CSC/ECE 506 Fall 2007/wiki1 7 a1: Difference between revisions

From Expertiza_Wiki
Jump to navigation Jump to search
 
(52 intermediate revisions by the same user not shown)
Line 1: Line 1:
Shared address space is an important class of the multiprocessors where all processors share a global memory. Communication between tasks running on different processors is performed through writing to and reading from the global memory. All interprocessor coordination and synchronization is also accomplished via the global memory. Hence, a shared memory computer system consists of a set of independent processors, a set of memory modules, and an interconnection network.
This article talks about the recent trends in shared address space and how the interconnection technology has changed in last 10 years. Later it highlights some of the existing SMPs and latest supercomputer.
==Recent Design Trends in Shared Address Space==
==Recent Design Trends in Shared Address Space==


One change in the organization of address spaces in the last 10 years has been the use of vector instructions to deliver fine grain data parallelism. Examples of this are SIMD (single intruction multiple data) and ...
Current high-performance multiprocessor platforms may be broken into two categories: distributed address-space and shared address-space. For programs running on a distributed address-space multiprocessor, data visible on one processing unit is not visible on the remaining processor units. In contrast, hardware mechanisms in high performance shared address-space environments ensure that all data is in principle visible to all processors. While a shared address-space can offer programmability advantages and better performance for fine-grained applications, the very mechanisms that create those advantages appear to prevent the machines from scaling to large numbers of processors.  


Another recent trend in address space partitioning has been towards making the address space more transparent to the programmer. This is achieved by allowing more of the memory architecture to be partitioned by the compiler instead of hardware. This allows more private data for the program to be stored locally in cache instead of remotely, reducing access times.  
To scale beyond the limits of traditional shared address space architectures, hybrids of shared and distributed architectures are becoming more prevalent. These architectures allow users to take advantage of two or more address spaces to gain the scalability of distributed architectures while retaining the benefits of the shared address-space architecture.
Memory is split into global, which hardware keeps coherent, and strictly local, which is left incoherent. Application programmers are given a mechanism for specifying, at a very high level, data that would benefit from localization – being moved from global to local memory. This transparency in memory partitioning takes place through improved compiler allocation mechanisms than the traditional hardware devices.  The following are benefits of localizing otherwise remote memory accesses.


1. Faster access to localized data.


Over the few years two major form of interconnection network has evolved:
2. Elimination of redundant protocol traffic.


1. UMA (Uniform Memory Access): UMA architecture uses bus-based symmetric multiprocessor, such as SGI challenge.  
3. Elimination of locking/contention for written shared data.


2. NUMA (Non-Uniform Memory Access): elements not symmetrically partitioned. The local memory controller determines whether to access its cache or exchange messages with a remote memory controller.
A trend in application design is the use of vector instructions. This allows programs to deliver more fine grain tasks which capitalize on the parallel processing nature of a shared address space architecture. Vector instructions, also known as SIMD (single intruction multiple data), have been optimized on a series of recent architectures including the cell processor in Toshiba/Sony's PlayStation 3 game console. The PS3 cell processor consists of one scalar and eight vector processors.


==Evolution of Interconnect Technology==
==Evolution of Interconnect Technology==
Line 18: Line 24:
The single shared bus system has been extended to multiple buses to connect multiple processors.  
The single shared bus system has been extended to multiple buses to connect multiple processors.  


A multiple bus multiprocessor system uses several parallel buses to interconnect multiple processors and multiple memory modules. A number of connection schemes are possible in this case. Among the possibilities are the multiple bus with full bus–memory connection (MBFBMC), multiple bus with single bus memory connection (MBSBMC), multiple bus with partial bus–memory connection (MBPBMC), and multiple bus with class-based memory connection (MBCBMC). The multiple bus with full bus–memory connection has all memory modules connected to all buses. The multiple bus with single bus–memory connection has each memory module connected to a specific bus. The multiple bus with partial bus–memory connection has each memory module connected to a subset of buses. The multiple bus with class-based memory connection has memory modules grouped into classes whereby each class is connected to a specific subset of buses. A class is just an arbitrary collection of memory modules.
A multiple bus multiprocessor system uses several parallel buses to interconnect multiple processors and multiple memory modules. A number of connection schemes are possible in this case. Among the possibilities are the multiple bus with full bus–memory connection (MBFBMC), multiple bus with single bus memory connection (MBSBMC), multiple bus with partial bus–memory connection (MBPBMC), and multiple bus with class-based memory connection (MBCBMC). The multiple bus with full bus–memory connection has all memory modules connected to all buses. The multiple bus with single bus–memory connection has each memory module connected to a specific bus. The multiple bus with partial bus–memory connection has each memory module connected to a subset of buses. The multiple bus with class-based memory connection has memory modules grouped into classes whereby each class is connected to a specific subset of buses. A class is just an arbitrary collection of memory modules. Example of different multiple bus scheme for for the case of N = 6 processors, M = 4 memory modules, and B = 4 buses has been illustrated in the figure below.
 
[[Image:Bus01.JPG]]
[[Image:Bus02.JPG]]


A typical bus-based design uses caches to solve the bus contention problem. Highspeed caches connected to each processor on one side and the bus on the other side mean that local copies of instructions and data can be supplied at the highest possible rate. If the local processor finds all of its instructions and data in the local cache, we say the hit rate is 100%. The miss rate of a cache is the fraction of the references that cannot be satisfied by the cache, and so must be copied from the global memory, across the bus, into the cache, and then passed on to the local processor. One of the goals of the cache is to maintain a high hit rate, or low miss rate under high processor loads. A high hit rate means the processors are not using the bus as much. Hit rates are determined by a number of factors, ranging from the application programs being run to the manner in which cache hardware is implemented.  
Figure 2.1 (a) Multiple bus with full bus–memory connection (MBFBMC); (b) multiple bus with single bus-memory connection (MBSBMC); (c) multiple bus with partial bus–memory connection (MBPBMC); and (d) multiple bus with class-based memory
connection (MBCBMC). [Source: Reference 8]


The use of multiple buses to connect multiple processors is a natural extension to the single shared bus system. A multiple bus multiprocessor system uses several parallel buses to interconnect multiple processors and multiple memory modules. A number
A typical bus-based design uses caches to solve the bus contention problem. Highspeed caches connected to each processor on one side and the bus on the other side mean that local copies of instructions and data can be supplied at the highest possible rate. If the local processor finds all of its instructions and data in the local cache, we say the hit rate is 100%. The miss rate of a cache is the fraction of the references that are not stored locally. The required data must be copied from the global memory, across the bus and into the cache. From there it is routed to the local processor. One of the goals of the cache is to maintain a low miss rate under high processor loads. A low miss rate means the processors do not have to use the bus as much. Miss rates are determined by a number of factors, ranging from the complexity and number of applications being run to the organization of the cache hardware that is implemented.
of connection schemes are possible in this case. Among the possibilities are the multiple bus with full bus–memory connection (MBFBMC), multiple bus with single bus memory connection (MBSBMC), multiple bus with partial bus–
memory connection (MBPBMC), and multiple bus with class-based memory connection (MBCBMC). The multiple bus with full bus–memory connection has all memory modules connected to all buses. The multiple bus with single bus–memory connection has each memory module connected to a specific bus. The multiple bus with partial bus–memory connection has each memory module connected to a subset of buses. The multiple bus with class-based memory connection has memory modules grouped into classes whereby each class is connected to a specific subset of buses. A class is just an arbitrary collection of memory modules.


== Current High End SMPs ==
== Current High End SMPs ==
Line 34: Line 42:
2. The Cray XT4 system can scale from 562 to 30,614 AMD Dual Core 2.6-GHz giving a peak performance varying from 5.6TFLOPS to 318 TFLOPS.  
2. The Cray XT4 system can scale from 562 to 30,614 AMD Dual Core 2.6-GHz giving a peak performance varying from 5.6TFLOPS to 318 TFLOPS.  


3.The IBM System p5 595 server uses fifth-generation 64-bit IBM POWER5 technology in up to 64 core symmetric multi-processing (SMP) configurations with IBM Advanced POWER5+(2.1/2.3GHz).  
3. The IBM System p5 595 server uses fifth-generation 64-bit IBM POWER5 technology in up to 64 core symmetric multi-processing (SMP) configurations with IBM Advanced POWER5+ (2.1/2.3GHz).


== Evolution of Supercomputers ==
== Evolution of Supercomputers ==


Since the Cray T3E, supercomputers have steadily improved their shared address space architectures. Many of the improvements in address space organization have been utilized by supercomputers. Since the T3E was released, Cray merged with SGI and acquired OctigaBay Systems Corporation, a Canadian company developing a high performance computing. As a leader in the development of supercomputers, Cray, Inc. has released several more machines superior to the T3E. The most noticeable feature in all newer supercomputers is the implementation of distributed shared memory (DSM). The DSM architecture removes memory contention and interconnect bottlenecks, giving all processors access to any memory location.  
Since the Cray T3E, supercomputers have steadily improved their shared address space architectures. Many of the improvements in address space organization have been utilized by supercomputers. Since the T3E was released, Cray merged with SGI and acquired OctigaBay Systems Corporation. As a leader in the development of supercomputers, Cray, Inc. has released several more machines superior to the T3E. The most noticeable feature in newer supercomputers is the implementation of distributed shared memory (DSM). Below are 10 of the fastest supercomputers in recent time.
 
[[Image:Table01.JPG]]
 
Figure 4.1 - Top 10 Fastest Supercomputers
 
Ranked as the fastest supercomputer in the world and released after Cray's T3E, IBM's Blue Genie supercomputer is currently able to reach speeds of 367 TeraFlops. A node in the Blue Gene architecture consists of a single ASIC, shown in Figure 4.2, along with a read/write locking mechanism, scratch DRAM, and L1 and L2 data caches. The ASIC contains two 700 MHz PowerPC 440 embedded processors. Each of the processors has two floating point units, DRAM controller and the logic to support multiple message sharing system. Theoretically a peak performance of 5.6 GFLOPS can be acheived by the dual FPUs. To reduce overhead, peer node CPUs are not cache coherent.
 
 
[[Image:Pic01.jpg]]
 
 
Figure 4.2 - Blue Gene node ASIC
 


Unlike most other systems, the Cray X1E system has a DSM architecture that integrates the memory and interconnect subsystems to allow memory references to be efficiently routed directly to the appropriate local or remote memory. Memory is physically distributed on individual modules, but all memory is directly addressable to and accessible by any MSP in the system through the use of load and store instructions. While each node functions like a traditional SMP node, its processors can also directly address memory on any other node—remote memory accesses go over the interconnect to request processors, bypassing the local cache. This mechanism is more scalable than traditional shared memory and it provides very low latencies and unprecedented interprocessor bandwidths. Each processor can have up to 2,048 outstanding memory references, allowing applications to tolerate global network latencies.  
Due to the nature of implementing the cores on ASIC chips, low power operation was maintained, allowing high area density and scalability of nodes. Each node operates in a range around 17 Watts. These properties allow up to 1024 nodes plus additional IO nodes in a standard 19" cabinet, within reasonable limits of electrical power supply and air cooling.  


The interconnect supporting these remote references includes routing logic and network ports on each compute module, and separate routing modules. A novel design effectively implements 16 independent 2D torus topologies within the interconnect, each called a slice. Processor or I/O memory references are first handled by the routing logic of the appropriate slice. If the address is local to the node, the routing logic accesses the node’s local memory. If the memory address is on a remote node, the request is routed using the network, and the routing logic on the remote node handles the request as a local reference. Each slice of the machine independently handles all memory accesses and routing for addresses that map to that slice. Each compute module accesses the network through a total of 32 network ports, two per slice, each of which supports 1.6 GB/second peak per direction. For large systems, half of these ports are connected to router modules which are connected to other compute or router modules to build up the interconnect. The scalability of this interconnect is further enhanced by the following features. Local memory references are cached, remote memory references are not, reducing the overhead normally associated with SMP coherence protocols. The Cray X1E system translates addresses at the destination node, requiring each node only keep track of translation information for its local memory.
A hybrid of bus and message passsing architectures are used in the Blue Gene as is necessary for high speed symmetric memory locations that allows scalability. Each node is part of a 3D toroidal bus network architecture, a bus network for collective communication, and a message passing global interrupt network.


== References ==
== References ==
Line 53: Line 74:


[http://www.cray.com/downloads/X1E_datasheet.pdf] Cray X1E
[http://www.cray.com/downloads/X1E_datasheet.pdf] Cray X1E
[http://en.wikipedia.org/wiki/Vector_processor] Wikipedia: Vector Processor
[http://en.wikipedia.org/wiki/Image:Blue_Gene_L_ASIC.png] Wikipedia: Blue Gene
7. McCurdy , C. & Fischer, C. (2003). User-controllable coherence for high performance shared memory multiprocessors. Proceedings of the Ninth ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming Archive, , 73-82.
8. Culler, D. E., Singh, J. P., & Gupta, A. (1999). Introduction: Convergence of Parallel Architectures: Shared Address Space. In D. E. Penrose (Ed.), A Tutor's Guide: Helping Writers One to One (pp. 28-37). San Francisco, CA: Morgan Kaufmann.
9. El-Rewini, Hesham. Abd-El-Barr, Mostafa. Advanced Computer Architecture and Parallel Processing (chapter 2 & 4), John Wiley & Sons, Inc. (US), 2005

Latest revision as of 03:27, 11 September 2007

Shared address space is an important class of the multiprocessors where all processors share a global memory. Communication between tasks running on different processors is performed through writing to and reading from the global memory. All interprocessor coordination and synchronization is also accomplished via the global memory. Hence, a shared memory computer system consists of a set of independent processors, a set of memory modules, and an interconnection network.

This article talks about the recent trends in shared address space and how the interconnection technology has changed in last 10 years. Later it highlights some of the existing SMPs and latest supercomputer.

Recent Design Trends in Shared Address Space

Current high-performance multiprocessor platforms may be broken into two categories: distributed address-space and shared address-space. For programs running on a distributed address-space multiprocessor, data visible on one processing unit is not visible on the remaining processor units. In contrast, hardware mechanisms in high performance shared address-space environments ensure that all data is in principle visible to all processors. While a shared address-space can offer programmability advantages and better performance for fine-grained applications, the very mechanisms that create those advantages appear to prevent the machines from scaling to large numbers of processors.

To scale beyond the limits of traditional shared address space architectures, hybrids of shared and distributed architectures are becoming more prevalent. These architectures allow users to take advantage of two or more address spaces to gain the scalability of distributed architectures while retaining the benefits of the shared address-space architecture. Memory is split into global, which hardware keeps coherent, and strictly local, which is left incoherent. Application programmers are given a mechanism for specifying, at a very high level, data that would benefit from localization – being moved from global to local memory. This transparency in memory partitioning takes place through improved compiler allocation mechanisms than the traditional hardware devices. The following are benefits of localizing otherwise remote memory accesses.

1. Faster access to localized data.

2. Elimination of redundant protocol traffic.

3. Elimination of locking/contention for written shared data.

A trend in application design is the use of vector instructions. This allows programs to deliver more fine grain tasks which capitalize on the parallel processing nature of a shared address space architecture. Vector instructions, also known as SIMD (single intruction multiple data), have been optimized on a series of recent architectures including the cell processor in Toshiba/Sony's PlayStation 3 game console. The PS3 cell processor consists of one scalar and eight vector processors.

Evolution of Interconnect Technology

The effectiveness of the shared memory approach depends on the latency incurred on the memory access as well as the bandwidth of the data transfer that can be supported. The interconnection network in a shared-memory multiprocessor has evolved over a decade to provide better aggregate bandwidth when more processors are added without increasing cost too much. Shared memory systems can be designed using bus-based or switch-based interconnection networks. The simplest network for shared memory systems is the bus and moreover the bus/cache architecture alleviates the need for expensive multiported memories and interface circuitry as well as the need to adopt a message-passing paradigm when developing application software. The single shared bus system has been extended to multiple buses to connect multiple processors.

A multiple bus multiprocessor system uses several parallel buses to interconnect multiple processors and multiple memory modules. A number of connection schemes are possible in this case. Among the possibilities are the multiple bus with full bus–memory connection (MBFBMC), multiple bus with single bus memory connection (MBSBMC), multiple bus with partial bus–memory connection (MBPBMC), and multiple bus with class-based memory connection (MBCBMC). The multiple bus with full bus–memory connection has all memory modules connected to all buses. The multiple bus with single bus–memory connection has each memory module connected to a specific bus. The multiple bus with partial bus–memory connection has each memory module connected to a subset of buses. The multiple bus with class-based memory connection has memory modules grouped into classes whereby each class is connected to a specific subset of buses. A class is just an arbitrary collection of memory modules. Example of different multiple bus scheme for for the case of N = 6 processors, M = 4 memory modules, and B = 4 buses has been illustrated in the figure below.

Figure 2.1 (a) Multiple bus with full bus–memory connection (MBFBMC); (b) multiple bus with single bus-memory connection (MBSBMC); (c) multiple bus with partial bus–memory connection (MBPBMC); and (d) multiple bus with class-based memory connection (MBCBMC). [Source: Reference 8]

A typical bus-based design uses caches to solve the bus contention problem. Highspeed caches connected to each processor on one side and the bus on the other side mean that local copies of instructions and data can be supplied at the highest possible rate. If the local processor finds all of its instructions and data in the local cache, we say the hit rate is 100%. The miss rate of a cache is the fraction of the references that are not stored locally. The required data must be copied from the global memory, across the bus and into the cache. From there it is routed to the local processor. One of the goals of the cache is to maintain a low miss rate under high processor loads. A low miss rate means the processors do not have to use the bus as much. Miss rates are determined by a number of factors, ranging from the complexity and number of applications being run to the organization of the cache hardware that is implemented.

Current High End SMPs

Symmetric multiprocessors (SMPs) are available from a wide range of workstation vendors in various configurations. With the introduction of dual-core devices, SMP is found in most new desktop machines and in many laptop machines. The most popular entry-level SMP systems use the x86 instruction set architecture and are based on Intel’s Xeon, Pentium D, Core Duo, and Core 2 Duo based processors or AMD’s Athlon64 X2, Quad FX or Opteron 200 and 2000 series processors. Servers use those processors and other readily available non-x86 processor choices including the Sun Microsystems UltraSPARC, Fujitsu SPARC64, SGI MIPS, Intel Itanium, Hewlett Packard PA-RISC, Hewlett-Packard (formerly Compaq formerly Digital Equipment Corporation) DEC Alpha, IBM POWER and Apple Computer PowerPC (specifically G4 and G5 series, as well as earlier PowerPC 604 and 604e series) processors.

1. The Sun Fire 12K server is a high-end data center server with up to 52 UltraSPARC III Cu 1.2-GHz processors in a symmetric multiprocessing architecture. The Sun Fire E52K server scales up to 72 UltraSPARC IV+ processors.

2. The Cray XT4 system can scale from 562 to 30,614 AMD Dual Core 2.6-GHz giving a peak performance varying from 5.6TFLOPS to 318 TFLOPS.

3. The IBM System p5 595 server uses fifth-generation 64-bit IBM POWER5 technology in up to 64 core symmetric multi-processing (SMP) configurations with IBM Advanced POWER5+ (2.1/2.3GHz).

Evolution of Supercomputers

Since the Cray T3E, supercomputers have steadily improved their shared address space architectures. Many of the improvements in address space organization have been utilized by supercomputers. Since the T3E was released, Cray merged with SGI and acquired OctigaBay Systems Corporation. As a leader in the development of supercomputers, Cray, Inc. has released several more machines superior to the T3E. The most noticeable feature in newer supercomputers is the implementation of distributed shared memory (DSM). Below are 10 of the fastest supercomputers in recent time.

Figure 4.1 - Top 10 Fastest Supercomputers

Ranked as the fastest supercomputer in the world and released after Cray's T3E, IBM's Blue Genie supercomputer is currently able to reach speeds of 367 TeraFlops. A node in the Blue Gene architecture consists of a single ASIC, shown in Figure 4.2, along with a read/write locking mechanism, scratch DRAM, and L1 and L2 data caches. The ASIC contains two 700 MHz PowerPC 440 embedded processors. Each of the processors has two floating point units, DRAM controller and the logic to support multiple message sharing system. Theoretically a peak performance of 5.6 GFLOPS can be acheived by the dual FPUs. To reduce overhead, peer node CPUs are not cache coherent.



Figure 4.2 - Blue Gene node ASIC


Due to the nature of implementing the cores on ASIC chips, low power operation was maintained, allowing high area density and scalability of nodes. Each node operates in a range around 17 Watts. These properties allow up to 1024 nodes plus additional IO nodes in a standard 19" cabinet, within reasonable limits of electrical power supply and air cooling.

A hybrid of bus and message passsing architectures are used in the Blue Gene as is necessary for high speed symmetric memory locations that allows scalability. Each node is part of a 3D toroidal bus network architecture, a bus network for collective communication, and a message passing global interrupt network.

References

[1] Sunfire12k

[2] Cray XT4

[3] IBM

[4] Cray X1E

[5] Wikipedia: Vector Processor

[6] Wikipedia: Blue Gene

7. McCurdy , C. & Fischer, C. (2003). User-controllable coherence for high performance shared memory multiprocessors. Proceedings of the Ninth ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming Archive, , 73-82.

8. Culler, D. E., Singh, J. P., & Gupta, A. (1999). Introduction: Convergence of Parallel Architectures: Shared Address Space. In D. E. Penrose (Ed.), A Tutor's Guide: Helping Writers One to One (pp. 28-37). San Francisco, CA: Morgan Kaufmann.

9. El-Rewini, Hesham. Abd-El-Barr, Mostafa. Advanced Computer Architecture and Parallel Processing (chapter 2 & 4), John Wiley & Sons, Inc. (US), 2005

Retrieved from "https://wiki.expertiza.ncsu.edu/index.php?title=CSC/ECE_506_Fall_2007/wiki1_7_a1&oldid=3537"