Parallel and Distributed Computing Chapter 6
- 3. Consistency Model Example: • Assume that the following case occurs: • The row X is replicated on nodes M and N • The client A writes row X to node M • After a period of time t, client B reads row X from node N • The consistency model determines whether client B will definitely see the write performed by client A, will definitely not, or cannot depend on seeing the write.
- 4. Conflicting operations • Read-write conflict: a read operation and a write operation act concurrently • Write-Write conflict: two concurrent write operation
- 5. Reasons for replication Two primary reasons for replicating data reliability and performance However there is price to be paid when data is replicated The problem is with replication that having multiple copies may lead to consistency model i.e. when copy is modified that copy becomes difference from the rest
- 6. Reasons for replication Therefore modifications have to be caried out on all copies to ensure consistency. Exactly how and when those modifications need to be carried out determines the price of replication.
- 7. Consistency Model • Provides formal view of how memory system will appears to programmer • Eliminates the gap between the behavior expected by the programmer and actual behavior supported by the system • It is used for contract between software and memory • It says that if the software agrees to obey certain rule then the memory promise to work properly • Memory consistence model for shared address space specifies that in which memory operations must appear to be performed i.e. visible to processors with respect to one other.
- 8. Consistency Model Consistency is different from coherence. Coherence deals with maintaining a global order in which writes to a single location or single variable are seen by all processors i.e. how memory accesses are coordinated among CPUs. Consistency deals with the ordering of operations to multiple locations with respect to all processors. i.e. when memory writes shows up at another CPU.
- 9. Consistency Model Consistency models define rules for the apparent order and visibility of updates. There are two methods to define and a categorize consistency models 1. issue: describes the restrictions that define how a process can issue operations 2. View: defines the order of operations visible to processes.
- 10. Consistency Model •Consistency Model: •Data Centric Consistency Model •Client Centric Consistency Model
- 11. Consistency Models 1. Strict Consistency Model 2. Sequential Consistency Model 3. Causal Consistency Model 4. PRAM and Processor 5. Weak Consistency Model 6. Eventual Consistency Model 7. Release Consistency Model
- 12. Strict Consistency Model 1. It Is most strongest consistency model, i.e. model of rules 2. A write to a variable by any processor needs to be seen instantaneously by all processors. 3. It is defined by following conditions 4. Any read to memory location x returns the value stored by the most recent write operation of x 5. Disadvantage: number of messages increase between processors, as single write operation may need to update on all processors 6. We can’t implement this model in all scenario. 7. It doesn't help in answering the question of conflict resolution in concurrent writes to the same data item, because it assumes concurrent writes to be impossible
- 13. Sequential Consistency Model 1. It is weaker memory model than strict consistency model. 2. It is the most intuitive (with out conscious) model greatly restrict the use of the many performance, optimization used by the uniprocessor hardware and compiler designers 3. Results of any execution is same as if the operations of all the processors were executed in some sequential order and the operations of each individual process appears in the sequence in order specified by its program
- 14. Sequential Consistency Model 1. Note that this rule says nothing about what order the events happen in 2. Just that happen in some order 3. All instructions are executed atomically and there are no recording between memory instructions 4. This is simple model however performance can’t be at its best 5. One difficulty to bring all operations execute in sequential order.
- 15. Sequential vs strict Consistency model
- 16. Causal consistency model 1. It represents a weaking of sequential consistency 2. It makes a distinguish between events that are potentially causally related to and those are not. 3. Condition: writes that are potentially causally related must be seen by all processing the same order. 4. Operations are said to related to another if one might have bee influenced by other operations 5. Operations that are not causally related are said to be concurrent.
- 18. Weak consistency model 1. Not necessary to show the change in memory done by every write operations to other processes. E.g. when a process executes in a critical sections. 2. Process to synchronization variables are sequentially consistent 3. No access to a synchronization variable allowed to perform until all previous writes are complete every where. 4. No data access write/read is allowed to be performed until all process synchronize variables has been performed
- 19. Eventual Consistency Model • Based on client centric Model • Many system: one or few processes perform updates, how frequently these updates be made available to other read-only process. • Examples: NIS (News information service), user information database, only system admin update database, users only read data. • Only naming authority allowed updates (hence no write-write conflicts)
- 20. Monotonic Reads • If a process reads the value of a data item x, any successive read operation on x by that process will always return that same value or a more recent value. • Once read, subsequent reads on that data items return same or more recent values. • Example: Each time you connect to a different email server, the server fletches all the updates form the server you previously visited. • Example: automatically reading you personal calendar updates from different servers. Monotonic reads guarantees that the user see all updates, no matter from which server the automatic reading takes place.
- 21. Monotonic Write A write operation by a process on a data item x is completed before any successive write operation on x by the same process. a write must be propagated to all replicas before a successive write by the same process Example: Updating a program at server S2, And ensuring that all components on which compilation and linking depends are also placed at S2. Example: maintaining versions of replicated files in the correct order everywhere (Propagate the previous version to the server where the newest version is installed)
- 22. PRAM (Pipelined RAM) consistency model 1. Weaker consistency model so for 2. Only write operations is considers 3. Writes done by a single processor are received by all the other processors in the order in which they were issued 4. Writes generates by different processors are concurrent. 5. Write operations performed by single process are in pipelined 6. Simple and easy to implement 7. Also called as FIFO Consistency Model
- 23. Released Consistency Model 1. Enhancement of weak consistency model. In which operations are performed in a synchronized way. 2. Use of two synchronization variables a) Acquire(use to tell system is entering) b) Release(use to tell system is just exited)