CSC/ECE 506 Spring 2014/2a - Revision history

Sasthan2: /* References */

2014-02-09T23:33:58Z

References

← Older revision		Revision as of 23:33, 9 February 2014
Line 291:		Line 291:
	*Nitzberg, B.; Lo, V. , [http://www.cdf.toronto.edu/~csc469h/fall/handouts/nitzberg91.pdf "Distributed Shared Memory: A Survey of Issues and Algorithms"]		*Nitzberg, B.; Lo, V. , [http://www.cdf.toronto.edu/~csc469h/fall/handouts/nitzberg91.pdf "Distributed Shared Memory: A Survey of Issues and Algorithms"]
	*Steven Cameron Woo , Moriyoshi Ohara , Evan Torrie , Jaswinder Pal Singh , Anoop Gupta, [http://dl.acm.org/citation.cfm?id=223990 "The SPLASH-2 programs: characterization and methodological considerations,"] Proceedings of the 22nd annual international symposium on Computer architecture, p.24-36, June 22-24, 1995, S. Margherita Ligure, Italy		*Steven Cameron Woo , Moriyoshi Ohara , Evan Torrie , Jaswinder Pal Singh , Anoop Gupta, [http://dl.acm.org/citation.cfm?id=223990 "The SPLASH-2 programs: characterization and methodological considerations,"] Proceedings of the 22nd annual international symposium on Computer architecture, p.24-36, June 22-24, 1995, S. Margherita Ligure, Italy

	*Hennessy, J.; Heinrich, M.; Gupta, A.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=747863 "Cache-Coherent Distributed Shared Memory: Perspectives on Its Development and Future Chanllenges,"] Proceedings of the IEEE, Volume: 87 Issue:3, pp.418 - 429, Mar 1999, Comput. Syst. Lab., Stanford Univ., CA		*Hennessy, J.; Heinrich, M.; Gupta, A.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=747863 "Cache-Coherent Distributed Shared Memory: Perspectives on Its Development and Future Chanllenges,"] Proceedings of the IEEE, Volume: 87 Issue:3, pp.418 - 429, Mar 1999, Comput. Syst. Lab., Stanford Univ., CA
	*J. Protic, M. Tomasevic, and V. Milutinovic, [http://media.wiley.com/product_data/excerpt/76/08186773/0818677376-2.pdf An Overview of Distributed Shared Memory]		*J. Protic, M. Tomasevic, and V. Milutinovic, [http://media.wiley.com/product_data/excerpt/76/08186773/0818677376-2.pdf An Overview of Distributed Shared Memory]
Line 299:		Line 298:
	*Kranz, D.; Johnson, K.; Agarwal, A.; Kubiatowicz, J.; Lim, B.; , [http://delivery.acm.org.prox.lib.ncsu.edu/10.1145/160000/155338/p54-kranz.pdf?ip=152.1.24.251&acc=ACTIVE%20SERVICE&CFID=81831968&CFTOKEN=62928147&__acm__=1327853249_029db7e958cb50bd47056f939f3296f7 "Integrating Message-Passing and Shared-Memory: Early Experience"] PPOPP '93 Proceedings of the fourth ACM SIGPLAN symposium on Principles and practice of parallel programming, pp.54 - 63		*Kranz, D.; Johnson, K.; Agarwal, A.; Kubiatowicz, J.; Lim, B.; , [http://delivery.acm.org.prox.lib.ncsu.edu/10.1145/160000/155338/p54-kranz.pdf?ip=152.1.24.251&acc=ACTIVE%20SERVICE&CFID=81831968&CFTOKEN=62928147&__acm__=1327853249_029db7e958cb50bd47056f939f3296f7 "Integrating Message-Passing and Shared-Memory: Early Experience"] PPOPP '93 Proceedings of the fourth ACM SIGPLAN symposium on Principles and practice of parallel programming, pp.54 - 63
	*Amza, C.; Cox, A.L.; Dwarkadas, S.; Keleher, P.; Honghui Lu; Rajamony, R.; Weimin Yu; Zwaenepoel, W.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=485843&tag=1 "TreadMarks: shared memory computing on networks of workstations"] Computer, Volume: 29 Issue:2, pp. 18 - 28, Feb 1996, Dept. of Comput. Sci., Rice Univ., Houston, TX		*Amza, C.; Cox, A.L.; Dwarkadas, S.; Keleher, P.; Honghui Lu; Rajamony, R.; Weimin Yu; Zwaenepoel, W.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=485843&tag=1 "TreadMarks: shared memory computing on networks of workstations"] Computer, Volume: 29 Issue:2, pp. 18 - 28, Feb 1996, Dept. of Comput. Sci., Rice Univ., Houston, TX


	*Protic, Jelica, Milo Tomagevic, and Veljk Milutinovic. [http://www.cs.rit.edu/~pns6910/docs/Distributed%20Shared%20Memory%20Systems/A%20survey%20of%20distributed%20shared%20memory%20systems.pdf "A Survey of Distributed Shared Memory Systems."] 28th Annual Hawaii International Conference on System Sciences. IEEE. Hawaii, 1995. Reading.		*Protic, Jelica, Milo Tomagevic, and Veljk Milutinovic. [http://www.cs.rit.edu/~pns6910/docs/Distributed%20Shared%20Memory%20Systems/A%20survey%20of%20distributed%20shared%20memory%20systems.pdf "A Survey of Distributed Shared Memory Systems."] 28th Annual Hawaii International Conference on System Sciences. IEEE. Hawaii, 1995. Reading.

	== Others ==		== Others ==
	<references> </references>		<references> </references>

Sasthan2: /* References */

2014-02-09T23:32:45Z

References

← Older revision		Revision as of 23:32, 9 February 2014
Line 282:		Line 282:
	==References==		==References==
	*Bienia, Christian, Sanjeev Kumar, and Kai Li.[http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4636090 "PARSEC vs. SPLASH-2: A quantitative comparison of two multithreaded benchmark suites on chip-multiprocessors."] Workload Characterization, 2008. IISWC 2008. IEEE International Symposium on. IEEE, 2008.		*Bienia, Christian, Sanjeev Kumar, and Kai Li.[http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4636090 "PARSEC vs. SPLASH-2: A quantitative comparison of two multithreaded benchmark suites on chip-multiprocessors."] Workload Characterization, 2008. IISWC 2008. IEEE International Symposium on. IEEE, 2008.
			*Matthew Chapman and Gernot Heiser. 2009. [http://www.usenix.org/event/usenix09/tech/full_papers/chapman/chapman_html/index.html vNUMA: a virtual shared-memory multiprocessor.] In Proceedings of the 2009 conference on USENIX Annual technical conference (USENIX'09). USENIX Association, Berkeley, CA, USA, 2-2.
	*Shan, H.; Singh, J.P.; Oliker, L.; Biswas, R.; , [http://escholarship.org/uc/item/76p9b40g#page-1 "Message passing vs. shared address space on a cluster of SMPs,"] Parallel and Distributed Processing Symposium., Proceedings 15th International , vol., no., pp.8 pp., Apr 2001		*Shan, H.; Singh, J.P.; Oliker, L.; Biswas, R.; , [http://escholarship.org/uc/item/76p9b40g#page-1 "Message passing vs. shared address space on a cluster of SMPs,"] Parallel and Distributed Processing Symposium., Proceedings 15th International , vol., no., pp.8 pp., Apr 2001
	*[http://cnx.org/content/m20649/latest/ An Introduction to the Partitioned Global Address Space (PGAS) Programming Model],Module by: Tim Stitt Ph.D.		*[http://cnx.org/content/m20649/latest/ An Introduction to the Partitioned Global Address Space (PGAS) Programming Model],Module by: Tim Stitt Ph.D.
			*Mutanga, Alfred. [http://isro.eu.com/Nov_Proc/pdf%201.pdf "CACHE PERFORMANCE EVALUATION FOR MULTIPROCESSOR SYSTEM." Nov 2013 ]
			*Jegou, Y. , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=1199404 Implementation of page management in Mome, a user-level DSM] Cluster Computing and the Grid, 2003. Proceedings. CCGrid 2003. 3rd IEEE/ACM International Symposium on, p.479-486, 21 May 2003, IRISA/INRIA, France
	*Protic, J.; Tomasevic, M.; Milutinovic, V.; , [http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=494605&isnumber=10721 "Distributed shared memory: concepts and systems,"] Parallel & Distributed Technology: Systems & Applications, IEEE , vol.4, no.2, pp.63-71, Summer 1996		*Protic, J.; Tomasevic, M.; Milutinovic, V.; , [http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=494605&isnumber=10721 "Distributed shared memory: concepts and systems,"] Parallel & Distributed Technology: Systems & Applications, IEEE , vol.4, no.2, pp.63-71, Summer 1996
	*Chandola, V. , [http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.135.5206&rep=rep1&type=pdf "Design Issues in Implementation of Distributed Shared Memory in User Space,"]		*Chandola, V. , [http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.135.5206&rep=rep1&type=pdf "Design Issues in Implementation of Distributed Shared Memory in User Space,"]
	*Nitzberg, B.; Lo, V. , [http://www.cdf.toronto.edu/~csc469h/fall/handouts/nitzberg91.pdf "Distributed Shared Memory: A Survey of Issues and Algorithms"]		*Nitzberg, B.; Lo, V. , [http://www.cdf.toronto.edu/~csc469h/fall/handouts/nitzberg91.pdf "Distributed Shared Memory: A Survey of Issues and Algorithms"]
	*Steven Cameron Woo , Moriyoshi Ohara , Evan Torrie , Jaswinder Pal Singh , Anoop Gupta, [http://dl.acm.org/citation.cfm?id=223990 "The SPLASH-2 programs: characterization and methodological considerations,"] Proceedings of the 22nd annual international symposium on Computer architecture, p.24-36, June 22-24, 1995, S. Margherita Ligure, Italy		*Steven Cameron Woo , Moriyoshi Ohara , Evan Torrie , Jaswinder Pal Singh , Anoop Gupta, [http://dl.acm.org/citation.cfm?id=223990 "The SPLASH-2 programs: characterization and methodological considerations,"] Proceedings of the 22nd annual international symposium on Computer architecture, p.24-36, June 22-24, 1995, S. Margherita Ligure, Italy
	*Jegou, Y. , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=1199404 Implementation of page management in Mome, a user-level DSM] Cluster Computing and the Grid, 2003. Proceedings. CCGrid 2003. 3rd IEEE/ACM International Symposium on, p.479-486, 21 May 2003, IRISA/INRIA, France
	*Hennessy, J.; Heinrich, M.; Gupta, A.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=747863 "Cache-Coherent Distributed Shared Memory: Perspectives on Its Development and Future Chanllenges,"] Proceedings of the IEEE, Volume: 87 Issue:3, pp.418 - 429, Mar 1999, Comput. Syst. Lab., Stanford Univ., CA		*Hennessy, J.; Heinrich, M.; Gupta, A.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=747863 "Cache-Coherent Distributed Shared Memory: Perspectives on Its Development and Future Chanllenges,"] Proceedings of the IEEE, Volume: 87 Issue:3, pp.418 - 429, Mar 1999, Comput. Syst. Lab., Stanford Univ., CA
	*J. Protic, M. Tomasevic, and V. Milutinovic, [http://media.wiley.com/product_data/excerpt/76/08186773/0818677376-2.pdf An Overview of Distributed Shared Memory]		*J. Protic, M. Tomasevic, and V. Milutinovic, [http://media.wiley.com/product_data/excerpt/76/08186773/0818677376-2.pdf An Overview of Distributed Shared Memory]
Line 296:		Line 299:
	*Kranz, D.; Johnson, K.; Agarwal, A.; Kubiatowicz, J.; Lim, B.; , [http://delivery.acm.org.prox.lib.ncsu.edu/10.1145/160000/155338/p54-kranz.pdf?ip=152.1.24.251&acc=ACTIVE%20SERVICE&CFID=81831968&CFTOKEN=62928147&__acm__=1327853249_029db7e958cb50bd47056f939f3296f7 "Integrating Message-Passing and Shared-Memory: Early Experience"] PPOPP '93 Proceedings of the fourth ACM SIGPLAN symposium on Principles and practice of parallel programming, pp.54 - 63		*Kranz, D.; Johnson, K.; Agarwal, A.; Kubiatowicz, J.; Lim, B.; , [http://delivery.acm.org.prox.lib.ncsu.edu/10.1145/160000/155338/p54-kranz.pdf?ip=152.1.24.251&acc=ACTIVE%20SERVICE&CFID=81831968&CFTOKEN=62928147&__acm__=1327853249_029db7e958cb50bd47056f939f3296f7 "Integrating Message-Passing and Shared-Memory: Early Experience"] PPOPP '93 Proceedings of the fourth ACM SIGPLAN symposium on Principles and practice of parallel programming, pp.54 - 63
	*Amza, C.; Cox, A.L.; Dwarkadas, S.; Keleher, P.; Honghui Lu; Rajamony, R.; Weimin Yu; Zwaenepoel, W.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=485843&tag=1 "TreadMarks: shared memory computing on networks of workstations"] Computer, Volume: 29 Issue:2, pp. 18 - 28, Feb 1996, Dept. of Comput. Sci., Rice Univ., Houston, TX		*Amza, C.; Cox, A.L.; Dwarkadas, S.; Keleher, P.; Honghui Lu; Rajamony, R.; Weimin Yu; Zwaenepoel, W.; , [http://ieeexplore.ieee.org.prox.lib.ncsu.edu/stamp/stamp.jsp?tp=&arnumber=485843&tag=1 "TreadMarks: shared memory computing on networks of workstations"] Computer, Volume: 29 Issue:2, pp. 18 - 28, Feb 1996, Dept. of Comput. Sci., Rice Univ., Houston, TX
	*David R. Cheriton. 1985. [http://doi.acm.org.prox.lib.ncsu.edu/10.1145/858336.858338 Preliminary thoughts on problem-oriented shared memory: a decentralized approach to distributed systems.] SIGOPS Oper. Syst. Rev. 19, 4 (October 1985), 26-33.
	*Matthew Chapman and Gernot Heiser. 2009. [http://www.usenix.org/event/usenix09/tech/full_papers/chapman/chapman_html/index.html vNUMA: a virtual shared-memory multiprocessor.] In Proceedings of the 2009 conference on USENIX Annual technical conference (USENIX'09). USENIX Association, Berkeley, CA, USA, 2-2.

	*Protic, Jelica, Milo Tomagevic, and Veljk Milutinovic. [http://www.cs.rit.edu/~pns6910/docs/Distributed%20Shared%20Memory%20Systems/A%20survey%20of%20distributed%20shared%20memory%20systems.pdf "A Survey of Distributed Shared Memory Systems."] 28th Annual Hawaii International Conference on System Sciences. IEEE. Hawaii, 1995. Reading.		*Protic, Jelica, Milo Tomagevic, and Veljk Milutinovic. [http://www.cs.rit.edu/~pns6910/docs/Distributed%20Shared%20Memory%20Systems/A%20survey%20of%20distributed%20shared%20memory%20systems.pdf "A Survey of Distributed Shared Memory Systems."] 28th Annual Hawaii International Conference on System Sciences. IEEE. Hawaii, 1995. Reading.

Sasthan2 at 23:24, 9 February 2014

2014-02-09T23:24:11Z

← Older revision		Revision as of 23:24, 9 February 2014
Line 221:		Line 221:
	== Performance in SPLASH ==		== Performance in SPLASH ==

	There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4636090 SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence ~~we have described performance~~ in SPLASH.		There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4636090 SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence Performance in SPLASH has been described.

	''' SPLASH and SPLASH-2 '''		''' SPLASH and SPLASH-2 '''

Pdpathak at 23:20, 9 February 2014

2014-02-09T23:20:13Z

Pdpathak at 23:17, 9 February 2014

2014-02-09T23:17:48Z

← Older revision		Revision as of 23:17, 9 February 2014
Line 38:		Line 38:
	Cache-coherent DSM architectures rely on a directory-based [http://en.wikipedia.org/wiki/Cache_coherency cache coherence] protocol where an extra directory structure keeps track of all blocks that have been cached by each processor. A coherence protocol can then establish a consistent view of memory by maintaining state and other information about each cached block. These states usually minimally include Invalid, Shared, and Exclusive. Furthermore, in a cache-coherent DSM machine, the directory is distributed in memory to associate with the cache block it describes in the physical local memory.		Cache-coherent DSM architectures rely on a directory-based [http://en.wikipedia.org/wiki/Cache_coherency cache coherence] protocol where an extra directory structure keeps track of all blocks that have been cached by each processor. A coherence protocol can then establish a consistent view of memory by maintaining state and other information about each cached block. These states usually minimally include Invalid, Shared, and Exclusive. Furthermore, in a cache-coherent DSM machine, the directory is distributed in memory to associate with the cache block it describes in the physical local memory.

	== ~~DSM Implementations~~ ==		== Classification of DSMs ==

			From an architectural point of view, DSMs are composed of several nodes connected via a network. Each of the nodes can be an individual machine or a cluster of machines. Each system has local memory modules that are either partially or completely part of the shared memory. There are many characteristics that can be used to classify DSMs. The most common two ways are by the algorithm used by the DSM and the type of implementation.

	From an architectural point of view, DSMs are composed of several nodes connected via a network. Each of the nodes can be an individual machine or a cluster of machines. Each system has local memory modules that are either partially or completely part of the shared memory. There are many characteristics that can be used to classify DSM implementations. One of them is based on the nature of the memory demarcation: Software, Hardware, and Hybrid.
	There are three types of DSM algorithm:		There are three types of DSM algorithm:
	* Single Reader/ Single Writer (SRSW)		* Single Reader/ Single Writer (SRSW)
Line 90:		Line 91:
	\|}		\|}

			Another way of classifying DSMs is the level at which the DSM is implemented. ''To achieve ease of programming, cost-effectiveness and scalability, DSM systems logically implement the shared memory model on physical distributed memories''. <ref> Distributed shared memory: concepts and systems </ref> The normal operations thus include looking-up the data in the local memory, getting the data if it is not present and action on write accesses to maintain the coherence of data. Lookup and action for coherence can be done in Software, hardware or combination of both. Thus DSMs can be classified into software, hardware and hybrid implementations.<ref> Distributed shared memory: concepts and systems </ref>

	=== Software DSMs ===		=== Software DSMs implementations ===

	Software support for DSMs started being explored in the mid-eighties as they are more flexible than their hardware counter-parts. They are implemented by using user-level software, OS, programming language, or combination of these.		Software support for DSMs started being explored in the mid-eighties as they are more flexible than their hardware counter-parts. They are implemented by using user-level software, OS, programming language, or combination of these.
Line 112:		Line 114:
	''' Linda ''' was introduced as an architecture-independent language where the shared memory is put as a ‘tuple space’, tuples mean storage and access units that have data in them. Linda also offers the use of replication as a method for partitioning.		''' Linda ''' was introduced as an architecture-independent language where the shared memory is put as a ‘tuple space’, tuples mean storage and access units that have data in them. Linda also offers the use of replication as a method for partitioning.

	=== Hardware DSMs ===		=== Hardware DSMs implementations ===
			Hardware DSMs though superior in performance, require additional hardware, are complex and expensive. They are mostly used in high-performance, large scale machines. Some examples of hardware DSMs are discussed below.

	''' Memnet '''		''' Memnet '''
Line 119:		Line 122:
	''' KSR1 ''' multiprocessor represents one of the early attempts to make DSM systems available on the market. It had a ring based hierarchical organization of clusters each with a local 32 Mbyte cache. Here since there was no main meory the problem of replacement of cache lines arose.		''' KSR1 ''' multiprocessor represents one of the early attempts to make DSM systems available on the market. It had a ring based hierarchical organization of clusters each with a local 32 Mbyte cache. Here since there was no main meory the problem of replacement of cache lines arose.

	=== Hybrid DSMs ===		=== Hybrid DSMs implementations ===
			Hybrid DSM implementations find a balance between the flexibility of software and high performance of the hardware. They are extensively used in mid-range systems. Following are some examples of hybrid DSMs.

	''' Alewife ''' is a specific hybrid system that implements the LimitLESS directory protocol. This protocol represents a hardware-based coherence scheme supported by a software mechanism The main advantage of this is that directory protocol is storage efficient while performing about as well as the full map directory protocol.		''' Alewife ''' is a specific hybrid system that implements the LimitLESS directory protocol. This protocol represents a hardware-based coherence scheme supported by a software mechanism The main advantage of this is that directory protocol is storage efficient while performing about as well as the full map directory protocol.

Sasthan2 at 22:59, 9 February 2014

2014-02-09T22:59:14Z

← Older revision		Revision as of 22:59, 9 February 2014
Line 12:		Line 12:

	Distributed memory systems are multi-processor systems in which each processor has its own individual memory. Tasks can only operate on a processor's local memory and if non-local data is required, the processor must communicate with one or more remote processors. Distributed memory systems started to flourish in the 1980s. The increasing performance in processors and network connectivity offered the perfect environment for parallel processing over a network of computers. This was a cheap way to put together massive computing power. The main drawback was going from sequential programs made for local memory to parallel programming in shared memory. This was where SAS provided the means to simplify programming by hiding the mechanisms to access distant memory located in other computers of the cluster.		Distributed memory systems are multi-processor systems in which each processor has its own individual memory. Tasks can only operate on a processor's local memory and if non-local data is required, the processor must communicate with one or more remote processors. Distributed memory systems started to flourish in the 1980s. The increasing performance in processors and network connectivity offered the perfect environment for parallel processing over a network of computers. This was a cheap way to put together massive computing power. The main drawback was going from sequential programs made for local memory to parallel programming in shared memory. This was where SAS provided the means to simplify programming by hiding the mechanisms to access distant memory located in other computers of the cluster.
	In 1985, Cheriton, in his article [http://doi.acm.org.prox.lib.ncsu.edu/10.1145/858336.858338 "Preliminary thoughts on problem-oriented shared memory: a decentralized approach to distributed systems"], introduced ideas for the application of shared memory techniques in distributed memory systems. Cheriton envisioned a system of nodes with a pool of shared memory using a common file namespace that could "decentralize the implementation of a service."
	Early distributed computer systems relied almost exclusively on message passing in order to communicate with one another, and this technique is still widely used today. In a message passing (MP) model, each processor's local memory can be considered as isolated from that of the rest of the system. Processes or objects can send or receive messages in order to communicate, and this can occur in a synchronous or asynchronous manner. In distributed systems, and particularly with certain types of programs, the message passing model can become overly burdensome to the programmer as tracking data movement and maintaining data integrity can become challenging with many control threads. A shared address or shared-memory system, however, can provide a programming model that simplifies data sharing via uniform mechanisms of data structure reads and writes on common memory. Current distributed systems seek to take advantage both SAS and MP programming model principles in hybrid systems.		Early distributed computer systems relied almost exclusively on message passing in order to communicate with one another, and this technique is still widely used today. In a message passing (MP) model, each processor's local memory can be considered as isolated from that of the rest of the system. Processes or objects can send or receive messages in order to communicate, and this can occur in a synchronous or asynchronous manner. In distributed systems, and particularly with certain types of programs, the message passing model can become overly burdensome to the programmer as tracking data movement and maintaining data integrity can become challenging with many control threads. A shared address or shared-memory system, however, can provide a programming model that simplifies data sharing via uniform mechanisms of data structure reads and writes on common memory. Current distributed systems seek to take advantage both SAS and MP programming model principles in hybrid systems.

Line 233:		Line 232:
	DSM helps provide an easy programming interface i.e. shared memory as well scalability of distributed memory machines. But the communication cost inherited in the underlying network is very expensive, thus limits the scalability of distributed shared memory systems and also create other problems. Page based DSM’s use the virtual memory mechanism in systems to access data in shared memories. Now when a single system is writing data and uses an invalid protocol to get consistency in data, it can cause ‘false sharing’. Two processes in a system can write different data that reside on the same virtual memory page, which can cause it to ping-pong between the two nodes. This shall just cause congesting and excess traffic. Due to this, the granularity of memory unit is restricted in a certain range to prevent false sharing or excessive message passing. Hence there is a need to model the communication with relaxed memory consistency models. This way the shared memory pages shall be consistent in reading data from well-defined points. This traffic can be reduced by sending only the changed data from a modified page.		DSM helps provide an easy programming interface i.e. shared memory as well scalability of distributed memory machines. But the communication cost inherited in the underlying network is very expensive, thus limits the scalability of distributed shared memory systems and also create other problems. Page based DSM’s use the virtual memory mechanism in systems to access data in shared memories. Now when a single system is writing data and uses an invalid protocol to get consistency in data, it can cause ‘false sharing’. Two processes in a system can write different data that reside on the same virtual memory page, which can cause it to ping-pong between the two nodes. This shall just cause congesting and excess traffic. Due to this, the granularity of memory unit is restricted in a certain range to prevent false sharing or excessive message passing. Hence there is a need to model the communication with relaxed memory consistency models. This way the shared memory pages shall be consistent in reading data from well-defined points. This traffic can be reduced by sending only the changed data from a modified page.

	'''Case Study ~~- 2008~~ - Roy and Chaudhary'''		'''Case Study - Roy and Chaudhary (2008)'''

	In 2008, [http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.57.1319 Roy and Chaudhary] compared the communication requirements of three different page-based DSM systems (CVM, Quarks, and Strings) that use virtual memory mechanisms to trap accesses to shared areas. Their study was also based on the SPLASH-2 suite of programs. In their experiments, Quarks was running tasks on separate processes (it does not support more than one application thread per process) while CVM and Strings were run with multiple application threads.		In 2008, [http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.57.1319 Roy and Chaudhary] compared the communication requirements of three different page-based DSM systems (CVM, Quarks, and Strings) that use virtual memory mechanisms to trap accesses to shared areas. Their study was also based on the SPLASH-2 suite of programs. In their experiments, Quarks was running tasks on separate processes (it does not support more than one application thread per process) while CVM and Strings were run with multiple application threads.

Sasthan2 at 22:55, 9 February 2014

2014-02-09T22:55:45Z

← Older revision		Revision as of 22:55, 9 February 2014
Line 276:		Line 276:

	==References==		==References==
			*Bienia, Christian, Sanjeev Kumar, and Kai Li.[http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4636090 "PARSEC vs. SPLASH-2: A quantitative comparison of two multithreaded benchmark suites on chip-multiprocessors."] Workload Characterization, 2008. IISWC 2008. IEEE International Symposium on. IEEE, 2008.
	*Shan, H.; Singh, J.P.; Oliker, L.; Biswas, R.; , [http://escholarship.org/uc/item/76p9b40g#page-1 "Message passing vs. shared address space on a cluster of SMPs,"] Parallel and Distributed Processing Symposium., Proceedings 15th International , vol., no., pp.8 pp., Apr 2001		*Shan, H.; Singh, J.P.; Oliker, L.; Biswas, R.; , [http://escholarship.org/uc/item/76p9b40g#page-1 "Message passing vs. shared address space on a cluster of SMPs,"] Parallel and Distributed Processing Symposium., Proceedings 15th International , vol., no., pp.8 pp., Apr 2001
	*[http://cnx.org/content/m20649/latest/ An Introduction to the Partitioned Global Address Space (PGAS) Programming Model],Module by: Tim Stitt Ph.D.		*[http://cnx.org/content/m20649/latest/ An Introduction to the Partitioned Global Address Space (PGAS) Programming Model],Module by: Tim Stitt Ph.D.

Sasthan2 at 22:53, 9 February 2014

2014-02-09T22:53:38Z

← Older revision		Revision as of 22:53, 9 February 2014
Line 216:		Line 216:
	== Performance in SPLASH ==		== Performance in SPLASH ==

	There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://dl.~~acm~~.org/~~citation~~.~~cfm~~?id=~~223990~~ SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence we have described performance in SPLASH.		There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=4636090 SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence we have described performance in SPLASH.

	''' SPLASH and SPLASH-2 '''		''' SPLASH and SPLASH-2 '''

Sasthan2 at 22:50, 9 February 2014

2014-02-09T22:50:34Z

← Older revision		Revision as of 22:50, 9 February 2014
Line 214:		Line 214:
	The authors found that the average speedup of the HLRC protocol is 5.97, and the average speedup of the SC (with MV) protocol is 4.5 if non-optimized allocation is used, and 6.2 if the granularity is changed dynamically.		The authors found that the average speedup of the HLRC protocol is 5.97, and the average speedup of the SC (with MV) protocol is 4.5 if non-optimized allocation is used, and 6.2 if the granularity is changed dynamically.

	== Performance ==		== Performance in SPLASH ==

	There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://dl.acm.org/citation.cfm?id=223990 SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence we have described performance in SPLASH.		There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://dl.acm.org/citation.cfm?id=223990 SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence we have described performance in SPLASH.

Sasthan2 at 22:42, 9 February 2014

2014-02-09T22:42:47Z

← Older revision		Revision as of 22:42, 9 February 2014
Line 216:		Line 216:
	== Performance ==		== Performance ==

	There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://dl.acm.org/citation.cfm?id=223990 SPLASH and SPLASH-2.].		There are numerous studies of the performance of shared memory applications in distributed systems. The vast majority of them use a collection of programs named [http://dl.acm.org/citation.cfm?id=223990 SPLASH and SPLASH-2.]. SPLASH is the most commonly used suite for scientific studies of parallel machines with shared memory. Hence we have described performance in SPLASH.

	''' SPLASH and SPLASH-2 '''		''' SPLASH and SPLASH-2 '''

← Older revision		Revision as of 23:20, 9 February 2014
Line 42:		Line 42:
	From an architectural point of view, DSMs are composed of several nodes connected via a network. Each of the nodes can be an individual machine or a cluster of machines. Each system has local memory modules that are either partially or completely part of the shared memory. There are many characteristics that can be used to classify DSMs. The most common two ways are by the algorithm used by the DSM and the type of implementation.		From an architectural point of view, DSMs are composed of several nodes connected via a network. Each of the nodes can be an individual machine or a cluster of machines. Each system has local memory modules that are either partially or completely part of the shared memory. There are many characteristics that can be used to classify DSMs. The most common two ways are by the algorithm used by the DSM and the type of implementation.

			=== DSM algorithms ===
	There are three types of DSM algorithm:		There are three types of DSM algorithm:
	* Single Reader/ Single Writer (SRSW)		* Single Reader/ Single Writer (SRSW)
Line 90:		Line 91:
	\|-		\|-
	\|}		\|}
			=== DSM implementations ===

	Another way of classifying DSMs is the level at which the DSM is implemented. ''To achieve ease of programming, cost-effectiveness and scalability, DSM systems logically implement the shared memory model on physical distributed memories''. <ref> Distributed shared memory: concepts and systems </ref> The normal operations thus include looking-up the data in the local memory, getting the data if it is not present and action on write accesses to maintain the coherence of data. Lookup and action for coherence can be done in Software, hardware or combination of both. Thus DSMs can be classified into software, hardware and hybrid implementations.<ref> Distributed shared memory: concepts and systems </ref>		Another way of classifying DSMs is the level at which the DSM is implemented. ''To achieve ease of programming, cost-effectiveness and scalability, DSM systems logically implement the shared memory model on physical distributed memories''. <ref> Distributed shared memory: concepts and systems </ref> The normal operations thus include looking-up the data in the local memory, getting the data if it is not present and action on write accesses to maintain the coherence of data. Lookup and action for coherence can be done in Software, hardware or combination of both. Thus DSMs can be classified into software, hardware and hybrid implementations.<ref> Distributed shared memory: concepts and systems </ref>

	=== Software DSMs implementations ===		==== Software DSMs implementations ====

	Software support for DSMs started being explored in the mid-eighties as they are more flexible than their hardware counter-parts. They are implemented by using user-level software, OS, programming language, or combination of these.		Software support for DSMs started being explored in the mid-eighties as they are more flexible than their hardware counter-parts. They are implemented by using user-level software, OS, programming language, or combination of these.
Line 114:		Line 116:
	''' Linda ''' was introduced as an architecture-independent language where the shared memory is put as a ‘tuple space’, tuples mean storage and access units that have data in them. Linda also offers the use of replication as a method for partitioning.		''' Linda ''' was introduced as an architecture-independent language where the shared memory is put as a ‘tuple space’, tuples mean storage and access units that have data in them. Linda also offers the use of replication as a method for partitioning.

	=== Hardware DSMs implementations ===		==== Hardware DSMs implementations ====
	Hardware DSMs though superior in performance, require additional hardware, are complex and expensive. They are mostly used in high-performance, large scale machines. Some examples of hardware DSMs are discussed below.		Hardware DSMs though superior in performance, require additional hardware, are complex and expensive. They are mostly used in high-performance, large scale machines. Some examples of hardware DSMs are discussed below.

Line 122:		Line 124:
	''' KSR1 ''' multiprocessor represents one of the early attempts to make DSM systems available on the market. It had a ring based hierarchical organization of clusters each with a local 32 Mbyte cache. Here since there was no main meory the problem of replacement of cache lines arose.		''' KSR1 ''' multiprocessor represents one of the early attempts to make DSM systems available on the market. It had a ring based hierarchical organization of clusters each with a local 32 Mbyte cache. Here since there was no main meory the problem of replacement of cache lines arose.

	=== Hybrid DSMs implementations ===		==== Hybrid DSMs implementations ====
	Hybrid DSM implementations find a balance between the flexibility of software and high performance of the hardware. They are extensively used in mid-range systems. Following are some examples of hybrid DSMs.		Hybrid DSM implementations find a balance between the flexibility of software and high performance of the hardware. They are extensively used in mid-range systems. Following are some examples of hybrid DSMs.