O Semantics for Energy-Aware Applications .........0.0c0.eeuceseeeesusees 117
`Andreas Weissel, Bjérn Beutel, and Frank Bellosa, University of Erlangen
`
`

`

`Physical Interface
`Session Chair: Jay Lepreau, University of Utah
`
`TAG: A Tiny AGgregation Service for Ad-Hoc Sensor Networks ....... 000.0 c ee cece eects 131
`Samuel Madden, Michael J. Franklin, and Joseph M. Hellerstein, University of California, Berkeley; Wei Hong,
`Intel Research, Berkeley
`
`Fine-Grained Network Time Synchronization Using Reference Broadcasts ........... 00 ee cece cee eens 147
`Jeremy Elson, Lewis Girod, and Deborah Estrin, University of California, Los Angeles
`
`Supporting Time-Sensitive Applications on a Commodity OS ......... 0... c eee eee eee cette eee 165
`Ashvin Goel, Luca Abeni, Charles Krasic, Jim Snow, and Jonathan Walpole, Oregon Graduate Institute
`
`Tuesday, December 10
`
`Virtual Machines
`Session Chair: Dawson Engler, Stanford University
`
`Memory Resource Management in VMware ESX Server ....... 0... cece teen eee eee e eens 181
`Carl A. Waldspurger, VMware Inc.
`
`Scale and Performancein the Denali Isolation Kernel .......... 0.000. eee eee eens 195
`Andrew Whitaker, Marianne Shaw, and Steven D. Gribble, University of Washington
`
`ReVirt: Enabling Intrusion Analysis Through Virtual-Machine Logging and Replay ..............-..00005 211
`George W. Dunlap, Samuel T. King, Sukru Cinar, Murtaza A. Basrai, and Peter M. Chen, University ofMichigan
`
`Cluster Resource Management
`Session Chair: Eric Brewer, University of California, Berkeley
`
`Integrated Resource Management for Cluster-based Internet Services ......... 0 0c cece eee eens 225
`Kai Shen, University of Rochester; Hong Tang, University of California, Santa Barbara; Tao Yang, University of
`California, Santa Barbara, and Ask Jeeves, Inc.; Lingkun Chu, University of California, Santa Barbara
`
`Resource Overbooking and Application Profiling in Shared Hosting Platforms .................0.. 0 e ee 239
`Bhuvan Urgaonkar and Prashant Shenoy, University ofMassachusetts; Timothy Roscoe, Intel Research, Berkeley
`
`AnIntegrated Experimental Environment for Distributed Systems and Networks .............. Se 255
`Brian White, Jay Lepreau, Leigh Stoller, Robert Ricci, Shashi Guruprasad, Mac Newbold, Mike Hibler, Chad
`Barb, and Abhijeet Joglekar, University of Utah
`
`Peer-to-Peer Infrastructure
`Session Chair: Robert Morris, Massachusetts Institute of Technology
`
`Scalability and Accuracy in a Large-Scale Network Emulator ........ 00.00 e cece ene 271
`Amin Vahdat, Ken Yocum, Kevin Walsh, Priya Mahadevan, Dejan Kostic, Jeff Chase, and David Becker, Duke
`University
`
`Pastiche: Making Backup Cheap and Easy ......... 0... c cece cece eee eee eee ee teen een teenie renee 285
`Landon P. Cox, Christopher D. Murray, and Brian D. Noble, University ofMichigan
`
`Secure Routing for Structured Peer-to-Peer Overlay Networks .. 0.0.0.0. 0.0 ccc cee eee enee299
`Miguel Castro, Microsoft Research; Peter Druschel, Rice University; Ayalvadi Ganesh and Antony Rowstron,
`Microsoft Research; Dan S. Wallach, Rice University
`
`

`

`
`
`Wednesday, December 11
`
`Network Behavior
`Session Chair: David Wetherall, University of Washington
`
`An Analysis of Internet Content Delivery Systems ........0000000 0 00sec cece cece eee een e evens 315
`Stefan Saroiu, Krishna P. Gummadi, Richard J. Dunn, Steven D. Gribble, and Henry M. Levy, University of
`Washington
`
`TCP Nice: A Mechanism for Background Transfers ........0.0.0 0.00.0 cc cece eee een ueeduvenveaueuvens 329
`Arun Venkataramani, Ravi Kokku, and Mike Dahlin, University of Texas at Austin
`
`The Effectiveness of Request Redirection on CDN Robustness ........00.0 ccc eeeneeeseueeuueeceeees 345
`Limin Wang, Vivek Pai, and Larry Peterson, Princeton University
`
`Migration
`Session Chair: Timothy Roscoe, Intel Research, Berkeley
`
`The Design and Implementation of Zap: A System for Migrating Computing Environments ..............-- 361
`Steven Osman, Dinesh Subhraveti, Gong Su, and Jason Nieh, Columbia University
`
`Optimizing the Migration of Virtual Computers .. 2.0.22... 000000e eee e eee eee ens 377
`Constantine P. Sapuntzakis, Ramesh Chandra, Ben Pfaff, Jim Chow, Monica S. Lam, and Mendel Rosenblum,
`Stanford University
`
`Luna: A Flexible Java Protection System .. 0.2.0... 0. ccc cee nee e esse ee nnaeeneeennns 391
`Chris Hawblitzel, Dartmouth College; Thorsten von Eicken, Expertcity, Inc.
`
`
`
`

`

`Index of Authors
`
`Abeni, Luca ...... a tsisatept ngs ovinwiepieaila on. 165
`Adya, Atul 2.2.20... ec ec eee eee 1
`Barb, Chad ...... 0... ccc ccc cece cence eens255
`Basrai, Murtaza A. .......c ccc eee eee teens 211
`Becker, David 22... .c cee e ieee ee eee e eee eee 271
`Bellosa, Frank ... 0.0.2.0. 0000 ccc cee eee 117
`Beutel, Bj6mn ... eee eee {17
`Bolosky, William J. 2.2... eee eee 1
`Castro, Miguel ...... 22... 2c ec eee eee 1, 299
`Cermak, Gerald... ccc cae eee eee ee eeee 1
`Chaiken, Ronnie .
`.
`. eiaa vaacecic eases eenee ee... 1
`Chandra, Ramesh ....... 0.00000 e eee ee eeeee 377
`Chase, Jeff... 2... scuiad cateaeawies Deedee27)
`Chen, Benjie .............-.0005- 1 SCRE | 31
`Chen, Peter M. ........ 0.2.02 cue cence eens 211
`Chou, Andy ... 2.00.0... . 2c eee eee eee eee75
`Chow, Jim . ... 2... .sarisace siatsarcaraveracare ane ierenececenas 377
`Chu, Lingkun 1.2.2.0...225
`Cinar, Sukm ..... . .aness be eee we ewes 211
`Cox, Alan .... 2... Jase Saieseta aia wiele awitemigure 2 89
`Cox, Landon P. ....... 0... ec ees 285
`Dahlin, Mike... 2... ce eee 329
`Dill, David L. 0... ee ee ee eee 75
`Douceur, John R. 2.1... ee eee eee l
`Druschel, Peter .... 0.0. cece enact een eee 89, 299
`Dunlap, George W. ..... 00.000 c eee eee eee eee211
`Dunn,Richard J... 2.0... 00.02 ee eee 315
`Elson, Jeremy ..... 0... 0c eee cee ee eee eens 147
`Engler, Dawson R. «1... eee eee ees15
`Estrin, Deborah .. 2.0... 602s ee ee eee eee 147
`Flautner, Krisztidn 2.0.0 ..05 000 cece ccc eee 105
`Franklin, Michael J... 2.0... cc eee ee eee 131
`Ganesh, Ayalvadi .........0 50: e ences299
`Gil, Thomer M... 2... eee 31
`Girod, Lewis .........--00 cece ee eee eee ee eens 147
`Goel, ASHViN 0.0.0... ccc ee eee 165
`Gribble, Steven D. ... 2.2... eee 195, 315
`Gummadi, Krishna P. .......... 000.00 c eee eee 315
`Guruprasad, Shashi... ......0.. 060 ce cee euee255
`Hawblitzel, Chris ... 2... 200.0200 c ee eee eae 391
`Hellerstein, Joseph M. 2... 2.00. ccc eee 131
`Hibler, Mike ..... .. Seeees dee Ove SeeeSies Peed 255
`Hong, Wei... 0... eee ee eee ees 131
`Howell, Jon... 0.0.6.0. eee eeJoven eee eee e ee 1
`Tyer, Sitaram ..... 20... eee aie cesta teceeatca.atcare Ecaveconk89
`Joglekar, Abhijeet... 2. ccc eee eee ees255
`Karamanolis, Christos ......0.. 0000 cece usecase 15
`Karlsson, Magnus .......00.0 00 ccc cece eee ees 15
`King, Samuel T.
`.. 2.0.2 cc eee211
`Kokku, Ravi 2... 0... cee eee eee eee ee 329
`Kostié, Dejan «2... ee eee eee271
`Krasic, Charles 2... 0c cece e cee eee eens 165
`Kumar, Sanjeev... cece ee eee ees61
`Lam, Monica S. 2... cece ee eee 377
`
`Lepreau, Jay .......... Sie Same ae255
`Levy, Henry M.
`.. 2.6 ce ee eee 315
`Li, Kai. cc ee ene eens 61
`Lorch, JacobR........ 0.06.45 eee eee eee 1
`Madden, Samuel .... 2.0... .0. 00 ccc ee ee ee eee 131
`Mahadevan,Priya ............00 002s eee ee eeee271
`Mahalingam, Mallik .............0....0..000.. 15
`Morris, Robert
`........0 0.025520 e eee eee eee 31
`Mudge, Trevor .... 22... ccc ee ee eee eee 105
`Murray, Christopher D. «1... 0. eee eee eee285
`Musuvathi, Madanlal .......... 00.0 c ce eee eee75
`Muthitacharoen, Athicha ......0. 00.0 cece ee eee 31
`Navarro, Juan .. 2... eee eee 89
`Newbold, Mac ........ 0.000 c cece ence eee eee255
`Nieh, Jason 2.2.2... eee eee ee tees 361
`Noble, Brian D. 2... 2... ccc eee 285
`Osman, Steven .... 2.0... e eee eee eee 361
`Pai, Vivek 2.0.0.0... ccc cee eens 345
`Pang, Ruoming ....... 2.22.00 e see eee eee eens45
`Park, David Y. W.
`.. 0. eee eee eee tee eee75
`Peterson, Larry ......... 00.2000 c esc ueuene45, 345
`Pfaff, Ben... 2... te ees 377
`Qie, Xiaohu ... 0... eee45
`Ricci, Robert 20... 0.0. eee eee eee 255
`Roscoe, Timothy 2.0.0... ccc cee eee ee eee239
`Rosenblum, Mendel .........2..00000e cee eee 377
`Rowstron, Antony .......... 2.250000 e eee euee299
`Saito, Yasushi ...........0. 00200: e see e een eeee 15
`Sapuntzakis, Constantine P. .. 2.2.2... eee eee 377
`Saroiu, Stefan ......0. 0... cee eee ees 315
`Shaw, Marianne ....... 0.2.00 cee eee eee eee eee 195
`Shen, Kai ............ . séa@menesaanmeces esrecen - 225
`Shenoy, Prashant ..... 02... 55 cc cee cece eens239
`Snow, Jim ............. see beeutea te Haier - 165
`Stoller, Leigh ... 00... 0.20 ccc cee eee255
`Su, Gong .............. boobies tees Bea wane - 361
`Subhraveti, Dinesh ............00 cee eee eens 361
`Tang, Hong .... 1.2... eet eee eee225
`Theimer, Marvin ... 2.0... 20000 -e eee eee eee eee ]
`Urgaonkar, Bhuvan 0.2... cece eee eee239
`Vahdat, AMIN ......... . wasers eaters een - 271
`Venkataramani, Arun .......... SMNeTeaTINN BN 329
`von Eicken, Thorsten .........-.0.2000000008- 391
`Waldspurger, Carl A. 1... 06. c cece ee 181
`Wallach, Dan S.
`... 2... ce es299
`Walpole, Jonathan... 1... cee eee eee ee eee 165
`Walsh, Kevin 2.0.0... 0... cece cece eee eee 271
`Wang, Limin ..... 0... ccc ee ee eee eee 345
`Wattenhofer, Roger P. .. 0.22.22 eee eee 1
`Weissel, Andreas .......... 002200020 eeu eeueees 117
`Whitaker, Andrew ... 0... 0.00 ccc ee eee ee 195
`White, Brian ..... 0.0... eee ee eee 255
`Yang, TaO 0... eee ete225
`Yocum, Ken ........- cc eee eee eee 271
`
`reezas,aeg+e
`
`

`

`
`
`oe
`
`The Design and Implementation of Zap:
`A System for Migrating Computing Environments
`
`Steven Osman, Dinesh Subhraveti, Gong Su, and Jason Nieh
`Department of Computer Science
`Columbia University
`{sto8, dinesh, gongsu, nieh}@cs.columbia.edu
`
`Abstract
`We have created Zap, a novel system for transparent migration of legacy and networked applications. Zap provides a
`thin virtualization layer on top of the operating system that introduces pods, which are groupsof processes that are pro-
`vided a consistent, virtualized view of the system. This decouples processes in pods from dependencies to the host oper-
`ating system and otherprocesses on the system. By integrating Zap virtualization with a checkpoint-restart mechanism,
`Zap can migrate a pod of processes as a unit among machines running independent operating systems without leaving
`behind any residualstate after migration. We have implemented a Zap prototype in Linux that supports transparent mi-
`gration of unmodified applications without any kernel modifications. We demonstrate that our Linux Zap prototype can
`provide general-purpose process migration functionality with low overhead. Our experimental results for migrating
`pods used for running a standard user’s X windows desktop computing environment and for running an Apache web
`server show that these kinds of pods can be migrated with subsecond checkpoint andrestart latencies.
`
`1 Introduction
`Process migrationis the ability to transfer a process from
`one machineto another.It is a useful facility in distrib-
`uted computing environments, especially as computing
`devices become more pervasive and Internet access be-
`comes more ubiquitous. The potential benefits of process
`migration, amongothers, are fault resilience by migrating
`processes off of faulty hosts, data access locality by mi-
`grating processes closer to the data, better system re-
`sponse time by migrating processes closer to users, dy-
`namic load balancing by migrating processes to less
`loaded hosts, and improved service availability and ad-
`ministration by migrating processes before host mainte-
`nanceso that applications can continue to run with mini-
`mal downtime.
`
`substantial
`Although process migration provides
`potential benefits and many approaches have been con-
`sidered [24], achieving process migration functionality
`has been difficult in practice. Toward this end, there are
`four important goals that need to be met. First, given the
`large numberof widely used legacy applications, applica-
`tions should be able to migrate and continue to operate
`correctly without modification, without requiring that
`they be written using uncommon languages or toolkits,
`and without restricting their use of common operating
`system services. For example, networked applications
`should be able to maintain their network connections
`even after being migrated. Second, migration should
`leverage the large existing installed base of commodity
`operating systems. It should not necessitate use of new
`operating systems or substantial modifications to existing
`ones. Third, migration should maintain the independence
`
`of independent machines. It should avoid creating resid-
`ual dependencies that limit the utility of process migra-
`tion by requiring machines where a process was
`previously executed to continue to service a process even
`after it has migrated to another machine. Fourth, migra-
`tion should be fast and efficient. Overhead should be
`small for normal execution and migration.
`To overcome limitations in previous approaches to
`general-purpose process migration, we have created Zap.
`Zap provides a thin virtualization layer on top of the
`operating system that
`introduces a PrOcess Domain
`(pod) abstraction. A pod provides a group of processes
`with a private namespace that presents the process group
`with the samevirtualized view of the system. This virtu-
`alized view associates virtual identifiers with operating
`system resources such as process identifiers and network
`addresses. This decouples processes in a pod from depen-
`dencies on the host operating system and from other pro-
`cesses in the system.
`Zap virtualization is integrated with a checkpoint-
`restart mechanism that enables processes within a pod to
`be migrated as a unit to another machine. Since pods are
`independent and self-contained they can be migrated
`freely without leaving behind any residual state after
`migration, even when migrating network applications
`while preserving their network connections. Zap can
`therefore allow applications to continue executing after
`migration even if the machine on which they previously
`executed is no longer available. In using a checkpoint-
`restart approach, Zap not only supports process migra-
`tion, but also allows processes to be suspended to second-
`ary storage and transparently resumed at a later time.
`
`USENIX Association
`
`5th Symposium on Operating Systems Design and Implementation
`
`361
`
`
`
`

`

`Going beyond simple migration, this functionality can be
`useful in many ways, including fast creation of user ses-
`sions and simpler, more dynamic system configuration.
`Zap is designed to support migration of unmodified
`legacy applications while minimizing changesto existing
`operating systems. This is done by leveraging loadable
`kernel module functionality in commodity operating sys-
`tems that allows Zap to intercept system calls as needed
`for virtualization and save and restore kernel state as
`needed for migration. Zap’s compatibility with existing
`applications and operating systems makes it simple to
`deploy and use. We have implemented a Zap prototype as
`a loadable kernel module in Linux that supports transpar-
`ent migration, without any kernel modifications, among
`separate machines running independent Linux operating
`systems; it does not require a single-system image across
`all machines. Our experimental results on our Linux Zap
`prototype demonstrate that it can provide general-pur-
`pose process migration functionality with low overhead.
`This paper focuses on the design and implementation
`of the Zap virtualization and migration mechanisms.
`Section 2 describes related work. Section 3 describes the
`pod abstraction provided by Zap. Section 4 describes the
`architecture of Zap and the mechanismsthat support the
`pod abstraction. Section 5 presents an overview of our
`implementation of Zap in Linux. Section 6 presents
`experimental results evaluate the overhead associated
`with Zap virtualization and migration mechanisms and
`demonstrate the utility of Zap for migrating legacy and
`network applications. Finally, we present concluding
`remarks and directions for future work.
`
`2 Related Work
`Many research operating systems have been developed
`that implemented process migration mechanisms, with a
`focus on using migration for load balancing. These sys-
`tems include Accent [31], Amoeba [25], Charlotte [6],
`Chorus [33], MOSIX [7], Sprite [12], and V [11]. These
`operating systems provided a single system image across
`a cluster of machines and providing migration throughout
`the cluster through careful kernel design to provide a glo-
`bal namespace and location-transparent execution. Pro-
`cess state such as IPC, open files, and system calls in
`somecases are typically handled by forwarding requests
`to a home node on which the process originated. If the
`home node fails, migrated processes running on other
`nodes mayfail as well. Although providing a single clus-
`ter operating system can simplify system management,
`these kinds of systems require new operating systems or
`substantial changes to existing ones, which have limited
`their deployment. Furthermore, these approaches do not
`work in the context of increasingly common clusters of
`independent machines, each with its own operating sys-
`tem.
`
`Several systems have been developed to support pro-
`cess migration at the user-level and can be run on unmod-
`ified commercial operating systems. These systems
`include Condor [22], CoCheck [29],
`libckpt [28], and
`MPVM [10]. These systems are primarily intended for
`executing long-running applications on a cluster of
`machines. However, because there is no kernel support
`for process migration, these systems require processes to
`be well-behaved in order to migrate, which meansthat
`such processes cannot use commonoperating system ser-
`vices such as inter-process communication. This severely
`limits the kind of applications that can be used with such
`systems.
`Several systems have been developed that provide
`migration using object-based approaches. These systems
`include Abacus [5], Emerald [19], Globus [13], Legion
`[14], and Rover [18]. These systems are designed as pro-
`gramming languages or middleware toolkits that typi-
`cally require explicit programmer control
`to utilize
`migration. By operating at a higher-level of abstraction,
`these systems can reduce the amountofstate that needs
`to be recorded and moved to migrate an application.
`However,
`these systems require applications to be re-
`written using new programming language environments.
`Asa result, they cannot migrate legacy applications.
`Virtualization at the operating system level has been
`proposed as a mechanism for supporting process migra-
`tion. Zap virtualization was inspired in part by capsules
`[36], an abstraction that provided a private namespace to
`a group of processes that can be migrated as a unit. How-
`ever, capsules did not support migration of networked
`applications while preserving their open network connec-
`tions. Unlike Zap, implementing capsules required exten-
`sive operating system changes. Whereas the capsule
`approachrestructured the operating system to achieveits
`goals, Zap seeks to be compatible with existing operating
`systems by virtualizing the operating system interface
`while minimizing changesto the operating system. Oper-
`ating system virtualization has also been explored in vOS
`[9]
`to provide process migration for simple, non-net-
`worked Windows applications.
`Virtual machine monitors (VMMs)canalso be used as
`a mechanism for process migration [21, 35]. Virtual
`machine monitors such as VMware [1] virtua