Distributed Shared Memory Systems

Distributed Shared Memory Systems

• 2.
  Distributed Shared Memory Systems -byAnkit Gupta
• 3.
  What is aDistributed System?
• 4.
  What is DSM? The distributed shared memory (DSM) implements the shared memory model in Distributed Systems, which have no physical shared memory.  The shared memory model provides a virtual address space shared between all nodes.  This overcomes the high cost of communication in distributed systems. DSM systems move data to the location of Access.
• 5.
  Memory Mapping Manager Memory Mapping Manager Memory Mapping Manager Shared Memory NODE 1NODE 2 NODE 3
• 6.
  Purpose Of DSMResearch  Building less expensive parallel machines  Building larger parallel machines  Eliminating the programming difficulty of MPP and Cluster architectures  Generally break new ground:  New network architectures and algorithms  New compiler techniques  Better understanding of performance in distributed systems
• 7.
  Distributed Shared MemoryModels  Object Based DSM  Variable Based DSM  Page Based DSM  Structured DSM  Hardware Supported DSM
• 8.
  Object Based  Objectbased DSM  Probably the simplest way to implement DSM  Shared data must be encapsulated in an object  Shared data may only be accessed via the methods in the object Variable Based  Delivers the lowest distribution granularity  Closely integrated in the compiler  May be hardware supported
• 9.
  Hardware Based DSM Uses hardware to eliminate software overhead  May be hidden even from the operating system  Usually provides sequential consistency  May limit the size of the DSM system
• 10.
  Advantages of DSM (DistributedShared Memory)  Data sharing is implicit, hiding data movement (as opposed to ‘Send’/‘Receive’ in message passing model)  Passing data structures containing pointers is easier (in message passing model data moves between different address spaces)  Moving entire object to user takes advantage of locality difference  Less expensive to build than tightly coupled multiprocessor system: off-the- shelf hardware, no expensive interface to shared physical memory  Very large total physical memory for all nodes: Large programs can run more efficiently  No serial access to common bus for shared physical memory like in multiprocessor systems  Programs written for shared memory multiprocessors can be run on DSM systems with minimum changes
• 11.
  Issues faced indevelopment of DSM  Granularity  Structure of Shared memory  Memory coherence and access synchronization  Data location and access  Replacement strategy  Thrashing  Heterogeneity
• 12.
  Granularity  Granularity isthe amount of data sent with each update  If granularity is too small and a large amount of contiguous data is updated, the overhead of sending many small messages leads to less efficiency  If granularity is too large, a whole page (or more) would be sent for an update to a single byte, thus reducing efficiency
• 13.
  Structure of SharedMemory  Structure refers to the layout of the shared data in memory.  Dependent on the type of applications that the DSM system is intended to support.
• 14.
  Replacement Strategy  Ifthe local memory of a node is full, a cache miss at that node implies not only a fetch of accessed data block from a remote node but also a replacement.  Data block must be replaced by the new data block. - Example: LRU with access modes Private (local) pages to be replaced before shared ones Private pages swapped to disk Shared pages sent over network to owner Read-only pages may be discarded (owners have a copy)
• 15.
  Trashing  Thrashing occurswhen network resources are exhausted, and more time is spent invalidating data and sending updates than is used doing actual work.  Based on system specifics, one should choose write-update or write-invalidate to avoid thrashing.
• 16.
  Memory Coherence andAccess Synchronization  In a DSM system that allows replication of shared data item, copies of shared data item may simultaneously be available in the main memories of a number of nodes.  To solve the memory coherence problem that deal with the consistency of a piece of shared data lying in the main memories of two or more nodes.  DSM are based on - Replicated shared data objects - Concurrent access of data objects at many nodes  Coherent memory: when value returned by read operation is the expected value (e.g., value of most recent write)
• 17.
   Mechanism thatcontrol/synchronizes accesses is needed to maintain memory coherence  Sequential consistency: A system is sequentially consistent if - The result of any execution of operations of all processors is the same as if they were executed in sequential order, and - The operations of each processor appear in this sequence in the order specified by its program  General consistency: - All copies of a memory location (replicas) eventually contain same data when all writes issued by every processor have completed
• 18.
  Algorithms for implementingDSM  The Central Server Algorithm  The Migration Algorithm  The Read-Replication Algorithm  The full-Replication Algorithm
• 19.
  The Central ServerAlgorithm - Central server maintains all shared data  Read request: returns data item  Write request: updates data and returns acknowledgement message - Implementation  A timeout is used to resend a request if acknowledgment fails  Associated sequence numbers can be used to detect duplicate write requests  If an application’s request to access shared data fails repeatedly, a failure condition is sent to the application - Issues: performance and reliability - Possible solutions  Partition shared data between several servers  Use a mapping function to distribute/locate data
• 20.
  The Migration Algorithm -Operation  Ship (migrate) entire data object (page, block) containing data item to requesting location  Allow only one node to access a shared data at a time - Advantages  Takes advantage of the locality of reference  DSM can be integrated with VM at each node - Make DSM page multiple of VM page size - A locally held shared memory can be mapped into the VM page address space - If page not local, fault-handler migrates page and removes it from address space at remote node - To locate a remote data object:  Use a location server  Maintain hints at each node  Broadcast query - Issues  Only one node can access a data object at a time  Thrashing can occur: to minimize it, set minimum time data object resides at a node
• 21.
  The Read-Replication Algorithm Replicates data objects to multiple nodes  DSM keeps track of location of data objects  Multiple nodes can have read access or one node write access (multiple readers-one writer protocol)  After a write, all copies are invalidated or updated  DSM has to keep track of locations of all copies of data objects. Examples of implementations:  IVY: owner node of data object knows all nodes that have copies  PLUS: distributed linked-list tracks all nodes that have copies  Advantage  The read-replication can lead to substantial performance improvements if the ratio of reads to writes is large
• 22.
  The Full-Replication Algorithm -Extension of read-replication algorithm: multiple nodes can read and multiple nodes can write (multiple-readers, multiple-writers protocol) - Issue: consistency of data for multiple writers - Solution: use of gap-free sequencer • All writes sent to sequencer • Sequencer assigns sequence number and sends write request to all sites that have copies • Each node performs writes according to sequence numbers • A gap in sequence numbers indicates a missing write request: node asks for retransmission of missing write requests
• 23.
  Any Questions?