1. By Shruti poundarik

2.  Data Objects and Files are replicated to increase system performance and availability.  Increased system performance achieved through concurrent access of replicas.  High availability of data due to redundancy of data objects.  Parallelism and failure transparencies are desirable in distributed systems.  Not useful unless replication and concurrency transparency is provided.

3.  In database systems, atomicity is one of the ACID transaction properties. An atomic transaction is a series of database operations which either all occur, or all do not occur[1].  All or nothing.  In DFS (Distributed File System), replicated objects (data or file) should follow atomicity rules, i.e., all copies should be updated (synchronously or asynchronously) or none

4.  One-copy serializability: The effect of transactions performed by clients on replicated objects should be the same as if they had been performed one at a time on a single set of objects.[2]

5.  (FSA) File service agent, client interface  (RM) replica manager, provide replication functions [3].  Client chooses one or more FSA to access data object.  FSA acts as front end to replica managers RMs to provide replication transparency.  FSA contacts one or more RMs for actual updating and reading of data objects.

7. Three options for Read operations.  Read-one-primary, FSA only read from a primary RM to enforce consistency.  Read-one, FSA may read from any RM for concurrency.  Read-quorum, FSA must read from a quorum of RMs to decide the currency of data

8.  Object access operations may be Read or Update.  In this architecture read operation needs to be addressed to one of the replicas.  Replica’s transparent to the client.  File Services invoked by the client may be required by RM protocol to ensure data read is most recent.

9.  From systems view point write operations should be addressed to all replicas automatically. Scenarios for Write:-  Write-one-primary:-Only write to primary RM, primary RM update all other RMs  Write-all:- update to all RMs.  Write-all- available:- Write to all functioning RMs. Faulty RM need to be synched before bring online.

10.  Write-quorum:- Update to a predefined quorum of RMs  Write-gossip :- Update to any RM and lazily propagated to other RMs.

11. ◦ Both Read and write operations must be directed to primary replica manager. ◦ No replication issue. ◦ All operations are serialized by the primary RM. ◦ Secondary RMs supply redundancy in case of primary failures. ◦ Consistency is easy to achieve but not concurrency.

12.  To provide Concurrency, read operation performed at any RM site.  leads to Coherency problem since propagation from one RM to the other secondary RM leads to communication delay.  Therefore the propagation of updates must be made atomic.  Updates can be initiated at any RM,preferably the one closer to the requesting client.  Provides concurrency and coherency

13.  Achieves one copy serializability,execution of transaction on replicated objects is equivalent to execution of same transaction on non replicated objects.  In this the data objects are replicated to faulty and non faulty replicas.  This contradicts the purpose of replication,as atomic updates should be made available to non faulty replicas.

14.  Variation of Read One write All.  Atomic updates should be made available to non faulty replicas.  Therefore one copy serializability gets slightly complicated.

15.  Each read operation to replicated data object d must obtain a read quorum R(d) to perform read.  Each write operation needs a write quorum W(d) to complete write.  Client gets most recently completed update of data as version number attached to replicated object.  Read operation queries all R(d) replicas, replica with highest version number is returned.  Write operation advances version by 1.

16.  Updates are less frequent than reads,updates can be propagated lazily to replicas.  This read-one/write-gossip approach is gossip update propagation protocol  Both read and update operations are directed by FSA to any RM  FSA shields replication details from clients.

17.  Main purpose is to support high availability in an environment where failure of replicas are likely.  Disadvantages of File replication:- ◦ Contents of the file needs to be known before replication operation takes place. ◦ Existing System cant work in limited bandwidth networks. ◦ DFS replication will not work well when there are large number of changes to replicate [4].

18.  File Replication is optimized over limited bandwidth networks using remote differential compression.[5]  RDC Remote Differential Compression protocol heuristically negotiates a set of differences between a recipient and sender that have two sufficiently similar versions of the same file.  RDC optimizes communication between sender and recipient by having both sides sub divide all of the files into chunks and compute strong checksums or signatures for each chunks.

19.  RDC needs to be applied to compressed chunk files.  Windows Server uses Remote Differential Compression to propagate change only to save bandwidth [4].

20. [1] Wikipedia; http://en.wikipedia.org/wiki/Atomicityhttp://en.wikipedia.org/wiki/Atomicity [2] M. T. Harandi;J. Hou (modified: I. Gupta);"Transactions with Replication";http://www.crhc.uiuc.edu/~nhv/428/slides/repl-trans.ppt [3] Randy Chow,Theodore Johnson, “Distributed Operating Systems & Algorithms”, 1998 [4] "Overview of the Distributed File System Solution in Microsoft Windows Server 2003 http://technet2.microsoft.com/WindowsServer/en/library/d3afe6ee-3083- 4950-a093-8ab748651b761033.mspx?mfr=true [5 ] “Optimizing File Replication over Limited-Bandwidth Networks using Remote Differential Compression” IEEE Infocom Conference, 2006.