Distributed Systems CS 757
Popular in Course
Popular in ComputerScienence
verified elite notetaker
This 23 page Class Notes was uploaded by Abe Jones on Saturday September 12, 2015. The Class Notes belongs to CS 757 at West Virginia University taught by Goseva-Popstojanova in Fall. Since its upload, it has received 16 views. For similar materials see /class/202748/cs-757-west-virginia-university in ComputerScienence at West Virginia University.
Reviews for Distributed Systems
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 09/12/15
DISTRIBUTED FILE SYSTEMS Outline quot III Definition 11 Goals III Architectures III Communication D Namespaces III Replication D Case Study 1 Questions Distributed File System quot III What is 0 Distributed File System J A Distributed File System is d file system that is made up of several networked processes eoch responsible for only C portion of the file system Goals of a Distributed File System II III Transparency A Distributed File System should allow clients to access remote files in the same way that local files are accessed El Availability Resources should have a high probability of a successful access El Performance Access to resources should be as efficient as possible Transparency Goals E III Access Transparency Users should have a uniform way to access the file system I Location Transparency Users should not be aware of where resources are physically located a Migration Transparency Users should not be aware that resources may change physical storage locations Transparency Goals cont I III Relocation Transparency When a file moves the users should not notice even if they are using the file El Replication Transparency Users should not be aware of replicated files El Concurrency Transparency Each users should appear to have exclusive access to the file system El Failure Transparency The system should correct failures without users noticing Availability Goals III How can availability be improved I Replication I Can help to hide machine failures I Can help to hide network failures sometimes I Reliable networks I Reliable servers Performance Goals quot III How can performance be improved I Server Replication I Serverside Content Replication ll Client Caching Architectures n III Asymmetric El ClientServer I Remote Access Model I UploodDownlood Model m Cluster Bosed l Multiple file servers with possibility for global storage loyer III Symmetric n Peertopeer l Typically Use Distributed Hosh Tdbles Communication n III Benefits of using Remote Procedure Coll It Already Exists and is Proven to Work for Distributed File Systems Operating System Independence Network Independence Transport Protocol Independence Namespaces III III Local Namespace The user typically chooses where portions of the distributed file system will be mounted in the local file system Multiple servers may exists each with different naming structures CI Global Namespace Globally agreed upon namespace hierarchy I Each server can maintain their own naming structure but there is at least a mapping to a globally agreed structure The user typically mounts portions of the distributed file system in a standard location in the local file system Locol Nomespoce n Client A Server Client B Av A if Ar 9 l l Exported directory Exported directory mounted by client mounted by client Network Image reproduced from 1 Global Namespcrce Client A Naming inherited from server s name space magma afsquot Exported direc mounted by ciient Sewer Client B J V i 739 T 77 Yilr r 1 i y quoti tory Exported directory mounted by client Netwo Image reproduced from 1 1 rk Replication III III Distributed File System Benefits ll Performance ll Fault Tolerance ll Availability III Common Drawback II Overhead associated maintaining consistency Case Study IE El Distributed File System The Google File System GFS El Distributed Storage System Bigtable III Purpose of the Study To reveal some of the key properties of distributed file systems that Google researchers have leveraged for simplicity and performance in distributed storage systems The Google File System Architecture In III A Single Master I System Coordination ll Metocloto III Chunkservers ll 64MB chunks of data III Clients ll Uses Master to find Doto El Uses Chunkservers to get Doto The Google File System Anli mim a lIle nnmc chunk Index 391 l I M gt llmb 39 UFScliunl File Chunklem chunk handle chunk locations Legend l A Data messngu Imtruclionx 10 chunkxervcr 39 Cm lml NSquot3395 Chunkserver slate chunk handle byte range I I lFS chunksorwr l I FH chunksvrwr chunk dam quotquot39 I Luluxllle system l I Linux Ille xyslcm l 7 r 1 393 l l Image reproduced from 2 Bigtable Distributed Storage System III Architecture at A Single Master a Tablet Servers D Clients III Components D Google File System I Stores tablets and logs n Chubby Distributed Lock Service I Holds metadata handles master election D Scheduling System Big rable Data Model II III Multidimensional sparse sorted map 3D Spreadsheet row column Timestamp gt value contents Columns Rows quotwwwcnncomquot Timestamps Image reproduced from httpresearchyahoocom les6DeanGooglepdf Findings E El Key aspects of the Google File System that Google researchers leveraged for simplicity and performance in Bigtable Data Persistence Shared Data Access Architectural Similarities With Distributed Storage Systems Summary n I Distributed file systems should perform much the same as local file systems El Case Study Data persistence at the file system level can reduce or eliminate the need for redundancy at the data storage system level Shared data access can provide processes with a common view of the system which provides a form of global knowledge Similarities in the architectures of the file system and storage system can be beneficial for not only better utilization of hardware but also performance Ques rions References n 1 Andrew S Tanenbaum and Maarten Van Steen Distributed Systems Principles and Paradigms Pearson Prentice Hall 2007 2 Saniay Ghemawat Howard Gobiott and ShunTak Leung The Google File System Proceedings of the nineteenth ACM symposium on Operating systems principles October 1922 2003 Bolton Landing NY USA 3 Chang E Dean J Ghemawat 8 Hsieh W C Wallach D A Burrows M Chandra T Fikes A Gruber R E and Gruber R E 2006 Bigtable a distributed storage system for structured data In Proceedings of the 7th Conference on USENIX Symposium on Operating Systems Design and Implementation Vol 7 November 06 O8 2006