ENERGY-EFFICIENCY AND STORAGE FLEXIBILITY IN THE BLUE FILE SYSTEM E. B. Nightingale and J. Flinn University of Michigan.

Slides:



Advertisements
Similar presentations
Consistency and Replication Chapter 7 Part II Replica Management & Consistency Protocols.
Advertisements

Serverless Network File Systems. Network File Systems Allow sharing among independent file systems in a transparent manner Mounting a remote directory.
Study of Hurricane and Tornado Operating Systems By Shubhanan Bakre.
CS-550: Distributed File Systems [SiS]1 Resource Management in Distributed Systems: Distributed File Systems.
G Robert Grimm New York University Disconnected Operation in the Coda File System.
Copyright © Clifford Neuman - UNIVERSITY OF SOUTHERN CALIFORNIA - INFORMATION SCIENCES INSTITUTE CS582: Distributed Systems Lecture 13, 14 -
Web Caching Schemes1 A Survey of Web Caching Schemes for the Internet Jia Wang.
Other File Systems: AFS, Napster. 2 Recap NFS: –Server exposes one or more directories Client accesses them by mounting the directories –Stateless server.
Jeff Chheng Jun Du.  Distributed file system  Designed for scalability, security, and high availability  Descendant of version 2 of Andrew File System.
University of Pennsylvania 11/21/00CSE 3801 Distributed File Systems CSE 380 Lecture Note 14 Insup Lee.
Distributed File System: Data Storage for Networks Large and Small Pei Cao Cisco Systems, Inc.
File Systems (2). Readings r Silbershatz et al: 11.8.
NovaBACKUP 10 xSP Technical Training By: Nathan Fouarge
DESIGN AND IMPLEMENTATION OF THE SUN NETWORK FILESYSTEM R. Sandberg, D. Goldberg S. Kleinman, D. Walsh, R. Lyon Sun Microsystems.
A Low-Bandwidth Network File System A. Muthitacharoen, MIT B. Chen, MIT D. Mazieres, NYU.
A LOW-BANDWIDTH NETWORK FILE SYSTEM A. Muthitacharoen, MIT B. Chen, MIT D. Mazieres, New York U.
Energy Efficiency and Storage Flexibility in the Blue File System Edmund B Nightingale Jason Flinn University of Michigan.
Distributed File Systems Concepts & Overview. Goals and Criteria Goal: present to a user a coherent, efficient, and manageable system for long-term data.
Review Session for Fourth Quiz Jehan-François Pâris Summer 2011.
Distributed Systems. Interprocess Communication (IPC) Processes are either independent or cooperating – Threads provide a gray area – Cooperating processes.
Distributed File Systems
CHAPTER 2: COMPUTER-SYSTEM STRUCTURES Computer system operation Computer system operation I/O structure I/O structure Storage structure Storage structure.
Distributed File Systems Case Studies: Sprite Coda.
Distributed file systems, Case studies n Sun’s NFS u history u virtual file system and mounting u NFS protocol u caching in NFS u V3 n Andrew File System.
Distributed File Systems Overview  A file system is an abstract data type – an abstraction of a storage device.  A distributed file system is available.
CS 5204 (FALL 2005)1 Leases: An Efficient Fault Tolerant Mechanism for Distributed File Cache Consistency Gray and Cheriton By Farid Merchant Date: 9/21/05.
July 30, 2001Systems Architecture II1 Systems Architecture II (CS ) Lecture 8: Exploiting Memory Hierarchy: Virtual Memory * Jeremy R. Johnson Monday.
CEPH: A SCALABLE, HIGH-PERFORMANCE DISTRIBUTED FILE SYSTEM S. A. Weil, S. A. Brandt, E. L. Miller D. D. E. Long, C. Maltzahn U. C. Santa Cruz OSDI 2006.
Introduction to DFS. Distributed File Systems A file system whose clients, servers and storage devices are dispersed among the machines of a distributed.
Operating Systems David Goldschmidt, Ph.D. Computer Science The College of Saint Rose CIS 432.
SPECULATIVE EXECUTION IN A DISTRIBUTED FILE SYSTEM E. B. Nightingale P. M. Chen J. Flint University of Michigan.
DISCONNECTED OPERATION IN THE CODA FILE SYSTEM J. J. Kistler M. Sataynarayanan Carnegie-Mellon University.
Data Replication and Power Consumption in Data Grids Susan V. Vrbsky, Ming Lei, Karl Smith and Jeff Byrd Department of Computer Science The University.
CODA: A HIGHLY AVAILABLE FILE SYSTEM FOR A DISTRIBUTED WORKSTATION ENVIRONMENT M. Satyanarayanan, J. J. Kistler, P. Kumar, M. E. Okasaki, E. H. Siegel,
Fast Crash Recovery in RAMCloud. Motivation The role of DRAM has been increasing – Facebook used 150TB of DRAM For 200TB of disk storage However, there.
Mobile Data Access1 Replication, Caching, Prefetching and Hoarding for Mobile Computing.
A Low-bandwidth Network File System Athicha Muthitacharoen et al. Presented by Matt Miller September 12, 2002.
CS425 / CSE424 / ECE428 — Distributed Systems — Fall 2011 Some material derived from slides by Prashant Shenoy (Umass) & courses.washington.edu/css434/students/Coda.ppt.
Distributed File Systems
Distributed Systems CS Consistency and Replication – Part IV Lecture 13, Oct 23, 2013 Mohammad Hammoud.
Chapter 7: Consistency & Replication IV - REPLICATION MANAGEMENT By Jyothsna Natarajan Instructor: Prof. Yanqing Zhang Course: Advanced Operating Systems.
Distributed File Systems Group A5 Amit Sharma Dhaval Sanghvi Ali Abbas.
Solutions for the Fourth Problem Set COSC 6360 Fall 2014.
Distributed File Systems Questions answered in this lecture: Why are distributed file systems useful? What is difficult about distributed file systems?
Highly Available Services and Transactions with Replicated Data Jason Lenthe.
1 CEG 2400 Fall 2012 Network Servers. 2 Network Servers Critical Network servers – Contain redundant components Power supplies Fans Memory CPU Hard Drives.
THE EVOLUTION OF CODA M. Satyanarayanan Carnegie-Mellon University.
Solutions for Fourth Quiz COSC 6360 Fall First question What do we mean when we say that NFS client requests are: (2×10 pts)  self-contained? 
Mobile File Systems.
Solution to the Fourth COSC 6360 Quiz for Fall 2013
Chapter 7: Consistency & Replication IV - REPLICATION MANAGEMENT -Sumanth Kandagatla Instructor: Prof. Yanqing Zhang Advanced Operating Systems (CSC 8320)
NFS and AFS Adapted from slides by Ed Lazowska, Hank Levy, Andrea and Remzi Arpaci-Dussea, Michael Swift.
Energy Efficiency and Storage Flexibility in the Blue File System
Outline Midterm results summary Distributed file systems – continued
Today: Coda, xFS Case Study: Coda File System
CSE 451: Operating Systems Winter Module 22 Distributed File Systems
Cooperative Caching, Simplified
Distributed File Systems
Distributed File Systems
Energy Efficiency and Storage Flexibility in the Blue File System
Outline Announcements Lab2 Distributed File Systems 1/17/2019 COP5611.
Cary G. Gray David R. Cheriton Stanford University
CSE 451: Operating Systems Spring Module 21 Distributed File Systems
DESIGN AND IMPLEMENTATION OF THE SUN NETWORK FILESYSTEM
Distributed File Systems
CSE 451: Operating Systems Winter Module 22 Distributed File Systems
Outline Review of Quiz #1 Distributed File Systems 4/20/2019 COP5611.
Distributed File Systems
Page Cache and Page Writeback
Distributed File Systems
Presentation transcript:

ENERGY-EFFICIENCY AND STORAGE FLEXIBILITY IN THE BLUE FILE SYSTEM E. B. Nightingale and J. Flinn University of Michigan

Key ideas Reduce power consumption of portable devices –Two big culprits Disk drive Wireless Integrate flash drives

Dynamic storage hierarchy We should power down drives when they are idle –But this causes powering up delays To mask these delays introduce a dynamic storage hierarchy –Includes flash drives and remote server –Takes into consideration state of local disk drive

Examples Disk is powered upDisk is powered down Flash Disk Server Flash Server Disk The storage hierarchy depends on the status of the hard disk.

Incorporating flash drives To satisfy read requests, Blue FS looks first in the flash drive Implies –A write all/read any policy –Treating flash drives as caches Need a quick way to query flash drive contents

More on caches Unit of caching is block Can cache some blocks of a file but not all –Must maintain information on validity and currency of each cached block See details in paper

Write to all/read any All writes are propagated to all devices Disk Server

Advantages Write all allows read any –Future power savings Requires efficient writes –Propagates all updates to the server Unless in disconnected mode

Aggregating writes (I) Blue FS maintains one write queue per device Disk Server

Aggregating writes (II) When it writes to a device, it flushes the whole queue Disk Server

Aggregating writes (III) After the writes, each queue is empty Disk Server

More on device queues Device queues share among themselves all common blocks –Save space Blocks in device queues can be accessed through a hashing scheme –Always access most recent version of a block even when it is not yet saved on any device

Why? Makes a lot of sense for disk drive and remote server –Power up disk, do a few updates, then power down –Power up wireless connection, send a few updates then power down Delaying updates allows BlueFS to coalesce multiple updates to the same block

Using flash drives as caches Small flash drives cannot contain whole contents of file system –Especially true in 2004! Must have a fast way to find whether a file is cached or not – Enode

E-node Captures all the information Blue FS has about the validity of an object E-nodes are – Hashed by file id –Stored in an e-node cache managed by LRU replacement policy Default size is one MB

Cache consistency Uses –Optimistic replica control (Coda) – Callbacks (AFS) Changes –Blue FS maintains callbacks on a per-device basis, instead of on a per-client basis –Server queues invalidation messages when a device is disconnected.

Optimistic replication Optimistic replication control protocols allow access in disconnected mode –Tolerate temporary inconsistencies –Promise to detect them later –Provide much higher data availability Updated for Fall 2012 Optimistic replication control puts data availability above data consistency

Callbacks (I) When a client opens a file for the first time, server promises to notify it whenever it receives a new version of the file from any other client –Promise is called a callback Relieves the server from having to answer a call from the client every time the file is opened – Significant reduction of server workload Updated for Fall 2012

If machines could talk (I) Client I should contact the server each time I open any file Server 99% of these requests are useless!

If machines could talk (II) Client ?? Server You do not need to call me: I promise to call you back if anyone has modified the file

If machines could talk (III) Client What if I do not receive your callback? Server Though luck!

Callbacks (II) Callbacks can be lost! –Client will call the server every tau  minutes to check whether it received all callbacks it should have received –Cached copy is only guaranteed to reflect the state of the server copy up to tau minutes before the time the client opened the file for the last time

If machines could talk (IV) Client I will contact you from time to time to check for lost callbacks Server Sure as long as you do not do it too frequently

Disconnected mode Like Coda, Blue FS works in disconnected mode User has access to all files cached –On local disk drive –On flash drive if any Can even specify that some files must always be cached – Affinity

Disconnected mode (II) When a device is disconnected, server queues all callbacks that could not be delivered to the device – Speeds up reintegration If inconsistencies are discovered at reconnection time, will run a resynchronization process

Accessing large files When disk is powered down, Blue FS get first file blocks from remote server

Deciding which device to use If decisions are made for each individual access –Will never activate a disk that’s powered down Blue FS uses ghost hints –Measure penalty for not using a disk that is powered down for a given access –When sum of ghosts hints exceeds the cost of powering up the disk, disk is powered up

Blue FS architecture (II) Applications BlueFS Kernel Module Wolverine Linux Kernel Linux File Cache USB Stick Card Flash Micro- drive Local Disk To BlueFS Server VFS Operatio ns Up- call Updated for Fall 2012

Blue FS architecture (I) Most of Blue FS functionality is implemented in a user-level process –Wolverine Kernel-resident part of Blue FS intercepts I/O calls that have to be forwarded to Wolverine

Blue FS architecture (II) Applications BlueFS Kernel Module Wolverine Linux Kernel Linux File Cache USB Stick Card Flash Micro- drive Local Disk To BlueFS Server VFS Operations Up-call

Performance Will use a standard benchmark –Andrews/AFS benchmark Will compare interactive delays and power consumption against –NFS –Coda with synchronous server updates (Coda) –Coda with asynchronous server updates (Coda WD)

Interactive delay Ten times faster than NFS, 16% faster than Coda WD

Adding a 16MB flash drive Assumes that cache is fully loaded 48% shorter interactive delay 48%, 25% less energy

Energy efficiency Different benchmark run with half-full cache 76% shorter interactive delay, 55% less energy than Coda

Adding a 16MB flash drive Assumes that cache is fully loaded 48% shorter interactive delay, 25% less energy

Discussion Comparison with NFS is unfair –NFS was designed for LANs, not WANs! Comparison with Coda is fair but –Coda did not incorporate flash drives Did not exist then! –Coda did not try to minimize power consumption Portable devices did not exist then!

Conclusion Blue FS updates Coda by taking into consideration the arrival of –Wireless networks (near ubiquitous access) –Portable devices (limited autonomy) –Flash drives Retains Coda features such as optimistic replication, callbacks and disconnected mode operation