1. Distributed System and Middleware Distributed Systems Distributed File System Dr. Sunny Jeong. spjeong@uic.edu.hk Mr. Coling Zhang colinzhang@uic.edu.hk With Thanks to Prof. G. Coulouris, Prof. A.S. Tanenbaum and Prof. S.C Joo

2. Distributed System and Middleware Overview  Requirements for distributed file systems  transparency, performance, fault-tolerance, Consistency...  Design issues  possible options, architectures  file sharing, concurrent updates  Caching  Examples  Sun Network File System  Andrew File System 1

3. Distributed System and Middleware Distributed Services 2

4. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS A Distributed File System ( DFS ) is simply a classical model of a file system ( as discussed before ) distributed across multiple machines. The purpose is to promote sharing of dispersed files. This is an area of active research interest today. The resources on a particular machine are local to itself. Resources on other machines are remote. A file system provides a service for clients. The server interface is the normal set of file operations: create, read, etc. on files. Definitions

5. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS Clients, servers, and storage are dispersed across machines. Configuration and implementation may vary - a) Servers may run on dedicated machines, OR b) Servers and clients can be on the same machines. c) The OS itself can be distributed (with the file system a part of that distribution. a) A distribution layer can be interposed between a conventional OS and the file system. Clients should view a DFS the same way they would a centralized FS; the distribution is hidden at a lower level. Performance is concerned with throughput and response time. Definitions

6. Distributed System and Middleware Distributed file service  Basic services  persistent file storage of data and programs  operations on files (create, open, read, write…)  multiple remote clients within intranet  file sharing  typically one-copy update semantics over RPC  Many new developments  persistent object stores (storage of objects)  Persistent Java, CORBA, …  replication, whole-file caching  distributed multimedia (Tiger video file server) 5

7. Distributed System and Middleware Storage system and their properties SharingPersis- tence Distributed cache/replicas Consistency maintenance Example Main memory RAM File systemUNIX file system Distributed file systemSun NFS Web Web server Distributed shared memoryIvy (Chap. 16) Remote objects (RMI/ORB)CORBA Persistent object store 1 CORBA Persistent Object Service Persistent distributed object store PerDiS, Khazana 1 1 1 * “1” is for one-copy consistency 6

8. Distributed System and Middleware Characteristics of file systems  Operations on files ( =data + attributes)  create/delete  query/modify attributes  open/close  read/write  access control  Storage organization  directory structure (hierarchical, pathnames)  metadata (= file management information, data about data)  file attributes  directory structure information, etc 7

9. Distributed System and Middleware Characteristics of file systems  Persistently stored data sets( files = data + attributes)  Hierarchic name space visible to all processes  API with the following characteristics:  access and update operations on persistently stored data sets  sequential access model (with additional random facilities)  Sharing of data between users, with access control  Concurrent access:  certainly for read-only access  what about updates? 8

10. Distributed System and Middleware File attribute record structure File length Creation timestamp Read timestamp Write timestamp Attribute timestamp Reference count Owner File type Access control list(ACL) E.g. for UNIX: rw-rw-r-- User controlled updated by system: updated by owner: 9

11. Distributed System and Middleware File system Modules Concentrate on higher levels. 10

12. Distributed System and Middleware Distributed file system requirements [1/4]  Facilities  support the sharing of persistent storage and information  enable user programs to access files without copying them to a local disk  Transparency (clients unaware of the distributed nature)  access transparency - client unaware of distribution of files, same interface for local/remote files  location transparency - uniform file name space from any client workstation  mobility transparency - files can be moved from one server to another without affecting client  performance transparency - client performance not affected by load on service  scaling transparency - expansion possible if numbers of clients increase 11

13. Distributed System and Middleware [Distributed file system requirements – ctd[2/4]  Concurrent file updates  changes by one client do not affect another  Isolation  File-level or record-level locking  Other forms of concurrency control to minimise contention (Minimum Competition)  File replication  File service maintains multiple identical copies of files  Load-sharing between servers makes service more scalable  Local access has better response (lower latency)  Fault tolerance  Full replication is difficult to implement  Caching (of all or part of a file) gives most of the benefits (except fault tolerance) 12

14. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS Naming is the mapping between logical and physical objects.  Example: A user filename maps to.  In a conventional file system, it's understood where the file actually resides; the system and disk are known.  In a transparent DFS, the location of a file, somewhere in the network, is hidden.  File replication means multiple copies of a file; mapping returns a SET of locations for the replicas. Location transparency - a) The name of a file does not reveal any hint of the file's physical storage location. a) File name still denotes a specific, although hidden, set of physical disk blocks. b) This is a convenient way to share data. c) Can expose correspondence between component units and machines. Naming and Transparency

15. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS Location independence -  The name of a file doesn't need to be changed when the file's physical storage location changes. Dynamic, one-to-many mapping.  Better file abstraction.  Promotes sharing the storage space itself.  Separates the naming hierarchy from the storage devices hierarchy. Most DFSs today:  Support location transparent systems.  Do NOT support migration; (automatic movement of a file from machine to machine.)  Files are permanently associated with specific disk blocks. Naming and Transparency

16. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS The ANDREW DFS AS AN EXAMPLE:  Is location independent.  Supports file mobility.  Separation of FS and OS allows for disk-less systems. These have lower cost and convenient system upgrades. The performance is not as good. NAMING SCHEMES: There are three main approaches to naming files: 1. Files are named with a combination of host and local name. This guarantees a unique name. NOT location transparent NOR location independent. Same naming works on local and remote files. The DFS is a loose collection of independent file systems. Naming and Transparency

17. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS NAMING SCHEMES: 2. Remote directories are mounted to local directories. So a local system seems to have a coherent directory structure. The remote directories must be explicitly mounted. The files are location independent. SUN NFS is a good example of this technique. 3. A single global name structure spans all the files in the system. The DFS is built the same way as a local filesystem. Location independent. Naming and Transparency

18. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS IMPLEMENTATION TECHNIQUES:  Can Map directories or larger aggregates rather than individual files.  A non-transparent mapping technique: name ---->  A transparent mapping technique: name ----> file_identifier ---->  So when changing the physical location of a file, only the file identifier need be modified. This identifier must be "unique" in the universe. Naming and Transparency

19. Distributed System and Middleware File Service Design Options  State full  server holds information on open files, current position, file locks  open before access, close after access  better performance  shorter message, read-ahead possible  server failure  lose state  client failure  tables fill up  can provide file locks 18

20. Distributed System and Middleware File Service Design Options -ctd  Stateless  no state information held by server  file operations(idempotent) must contain all information needed (longer message)  simpler file server design  can recover easily from client or server crash  locking requires extra lock server to hold state 19

21. Distributed System and Middleware File Service Architecture Client Side File server Side 20

22. Distributed System and Middleware File Service Architecture Client computerServer computer Application program Application program Client module Flat file service Directory service Lookup AddName UnName GetNames Read Write Create Delete GetAttributes SetAttributes 21

23. Distributed System and Middleware File Server Architecture -ctd  Components (for openness):  Flat file service  Flat file service operations below on file contents  Have unique file identifiers (UFIDs)  translates UFIDs to file locations Read(FileId, i, n) -> Data —throws BadPosition If 1 ≤ i ≤ Length(File): Reads a sequence of up to n items from a file starting at item i and returns it in Data. Write(FileId, i, Data) — throws BadPosition If 1 ≤ i ≤ Length(File)+1: Writes a sequence of Data to a file, starting at item i, extending the file if necessary. Create() -> FileIdCreates a new file of length 0 and delivers a UFID for it. Delete(FileId) Removes the file from the file store. GetAttributes(FileId) -> Attr Returns the file attributes for the file. SetAttributes(FileId, Attr) Sets the file attributes (only those attributes that are not shaded in ). 22

24. Distributed System and Middleware File Server Architecture -ctd  Directory service  mapping between text-(file) names to UFIDs  Client module  API for file access, one per client computer  holds states: open files, positions  knows network location of flat file & directory server Flat file service Read(FileId, i, n) -> Data Write(FileId, i, Data) Create() -> FileId Delete(FileId) GetAttributes(FileId) -> Attr SetAttributes(FileId, Attr) Directory service Lookup(Dir, Name) -> FileId AddName(Dir, Name, FileId) UnName(Dir, Name) GetNames(Dir, Pattern) -> NameSeq 23

25. Distributed System and Middleware Flat file service RPC interface  Used by client modules, not user programs  FileId (UFID) uniquely identifies file  invalid if file not present or inappropriate access  Read/Write; Create/Delete; Get/SetAttributes  No open/close! (unlike UNIX)  access immediate with FileId  Read/Write identify starting point  Improved fault-tolerance  operations idempotent except Create, can be repeated (at-least-once RPC semantics)  stateless service 24

26. Distributed System and Middleware Access control  In UNIX file system  access rights are checked against the access mode (read, write, execute) in open  user identity checked at login time, cannot be tampered(=changed) with in non-distributed implementations.  In distributed (file) systems  Access rights must be checked at server  RPC unprotected  Forging identity possible, a security risk  user id typically passed with every request (e.g. Sun NFS)  stateless 25

27. Distributed System and Middleware Directory structure  Hierarchical  tree-like, pathnames from root  (in UNIX) several names per file (link operation)  Naming system  implemented by client module, using directory service  root has well-known UFID  locate file following path from root bigbobjon people export (root)... 26

28. Distributed System and Middleware File names Text name = directory pathname+file name  hostname:local name  not mobility transparent  uniform name structure (the same name space for all clients)  remote mount (e.g. Sun NFS)  remote directory inserted into local directory  relies on clients maintaining consistent naming conventions across all clients  all clients must implement same local tree  must mount remote directory into the same local directory 27

29. Distributed System and Middleware File names  Mount operation: mount(remotehost, remotedirectory, localdirectory)  A server maintains a table of clients who have mounted file systems at that server.  Each client maintains a table of mounted file systems holding:  Hard versus soft mounts 28

30. Distributed System and Middleware Remote mount jimjanejoeann users students usrvmunix ClientServer 2...nfs Remote mount staff bigbobjon people Server 1 export (root) Remote mount... x (root) Note: The file system mounted at /usr/students in the client is actually the sub-tree located at /export/people in Server 1; the file system mounted at /usr/staff in the client is actually the sub-tree located at /nfs/users in Server 2.  server-side : /export/people, : /nfs/users  client-side : mount -t nfs server1:/export/people /usr/students /* client: /usr/students(=people)/jon,… */  client-side : mount -t nfs server2:/nfs/users /usr/staff /* client:/usr/staff(=users)/jane, … */ 29

31. Distributed System and Middleware Directory service  Directory  conventional file (client of the flat file service)  mapping from text names to UFIDs  Operations  require FileId, machine readable UFID as parameter  locate file (LookUp)  add/delete file (AddName/UnName)  match file names to regular expression (GetNames) 30

32. Distributed System and Middleware Directory service operations Lookup(Dir, Name) -> FileId — throws NotFound Locates the text name in the directory and returns the relevant UFID. If Name is not in the directory, throws an exception. AddName(Dir, Name, File) —throws NameDuplicate If Name is not in the directory, adds (Name, File) to the directory and updates the file’s attribute record. If Name is already in the directory: throws an exception. UnName(Dir, Name) —throws NotFound If Name is in the directory: the entry containing Name is removed from the directory. If Name is not in the directory: throws an exception. GetNames(Dir, Pattern) -> NameSeq Returns all the text names in the directory that match the regular expression Pattern. 31

33. Distributed System and Middleware File sharing Multiple clients share the same file for read/write access.  One-copy update semantics  every read sees the effect of all previous writes  a write is immediately visible to clients who have the file open for reading  Problems!  caching: maintaining consistency between several copies difficult to achieve  serialize access by using file locks (affects performance )  trade-off between consistency and performance 32

34. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS CACHING  Reduce network traffic by retaining recently accessed disk blocks in a cache, so that repeated accesses to the same information can be handled locally.  If required data is not already cached, a copy of data is brought from the server to the user.  Perform accesses on the cached copy.  Files are identified with one master copy residing at the server machine, Copies of (parts of) the file are scattered in different caches. Cache Consistency Problem -- Keeping the cached copies consistent with the master file. Remote File Access

35. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS CACHING A remote service ((RPC) has these characteristic steps: a) The client makes a request for file access. b) The request is passed to the server in message format. c) The server makes the file access. d) Return messages bring the result back to the client. This is equivalent to performing a disk access for each request. Remote File Access

36. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS CACHE LOCATION: Caching is a mechanism for maintaining disk data on the local machine. This data can be kept in the local memory or in the local disk. Caching can be advantageous both for read ahead and read again. The cost of getting data from a cache is a few HUNDRED instructions; disk accesses cost THOUSANDS of instructions. The master copy of a file doesn't move, but caches contain replicas of portions of the file. Caching behaves just like "networked virtual memory". Remote File Access

37. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS CACHE LOCATION: What should be cached? files >>. Bigger sizes give a better hit rate; Smaller give better transfer times.  Caching on disk gives: — Better reliability.  Caching in memory gives: — The possibility of diskless work stations, — Greater speed, Since the server cache is in memory, it allows the use of only one mechanism. Remote File Access

38. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS CACHE UPDATE POLICY: A write through cache has good reliability. But the user must wait for writes to get to the server. Used by NFS. Delayed write - write requests complete more rapidly. Data may be written over the previous cache write, saving a remote write. Poor reliability on a crash.  Flush sometime later tries to regulate the frequency of writes.  Write on close delays the write even longer.  Which would you use for a database file? For file editing? Remote File Access

39. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS Example: NFS with Cachefs

40. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS CACHE CONSISTENCY : The basic issue is, how to determine that the client-cached data is consistent with what's on the server.  Client - initiated approach - The client asks the server if the cached data is OK. What should be the frequency of "asking"? On file open, at fixed time interval,...?  Server - initiated approach - Possibilities: A and B both have the same file open. When A closes the file, B "discards" its copy. Then B must start over. The server is notified on every open. If a file is opened for writing, then disable caching by other clients for that file. Get read/write permission for each block; then disable caching only for particular blocks. Remote File Access

41. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS COMPARISON OF CACHING AND REMOTE SERVICE:  Many remote accesses can be handled by a local cache. There's a great deal of locality of reference in file accesses. Servers can be accessed only occasionally rather than for each access.  Caching causes data to be moved in a few big chunks rather than in many smaller pieces; this leads to considerable efficiency for the network.  Cache consistency is the major problem with caching. When there are infrequent writes, caching is a win. In environments with many writes, the work required to maintain consistency overwhelms caching advantages.  Caching requires a whole separate mechanism to support acquiring and storage of large amounts of data. Remote service merely does what's required for each call. As such, caching introduces an extra layer and mechanism and is more complicated than remote service. Remote File Access

42. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS STATEFUL VS. STATELESS SERVICE: Stateful: A server keeps track of information about client requests.  It maintains what files are opened by a client; connection identifiers; server caches.  Memory must be reclaimed when client closes file or when client dies. Stateless: Each client request provides complete information needed by the server (i.e., filename, file offset ).  The server can maintain information on behalf of the client, but it's not required.  Useful things to keep include file info for the last N files touched. Remote File Access

43. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS STATEFUL VS. STATELESS SERVICE: Performance is better for stateful.  Don't need to parse the filename each time, or "open/close" file on every request.  Stateful can have a read-ahead cache. Fault Tolerance: A stateful server loses everything when it crashes.  Server must poll clients in order to renew its state.  Client crashes force the server to clean up its encached information.  Stateless remembers nothing so it can start easily after a crash. Remote File Access

44. Distributed System and Middleware DISTRIBUTED FILE SYSTEMS FILE REPLICATION:  Duplicating files on multiple machines improves availability and performance.  Placed on failure-independent machines ( they won't fail together ). Replication management should be "location-opaque".  The main problem is consistency - when one copy changes, how do other copies reflect that change? Often there is a tradeoff: consistency versus availability and performance.  Example: "Demand replication" is like whole-file caching; reading a file causes it to be cached locally. Updates are done only on the primary file at which time all other copies are invalidated.  Atomic and serialized invalidation isn't guaranteed ( message could get lost / machine could crash. ) Remote File Access

45. Distributed System and Middleware Example: Sun NFS (1985)  An industry standard for file sharing on local networks since the 1980s  An open standard with clear and simple interfaces  Closely follows the abstract file service model defined above  Supports many of the design requirements already mentioned:  transparency  heterogeneity  efficiency  fault tolerance  Limited achievement of:  concurrency  replication  consistency  security 44

46. Distributed System and Middleware Example: Sun NFS (1985)  Structure of flat file, client & directory service  NFS protocol  RPC based, OS independent (originally UNIX)  NFS server  stateless (no open/close)  no locks or concurrency control  no replication with updates  Virtual file system, Remote mount  Access control (user id with each request)  security loophol  modify RPC to impersonate users  Client and Server caching 45

47. Distributed System and Middleware Sun NFS architecture UNIX kernel protocol Client computerServer computer system calls LocalRemote UNIX file system NFS client NFS server UNIX file system Application program Application program NFS UNIX UNIX kernel Virtual file system Other file system Operations on remote files 46

48. Distributed System and Middleware File identifier (FileId)  Simple Solution  i-node (number identifying file within file system)  file migration requires finding and changing all FileIds  UNIX reuses i-node numbers after file is deleted (i-generation. no)  NFS file handle  Virtual file system uses i-node if local, file handle(fh) if remote. Server addressIndex IP address.socketi-node number File system identifieri-node gener. no.i-node no. File handle(fh) fh = file handle: Filesystem identifieri-node numberi-node generation no 47

49. Distributed System and Middleware NFS Server Operations (simplified) read(fh, offset, count) -> attr, data write(fh, offset, count, data) -> attr create(dirfh, name, attr) -> newfh, attr remove(dirfh, name) -> status getattr(fh) -> attr setattr(fh, attr) -> attr lookup(dirfh, name) -> fh, attr rename(dirfh, name, todirfh, toname) link(newdirfh, newname, dirfh, name) readdir(dirfh, cookie, count) -> entries symlink(newdirfh, newname, string) -> status readlink(fh) -> string mkdir(dirfh, name, attr) -> newfh, attr rmdir(dirfh, name) -> status statfs(fh) -> fsstats fh = file handle: Filesystem identifieri-node numberi-node generation no i-node contains information of files 48

50. Distributed System and Middleware Caching in NFS  Indispensable for performance (necessary)  Caching  Retains recently the used data (file pages, directories, file attributes) in cache  updates data in cache for speed  block size typically 8kbytes  Server caching  cache in server memory (UNIX kernel)  Client caching  cache in client memory, local disk 49

51. Distributed System and Middleware Server caching  Store data in server memory  Read-ahead: anticipate which pages to read  Delayed write  update in cache; write to disk periodically (UNIX sync to synchronize cache) or when space needed  which contents seen by users depends on timing  Write through  cache and write to disk (reliable, poor performance), whenever updated  Write on close  write to disk only when commit received (fast but problems with files open for a long time) 50

52. Distributed System and Middleware Client caching  Potential consistency problems!  different versions, portions of files, … since writes delayed  clients poll server to check if copy still valid  Timestamp method  Tag with latest time of validity check and modification time  copy valid if time since last check less than freshness interval, or modification time on server the same  choose freshness interval adaptively, 3~30 sec for files, 30~60 sec for directories  for small freshness interval, potential heavy load on Network 51

53. Distributed System and Middleware Client caching ctd  Reads  perform validity check whenever cache entry(input) used  if not valid, request data from server  several optimizations to reduce traffic  Recent updates not always visible (timing!)  Writes  when page modified, marked as dirty  dirty pages flushed asynchronously, periodically (client’s synch) and on close  Not truly one-copy update semantics... 52

54. Distributed System and Middleware NFS summary  Transparency  Access transparency  providing application programming interface(= local system interface)  Location transparency  supporting a single network file name space  Mobility transparency  migration transparency  Scalability  To handle very large-world loads efficiently  File replication  NSF : read-only replica  supporting file replication with updates  Hardware and operating system - heterogeneity  Fault tolerance 53

55. Distributed System and Middleware Example: Andrew File System(AFS)  Overview  A distributed computing environment (Andrew) under development since 1983 at Carnegie-Mellon University, purchased by IBM and released as Transarc DFS, now open sourced as OpenAFS.  Information sharing on a large scale via transparency  NFS compatible(called NSF-2)  File reference by NFS-style file handle 54

56. Distributed System and Middleware  AFS tries to solve complex issues such as uniform name space, location- independent file sharing, client-side caching (with cache consistency), secure authentication (via Kerberos)  Also includes server-side caching (via replicas), high availability  Can span 5,000 workstations  Scalable  Whole-file serving (> 64kbytes)  Whole-file caching (on local client disk, 100s of recently used files)  Characteristics of AFS  local-cached copy  providing sufficient cache storage  UNIX based on file size and referencing locality DISTRIBUTED FILE SYSTEMS

57. Distributed System and Middleware AFS Software architecture  Two software components  Vice(user-level UNIX processing running in server, server module)  Venus( user-level process running in a client, client module) 56

58. Distributed System and Middleware SHARED NAME SPACE:  The server file space is divided into volumes. Volumes contain files of only one user. It's these volumes that are the level of granularity attached to a client.  A vice file can be accessed using a fid =. The fid doesn't depend on machine location. A client queries a volume-location database for this information.  Volumes can migrate between servers to balance space and utilization. Old server has "forwarding" instructions and handles client updates during migration.  Read-only volumes ( system files, etc. ) can be replicated. The volume database knows how to find these. DISTRIBUTED FILE SYSTEMS Andrew File System

59. Distributed System and Middleware FILE OPERATIONS AND CONSISTENCY SEMANTICS:  Andrew caches entire files form servers A client workstation interacts with Vice servers only during opening and closing of files  Venus – caches files from Vice when they are opened, and stores modified copies of files back when they are closed  Reading and writing bytes of a file are done by the kernel without Venus intervention on the cached copy  Venus caches contents of directories and symbolic links, for path-name translation  Exceptions to the caching policy are modifications to directories that are made directly on the server responsibility for that directory DISTRIBUTED FILE SYSTEMS Andrew File System

60. Distributed System and Middleware  Clients have a partitioned space of file names: a local name space and a shared name space  Dedicated servers, called Vice, 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  Servers collectively are responsible for the storage and management of the shared name space DISTRIBUTED FILE SYSTEMS Andrew File System

61. Distributed System and Middleware  Clients and servers are structured in clusters interconnected by a backbone LAN  A cluster consists of a collection of workstations and a cluster server and is connected to the backbone by a router  A key mechanism selected for remote file operations is whole file caching  Opening a file causes it to be cached, in its entirety, on the local disk DISTRIBUTED FILE SYSTEMS Andrew File System

62. Distributed System and Middleware IMPLEMENTATION – Flow of a request:  Deflection of open/close:  The client kernel is modified to detect references to vice files.  The request is forwarded to Venus with these steps:  Venus does pathname translation.  Asks Vice for the file  Moves the file to local disk  Passes inode of file back to client kernel.  Venus maintains caches for status ( in memory ) and data ( on local disk.)  A server user-level process handles client requests.  A lightweight process handles concurrent RPC requests from clients.  State information is cached in this process.  Susceptible to reliability problems. DISTRIBUTED FILE SYSTEMS Andrew File System

63. Distributed System and Middleware New developments -ctd  AFS enhancements  DCE/DFS standards, adopts a similar Spritely NFS and NQNFS to callbacks  improving in storage organization  Redundant array of inexpensive(RAID)  Log-structure file storage(LFS)  New design approaches(UC of Berkeley)  xFS (serverless network architecture, file serving responsibility distributed across LAN)  Frangipni( h igh scalable distributed file system, Digital System Research Center, 1997) 62

64. Distributed System and Middleware Summary  File service  crucial to the running of a distributed system  performance, consistency and easy recovery essential  Design issues  separate flat file service from directory service and client module  stateless for performance and fault-tolerance  caching for performance  concurrent updates difficult with caching  approximation of one-copy update semantics  Case studies  SUN-NFS  AFS  Recent advances 63