Advanced Operating Systems - Spring 2009Lecture 21 – Monday April 6st, 2009Dan C. Marinescu

Email: dcm@cs.ucf.eduOffice: HEC 439 B. Office hours: M, Wd 3 – 4:30 PM.

TA: Chen YuEmail: yuchen@cs.ucf.eduOffice: HEC 354. Office hours: M, Wd 1.00 – 3:00 PM.

1

Last, Current, Next Lecture Last time:

Distributed File System Today

Andrew File System Network and Distributed Operating Systems Multiple Access Networks

Next time: Interconnection Networks

2

Stateless vs. stateful serversStateless service:

longer request messages slower request processing additional constraints imposed on DFS design

Some environments require stateful service A server employing server-initiated cache validation

cannot provide stateless service, since it maintains a record of which files are cached by which clients

UNIX use of file descriptors and implicit offsets is inherently stateful; servers must maintain tables to map the file descriptors to inodes, and store the current offset within a file

File ReplicationReplicas reside on failure-independent machines Improves availability and can shorten service timeNaming scheme maps a replicated file name to a particular

replica Existence of replicas should be invisible to higher levels Replicas must be distinguished from one another by different

lower-level namesUpdates an update to any replica must be reflected on all

replicasDemand replication – reading a nonlocal replica causes it to be

cached locally, thereby generating a new nonprimary replica.

Andrew File System (AFS)AFS tries address issues such as:

uniform name space location-independent file sharing client-side caching (with cache consistency) secure authentication (via Kerberos) server-side caching (via replicas) high availability scalability can span 5,000 workstations

History A distributed computing environment developed since 1983 at

CMU Purchased by IBM and released as Transarc DFS, Now open source OpenAFS

AFS (cont’d)

Clusters clients and servers form clusters interconnected by a backbone LAN a cluster several workstations and a cluster server connected to

the backbone by a routerClients see a partitioned space of file names:

a local name space and a shared name space the local name space is the root file system of a workstation, from

which the shared name space descendsServers collectively are responsible for the storage and

management of the shared name space.Opening a file causes it to be cached, in its entirety, on the local

disk

AFS (cont’d) Vice dedicated servers present the shared name space to the clients

as an homogeneous, identical, and location transparent file hierarchy. Workstations

run the Virtue protocol to communicate with Vice are required to have local disks where they store their local name space

Andrew’s volumes small units associated with the files of a single client

fid (96 bits) identifies a Vice file or directory; three components: volume number vnode number – index into an array containing the inodes of files in a single

volume uniquifier – allows reuse of vnode numbers, thereby keeping certain data

structures, compact Fids are location transparent; therefore, file movements from server to

server do not invalidate cached directory contents Location information

kept on a volume basis replicated on each server

File Operations in AFSAndrew caches entire files form servers.

A client workstation interacts with Vice servers only during opening and closing of files

A component called Venus caches files from Vice when they are opened, and stores modified

copies of files back when they are closed Caches contents of directories and symbolic links, for path-name

translation Venus manages two separate caches:

Status cache kept in virtual memory to allow rapid servicing of stat (file status returning) system calls

Data cache on the local disk; the UNIX I/O buffering mechanism does some caching that are transparent to Venus

LRU algorithm used to keep each of them bounded in size Exceptions to the caching policy are modifications to directories

that are made directly on the server responsibility for that directory

Reading and writing to a file done by the kernel without Venus intervention on the cached copy

AFS implementation

Client processes are interfaced to a UNIX kernel with the usual set of system calls

Venus carries out path-name translation component by component

The UNIX file system is used as a low-level storage system for both servers and clientsThe client cache is a local directory on the workstation’s

diskBoth Venus and server processes access UNIX files directly by

their inodes to avoid the expensive path name-to-inode translation routine

Distributed systemsDistributed system collection of heterogeneous systems

(different processor architecture, OS, libraries, applications) linked to each other by an interconnection network.

Communication message passing.Advantages of distributed systems

Resource sharing better utilization of resources Fault-tolerance systems fail independently, increase

redundancy Scalability the system can grow in time Supports collaborative environments in enterprise computing,

engineering (e.g., CAD systems), science (e.g., GRID), etc.Problems

Resource management more difficult. Hard to manage autonomous systems New services necessary e.g., resource discovery

Security Harder to construct distributed applications

Distributed system architectureService-oriented architectures set of services provided

by autonomous service providers. Based upon client-server paradigm and request-response communication GRID, semantic Web

User-Coordinator-Executor architecture multiple sites provide computing resources ; the coordinator acts as an agent of the user and starts applications at participating sites and then monitors the execution. Potential use in high performance computing.

Peer-to-peer architectures the systems function simultaneously as client and server

Service oriented-distributed systems

Autonomous vs non-autonomous systemsAutonomous systems Resources of individual systems

controlled by the local operating systems. Often in distinct administrative domains. Open system new resources added or removed continually Scalable. No one tries to maintain common state

Non-autonomous systems Resources controlled by a Network operating system Users are aware of multiplicity of

machines. Access to resources of various systems done explicitly by:

Remote logging into the appropriate remote machine (telnet, ssh) Remote Desktop (Microsoft Windows) Transferring data from remote machines to local machines, via the File

Transfer Protocol (FTP) mechanism Distributed Operating System Users not aware of multiplicity of

machines Access to remote resources similar to access to local resources Common state.

Data Migration – transfer data Computation Migration – transfer the computation, rather than data

Process migrationPossible only if a homogeneous architecture.Load balancing.Synchronization across the network.Coordination