Distributed Shared Memory-jhgfdsserty.pdf

Distributed Shared Memory
( DSM )

Distributed shared memory
DSM paradigm provides process with shared address space
 Primitives for shared memory:
– Read(address)
– Write(address , data)
Shared memory paradigm gives the systems illusion of
physically shared memory
DSM refers to shared memory paradigm applied to loosely
coupled distributed memory systems

Cont….
Shared memory exists only virtually
Similar concept to virtual memory
DSM also known as DSVM
DSM provides a virtual address space shared among
processes on loosely coupled processors
DSM is basically an abstraction that integrates the local
memory of different machine into a single logical entity
shared by cooperating processes

DSM architecture
Each node of the system consist of one or more CPUs and
memory unit
Nodes are connected by high speed communication
network
Simple message passing system for nodes to exchange
information
Main memory of individual nodes is used to cache pieces
of shared memory space

Cont….
Memory mapping manager routine maps local memory to
shared virtual memory
Shared memory of DSM exist only virtually
Shared memory space is partitioned into blocks
Data caching is used in DSM system to reduce network
latency
The basic unit of caching is a memory block
The missing block is migrate from the remote node to the
client process’s node and operating system maps into the
application’s address space

Cont….
Data block keep migrating from one node to another on
demand but no communication is visible to the user
processes
If data is not available in local memory network block
fault is generated.

Design and implementation issues
Granularity
 Structure of Shared memory
 Memory coherence and access synchronization
 Data location and access
Replacement strategy
 Thrashing
 Heterogeneity

Cont….
Granularity:
Granularity refers to the block size of DSM
The unit of sharing and the unit of data transfer
across the network when a network block fault occurs
Possible unit are a few word , a page or a few pages
Structure of Shared memory:
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

Cont…
Memory coherence and access 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

Cont….
Data location and access:
To share data in a DSM, should be possible to locate and
retrieve the data accessed by a user process.
Replacement strategy:
If the 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

Cont…
Thrashing:
Data block migrate between nodes on demand. Therefore
if two nodes compete for write access to a single data item
the corresponding data block may be transferred back.
Heterogeneity:
The DSM system built for homogeneous system need not
address the heterogeneity issue

Granularity
Most visible parameter in the design of DSM system is
block size
Factors influencing block size selection:
Sending large packet of data is not much more expensive
than sending small ones
Paging overhead: A process is likely to access a large
region of its shared address space in a small amount of
time
Therefore the paging overhead is less for large block size
as compared to the paging overhead for small block size

Cont…
Directory size:
The larger the block size, the smaller the directory
Ultimately result in reduced directory management
overhead for larger block size
Thrashing:
The problem of thrashing may occur when data item in
the same data block are being updated by multiple node at
the same time
Problem may occur with any block size, it is more likely
with larger block size

Cont…
False sharing:
Occur when two different processes access two unrelated
variable that reside in the same data block

Cont…
The larger is the block size the higher is the probability of
false sharing
False sharing of a block may lead to a thrashing problem

Cont…
Using page size as block size:
Relative advantage and disadvantages of small and large
block size make it difficult for DSM designer to decide on
a proper block size
Following advantage:
It allows the use of existing page fault schemes to trigger a
DSM page fault
It allows the access right control
Page size do not impose undue communication overhead
at the time of network page fault
Page size is a suitable data entity unit with respect to
memory contention

Structure of shared-memory space
Structure defines the abstract view of the shared memory
space
The structure and granularity of a DSM system are closely
related
Three approach:
No structuring
Structuring by data type
Structuring as a database

Cont…
No structuring:
The shared memory space is simply a linear array of words
Advantage:
Choose any suitable page size as the unit of sharing and a
fixed grain size may be used for all application
Simple and easy to design such a DSM system

Cont….
Structuring by data type:
The shared memory space is structured either as a
collection of variables in the source language
The granularity in such DSM system is an object or a
variable
DSM system use variable grain size to match the size of
the object/variable being accessed by the application

Cont…
Structuring as a database:
Structure the shared memory like a database
Shared memory space is ordered as an associative memory
called tuple space
To perform update old data item in the DSM are replaced
by new data item
Processes select tuples by specifying the number of their
fields and their values or type
Access to shared data is nontransparent. Most system they
are transparent

Consistency Models
Consistency requirement vary from application to
application
A consistency model basically refers to the degree of
consistency that has to be maintained for the shared
memory data
Defined as a set of rules that application must obey if they
want the DSM system to provide the degree of consistency
guaranteed by the consistency model
If a system support the stronger consistency model then
the weaker consistency model is automatically supported
but the converse is not true

Consistency Models
Types:
Strict Consistency model
Sequential Consistency model
Causal consistency model
Pipelined Random Access Memory consistency
model(PRAM)
Processor Consistency model
Weak consistency model
Release consistency model

Strict consistency model
This is the strongest form of memory coherence having
the most stringent consistency requirement
Value returned by a read operation on a memory address
is always same as the value written by the most recent
write operation to that address
All writes instantaneously become visible to all processes
Implementation of the strict consistency model requires
the existence of an absolute global time

Cont….
Absolute synchronization of clock of all the nodes of a
distributed system is not possible
Implementation of strict consistency model for a DSM
system is practically impossible
If the three operations read(r1), write(w1), read(r2) are
performed on a memory location in that order
Only acceptable ordering for a strictly consistency
memory is (r1, w1, r2)

Sequential Consistency model
Proposed by Lamport [1979]
A shared memory system is said to support the sequential
consistency model if all processes see the same order
Exact order of access operations are interleaved does not
matter
If the three operations read(r1), write(w1), read(r2) are
performed on a memory location in that order
Any of the orderings (r1, w1, r2), (r1, r2, w1), (w1, r1, r2), (w1,
r2, r1), (r2, r1, w1), (r2, w1, r1) is acceptable provided all
processes see the same ordering

Cont…
The consistency requirement of the sequential consistency
model is weaker than that of the strict consistency model
 A sequentially consistency memory provide one-copy
/single-copy semantics
 sequentially consistency is acceptable by most
applications

Causal Consistency Model
Proposed by Hutto and Ahamad (1990)
All processes see only those memory reference operations
in the correct order that are potentially causally related
Memory reference operations not related may be seen by
different processes in different order
Memory reference operation is said to be related to
another memory reference operation if one might have
been influenced by the other
Maintaining dependency graphs for memory access
operations

Pipelined Random Access Memory
Consistency model
Proposed by Lipton and Sandberg (1988)
Provides a weaker consistency semantics than the
consistency model described so far
Ensures that all write operations performed by a single
process are seen by all other processes in the order in
which they were performed
All write operations performed by a single process are in a
pipeline
Write operations performed by different processes can be
seen by different processes in different order

Cont…
If w11 and w12 are two write operations performed by a
process P1 in that order, and w21 and w22 are two write
operations performed by a process P2 in that order
A process P3 may see them in the order [(w11,w12), (w21,w22)]
and another process P4 may see them in the order [(w21,w22),
(w11,w12)]
Simple and easy to implement and also has good performance
PRAM consistency all processes do not agree on the same
order of memory reference operations

Processor consistency model
Proposed by Goodman [1989]
Very similar to PRAM model with additional restriction of
memory coherence
Memory coherence means that for any memory location
all processes agree on the same order of all write
operations performed on the same memory location (no
matter by which process they are performed) are seen by
all processes in the same order
If w12 and w22 are write operations for writing the same
memory location x, all processes must see them in the
same order- w12 before w22 or w22 before w12
Processes P3 and P4 must see in the same order, which
may be either [(w11,w12), (w21,w22)] or [(w21,w22),
(w11,w12)]

Weak consistency model
Proposed by Dubois [1988]
Common characteristics to many application:
1 It is not necessary to show the change in memory done by
every write operation to other processes eg. when a
process executes in a critical section
2 Isolated accesses to shared variable are rare
 Better performance can be achieved if consistency is
enforced on a group of memory reference operations
rather than on individual memory reference operations

Cont…
DSM system that support the weak consistency model
uses a special variable called a synchronization variable
Requirements:
1. All accesses to synchronization variables must obey
sequential consistency semantics
2. All previous write operations must be completed
everywhere before an access to a synchronization variable
is allowed
3. All previous accesses to synchronization variables must be
completed before access to a nonsynchronization variable
is allowed

Release consistency model
Enhancement of weak consistency model
Use of two synchronization variables
Acquire (used to tell the system it is entering CR)
Release (used to tell the system it has just exited CR)
Acquire results in propagating changes made by other
nodes to process's node
Release results in propagating changes made by the
process to other nodes

Cont….
Barrier defines the end of a phase of execution of a group
of concurrently executing processes
Barrier can be implemented by using a centralized barrier
server
Requirements:
1 All accesses to acquire and release synchronization
variable obey processor consistency semantics
2 All previous acquires perform by a process must be
completed successfully before the process is allowed to
perform a data access operation on the memory

Cont…
3 All previous data access operations performed by a
process must be completed successfully before a release
access done by the process is allowed
 A variation of release consistency is lazy release
consistency proposed by Keleher [1992]

Implementing sequential
consistency model
Most commonly used model
Protocols for implementing the sequential consistency
model in the DSM system depend to a great extent on
whether the DSM system allows replication and/or
migration of shared memory data blocks
Strategies:
Nonreplicated, Nonmigrating blocks (NRNMB)
Nonreplicated, migrating blocks (NRMB)
Replicated, migrating blocks (RMB)
Replicated, Nonmigrating blocks (RNMB)

NRNMBs
Simplest strategy for implementing a sequentially
consistency DSM system
Each block of the shared memory has a single copy whose
location is always fixed

Cont…
Enforcing sequential consistency is simple in this case
Method is simple and easy to implement and suffers
following drawback:
 Serializing data access creates a bottleneck
 Parallelism, which is a major advantage of DSM is not
possible with this method
Data locating in the NRNMB strategy :
There is a single copy of each block in the entire system
The location of a block never changes
Hence use Mapping Function

NRMBS
Each block of the shared memory has a single copy in the
entire system
In this strategy only the processes executing on one node
can read or write a given data item at any one time and
ensures sequential consistency

Cont…
Advantage :
No communications cost are incurred when a process
accesses data currently held locally
It allows the applications to take advantage of data access
locality
Drawbacks:
It is prone to thrashing problem
The advantage of parallelism can not be availed in this
method also

Data locating in the NRMB strategy
There is a single copy of each block, the location of a
block keeps changing dynamically
Following method used :
1. Broadcasting
2.Centralized server algorithm
3. Fixed distributed server algorithm
4.Dynamic distributed server algorithm

Broadcasting
Each node maintains an owned blocks table that contains
an entry for each block for which the node is the current
owner

Cont….
When a fault occurs , the fault handler of the faulting
node broadcasts a read/write request on the network
Disadvantage:
It does not scale well

Cont…
A centralized server maintains a block table that contains
the location information for all block in the shared
memory space
Drawback:
A centralized server serializes location queries, reducing
parallelism
The failure of the centralized server will cause the DSM
system to stop functioning

Fixed distributed server algorithm

Cont..
Scheme is a direct extension of the centralized server
scheme
It overcomes the problems of the centralized server
scheme by distributing the role of the centralized server
Whenever a fault occurs, the mapping functions is used by
the fault handler of the faulting node to find out the node
whose block manager is mapping the currently accessed
block

Dynamic distributed-server algorithm

Cont…
Does not use any block manager and attempts to keep
track of the ownership information of all block in each
node
Each node has a block table that contains the ownership
information for all block
A field gives the node a hint on the location of the owner
of a block and hence is called the probable owner
When fault occurs, the faulting node extracts from its
block table the node information

Replicated, migrating blocks
A major disadvantage of the non replication strategies is
lack of parallelism
To increase parallelism, virtually all DSM system replicate
blocks
Replication tends to increase the cost of write operation
because for a write to a block all its replica must be
invalidated or updated to maintain consistency
If the read/write ratio is large, the extra expense for the
write operation may be more than offset by the lower
average cost of the read operation

Cont…
Two basic protocols that may be used for ensuring
sequential consistency in this case are:
write-invalidate:
When a write fault occurs fault handler copies the
accessed block
If one of the nodes that had a copy of the block before
invalidation tries to perform a memory access operation
(read/write) on the block after invalidation, a cache miss
will occur and the fault handler of that node will have to
fetch the block again from a node having a valid copy of
the block, therefore the scheme achieves sequential
consistency

Write-update
A write operation is carried out by updating all copies of
the data on which the write is perform
When write fault occurs the fault handler copies the
accessed block from one of the block’s current node to its
owe node
The write operation completes only after all the copies of
the block have been successfully updated
Sequentially consistency can be achieved by using a
mechanism to totally order the write operations of all the
node
The intended modification of each write operation is first
sent to the global sequencer

Cont…
Sequence number to the modification and multicasts the
modification with this sequence number to all the nodes
where a replica of the data block to be modified is located
The write operations are processed at each node in
sequence number order
If the verification fails, node request the sequencer
for a retransmission of the missing modification
Write-update approach is very expensive
In write-invalidate approach, updates are only
propagated when data are read and several updates
can take place before communication is necessary

Cont…
Read request:
If there is a local block containing the data and if it is
valid, the request is satisfied by accessing the local copy of
data
Otherwise, the fault handler of the requesting node
generates a read fault
Write request:
If there is a local block containing the data and if it is valid
and writable, the request is immediately satisfied by
accessing the local copy of the data
Otherwise, the fault handler of the requesting node
generates a write fault and obtain a valid copy of the block

Data Locating in the RMB strategy
Data-locating issues are involved in the write-invalidate
protocol used with the RMB strategy:
1. Locating the owner of a block, the most recent node to have
write access to it
2. Keeping track of the node that are currently have a valid
copy of the block
Following algorithms may be used:
1. Broadcasting
2. Centralized-server algorithm
3. Fixed distributed-server algorithm
4. Dynamic distributed-server algorithm

Each node has an owned block table
This table of a node has an entry for each block for which
the node is the owner
Each entry of this table has a copy-set field, contains a
valid copy of the corresponding block
When the read fault occurs, the faulting node sends a
broadcast read request on the network
When a write fault occurs the node send a broadcast write
request on the network

Cont…
The copy-set field of the entry is initialized to
indicate that there are no other copies of the block
Method suffers from disadvantage mentioned during
the description of the data-locating mechanism for
the NRMB strategy

Cont…
This is similar to the centralized-server algorithm of the
NRMB strategy
Each entry of the block table, managed the centralized
server, has an owner-node field
Copy-set field that contains a list of nodes having a valid
copy of the block
When read/write fault occurs, the node send a read/write
fault request for the accessed block to the centralized
server

Fixed distributed-server algorithm

Cont…
This scheme is a direct extension of the centralized-server
scheme
Role of the centralized server is distributed to several
distributed servers
There is a block manager on several node
When fault occurs, the mapping function is used to find
the location of the block manager
Advantage is same as the data locating mechanism for the
NRMB strategy

Cont…
Similar manner as the dynamic distributed server
algorithm of the NRMB strategy
Each node has a block table that contains an entry for all
block in the shared memory space
When a fault occurs, the fault handler of the faulting
node extracts the probable owner node information, send
a request for the block to that node
 On receiving the block and copy set information, node
and sends an in validation request to all nodes in the copy-
set.

Cont…
To reduce the length of the chain of node to be traversed
to reach the true owner of a block
Block table is updated as follows:
1. whenever the node receives and invalidation request
2. whenever the node relinquishes ownership, that is on a
write fault.
3. whenever the node forwards a fault request.
 First two cases the probable owner field is changed
Third case the probable owner field is changed to the
original faulting node.

Replicated, Nonmigrating Block
A shared memory block may be replicated at multiple
node of the system but the location of each replica is
fixed .
 All replicas of a block are kept consistent by updating
them all in case of a write excess.
 Sequential consistency is ensured by using a global
sequencer to sequence the write operation of all nodes.

Data locating in the RNMB strategy
 Following characteristics:
1. The replica location of a block never change
2. All replicas of a data block are kept consistent
3. Only a read request can be directly send to one of the
node having a replica.

Cont….
 A write operation on a block is sent to sequencer
 The sequencer assign the next sequence number to
the requested modification
It then multicast the modification with this sequence
number to all the nodes listed in the replica set field.

Munin: A Release Consistent DSM System
Structure of Shared-Memory Space:
 The shared memory space of Munin is structured as a
collection of shared variables.
 The shared variables are declared with the keywords
shared so the compiler can recognized them.
 Each shared variable, by default, is placed by the compiler
on a separate page.
 Unit of data transfer across the network by MMU
hardware.

Implementation of Release
Consistency
 Release consistency application must be modeled around
critical sections.
 Munin provides two such synchronization mechanism.
 A locking mechanism and a barrier mechanism.
The locking mechanism uses lock synchronization
variables with acquirelock and releasedlock primitives for
accessing these variables.
The acquirelock primitive with a lock variable as its
parameter is executed by a process to enter a critical
section .

Cont…
Releaselock primitive with the same lock variable
parameter is executed by the process to exit from the
critical section.
When a process makes an acquirelock request the system
first checks if the lock variable on the local node.
 If not, the probable owner mechanism is used to find the
location owner of the lock variable and request is sent to
that node.

Cont….
 The barrier mechanism uses barrier synchronization
variables with a waitAtBarrier primitive for accessing these
variables.
Barriers are implemented by using the centralized barrier
server mechanism.
 In a network fault occur, the probable owner based
dynamic distributed server algorithm is used in Munin to
locate a page containing the accessed shared variables.

Annotation for shared variables
 The release consistency of Munin allow applications to
have better performance than sequentially consistent DSM
system.
 For further performance improvements, Munin defines
several standard annotation for shared variables.
The standard annotations and consistency protocol for
variables for variable for each type are as follows:

Cont…
1. Read-only
2. Migratory
3. Write-shared
4. Producer-consumer
5. Result
6. Reduction
7. Conventional

Cont…
Read-only
 Shared-data variables annotated as read-only are immutable
data items
 These variable are read but never written after initialization
 Read-only variables are protected by the MMU hardware
Migratory
 Shared variable that are accessed in phases, where each
phase correspondence to a series of accesses by a single
process, may be annotated as migratory variables
 The locking mechanism is used to keep migratory variable
consistent

Cont…
 Page migrate from one node to another on a demand basis,
but page are not replicated
 Only one copy of a page containing a migrating variable
exists in the system
 Write-shared:
 A programmer may use this annotation with a shared
variable to indicate to the system that the variable is
updated concurrently by multiple processes
 Munin avoid the false sharing problem of write shared
variable

Cont…
Write shared variable is replicated on all node
The process may perform several write to the page before
releasing it
When page is released, the system perform a word by
word comparison of the original page
When a node receives the differences of a modified page,
the system check if the local copy of the page was
modified

Cont…
Producer-consumer:
Shared variable that written (producer) by only one
process and read(consumed) by fixed set of other
processes may be annotated to be of producer-
consumer type
Minin uses an “ eager object movement” mechanism
Producer may send several update together by using
locking mechanism
In this case procedure acquires a synchronization
lock, make several update on the variable

Cont…
Result:
Result variable are just the opposite of producer-
consumer variable
They are written by multiple processes but read by only
one process
The variable is read only when all its parts have been
written
Munin uses a special write-update protocol for result
variable

Cont…
Reduction:
Shared variable that must be atomically modified may be
annotated to be of reduction type
 Example: parallel computation application, a global
minimum must be atomically fetched and modified
 For better performance a reduction variable is stored at a
fixed owner
Conventional:
Shared variable that are not annotated as one of the above
types are conventional variables
 Page containing a conventional variable is dynamically
moved to the location of a process that want to perform a
write operation on the variable
 Uses Write Invalidate

Replacement strategy
In DSM system that allow shared memory block to be
dynamically migrated/replicated
Following issue:
1. Which block should be replaced to make space for a newly
required block?
2. Where should the replaced block be placed?

Which block to replace
Classification of replacement algorithms:
1. Usage based verses non-usage based
2. Fixed space verses variable space
Usage based verses non-usage based:
 Uses based algorithms keep track of the history of usage of a
cache line and use this information to make replacement
decisions eg. LRU algorithm
 Non-usage-based algorithms do not take the record of use
of cache lines into account when doing replacement. First in
first out and Random (random or pseudorandom) belong to
this class

Cont…
Fixed space versus variable space:
Fixed-space algorithms assume that the cache size is fixed
while variable space algorithm are based on the
assumption that the cache size can be changed
dynamically depending on the need
In a variable space algorithm, a fetch does not imply a
replacement, and a swap-out can take place without a
corresponding fetch
Variable space algorithms are not suitable for a DSM
system

Cont…
In DSM system of IVY, each memory block of a node is
classified into one of the following five types:
Unused: a free memory block that is not currently
being used
Nil: a block that has been invalidated
Read-only: a block for which the node has only read
access right
Read-owned: a block for which the node has only
read access right but is also the owner of the block
Writable: a block for which the node has write access
permission

Cont…
Based on this classification of block, priority is used:
1. Both unused and nil block have the highest
replacement priority
2. The read-only block have the next replacement priority
3. Read-owned and writable block for which replica(s)
exist on some other node(s) have the next replacement
priority
4. Read-owned and writable block for which only this
node has a copy have the lowest replacement priority

Where to place a replaced block
 Once a memory block has been selected for replacement,
it should be ensured that if there is some useful
information in the block , it should not be lost.
 The two commonly used approaches for storing a useful
block as follow:
 Using secondary store:
 The block is simply transferred on to a local disc.
 Advantages: it does not waste any memory space

Cont…
Using the memory space of other nodes:
 It may be faster to transfer a block over the network than to
transfer it to a local disc.
 Methods require each node to maintain a table of free
memory space in all other nodes.

Thrashing
 Thrashing is said to occur when the system spends a large
amount of time transferring shared data blocks from one
node to another
Thrashing may occur in following situation:
1. When interleaved data accesses made by processes on two
or more nodes
2. When blocks with read only permissions are repeatedly
invalidated soon after they are replicated
Thrashing degrades system performance considerably

Cont…
Methods for solving Thrashing problems:
Providing application controlled locks. locking data to
prevent other node from accessing that data for a short
period of time can reduce Thrashing.
Nailing a block to a node for a minimum amount of time
disallow a block to a be taken away from a node until a
minimum amount of time t elapses after its allocation to
that node.
Drawback:
It is very difficult to choose the appropriate value for the
time.

Other approaches to DSM
There are three main approaches for designing a DSM
system
1. Data caching managed by the operating system
2. Data caching managed by MMU hardware
3. Data caching managed by the language runtime system
In the first approach each node has its own memory and
access to a word in another nodes memory causes a trap to
the operating system then fetches and acquires the page.

Cont….
The second approach is to manage caching by MMU
Used in multiprocessors having hardware caches
DSM implementation is done either entirely or mostly
in hardware
The third approach is manage caching by language
runtime system
The DSM is structured not as a raw linear memory of
bytes from zero to total size of the combined memory
of all machine

Cont…
 Placement and migration of shared variables/objects are
handled by the language runtime system in cooperation
with the operating system.
 Advantage :Programming language may be provided with
features to allow programmers to specify the uses patterns
of shared variables/objects for their application

Heterogeneous DSM
A heterogeneous computing environment allow the
applications to exploit the best of all characteristics
features of several different types of computers.
Measurements made on their experimental prototype
heterogeneous DSM, called Mermaid.
Heterogeneous DSM is not only feasible but can also be
comparable in performance to its homogeneous
counterpart
The two issues in building a DSM system:
 Data conversion and selection of block size
 Machines of different architecture may used different bytes
ordering and floating point representation.

Structuring the DSM system as a collection
of source language objects
It is structured as a collection of variables or object so that
the unit of data migration is an object instead of a block
A suitable conversion runtime is used to translate the
object before migrating its to requesting node.
This method of data conversion is used in the Agora
shared memory system.
Array may easily be too large to be treated as unit of
sharing and data migration.

Allowing only one type of data in a
block
This mechanism is used in Mermaid which uses a
page size as its block size therefore , a page can
contain only one type of data.
Whenever a page is moved between two machines of
different architecture, a routine converts the data in
the page to the appropriate format
Limitations :
Allowing a page to content data of only one type may
lead to wastage of memory due to fragmentation,
resulting increased paging activity

Cont…
Compilers used of different types of machine must be
compatible
Another problem that entire pages are converted even
thought only a small portion may be accessed before it is
transferred away
The mechanism is not fully transparent
Another serious problem associated, accuracy of floating
point value in numerical applications

Block size selection
In the Homogeneous DSM system, the block size is
usually the same size as a native virtual memory (VM)
MMU hardware can be used to trigger a DSM block fault.
In heterogeneous, the virtual memory page size may be
different for machine of different types.
Block size selection become a complicated task
Following algorithm for block size selection:
1. Largest page size algorithm
2. Smallest page size algorithm
3. Intermediate page size algorithm

Cont…
In this method the DSM block size is taken as the largest
VM page size of all machines
Algorithm suffers from the same false sharing and
thrashing problem
The DSM block size is taken as the smallest VM page size
of all machines
This algorithm reduced data contention, its suffer from
the increase communication and block table management
over heads.
To balance between the problem of large and small size
blocks a heterogeneous DSM system, choose largest VM
page size and smallest VM page size

Advantage of DSM
Simpler Abstraction:
The shared memory programming paradigm shields the
application programmers from many such low level concern
Advantage:
 It is simple abstraction its provide to the application
programmers of loosely coupled distributed memory
machine.
Better portability of distributed application
programs:
The access protocol used in case of DSM is consistent with
the way sequential application access data this allows for a
more natural transition from sequential to distributed
application.

Cont…
Better performance of some application:
 The performance of application that use DSM is expected to
worse then if they use message passing directly
 Not always true, and it found that sub application using
DSM can even out perform their message passing
counterparts
This is possible for three reasons
Locality of data : the computation model of DSM is to
make the data more accessible by moving it around
Ultimately results in reduce overall communication cost
for such application

Cont…
On demand data moment:
 The computation data modeled of DSM also facilitates on
demand moment of data as they are being accessed
 Time needed for the data exchange phase is often dictated
by the throughput of existing communication bottlenecks
 On demand data movements facility provided by DSM
eliminates the data exchange phase
Large memory space:
 DSM facility, the total memory size is the sum of the
memory size of all the nodes in the system
 Paging and swapping activities, which involve disk access,
are greatly reduced

Flexible communication environment
The message passing paradigm requires recipients
identification and coexistence of the sender and receiver
processes
The shard memory paradigm of DSM provides a more
flexible communication environment in which the sender
process need not specify the identity of the receiver
processes of data

Ease of process migration
Migration of a process from one node to another in a
distributed system to be tedious and time consuming
The computation model of DSM provides the facility of on
demand migration of data between processors

Distributed Shared Memory-jhgfdsserty.pdf

Recommended

More Related Content

Similar to Distributed Shared Memory-jhgfdsserty.pdf (20)

Recently uploaded (20)

Distributed Shared Memory-jhgfdsserty.pdf