Academia.eduAcademia.edu

Figure 3 – uploaded by M. Gzara

See full PDFdownloadDownload figure
TABLE III. UTILITY VS. GRID COMPUTING

Table 3 III. UTILITY VS. GRID COMPUTING

Related Figures (16)

Figure 1. Serial computing (a) vs Parallel computing (b) [13] a single processor (Figure 1). While in_ parallel computing the problem is broken into discrete parts that can be solved concurrently. Instructions from each part execute simultaneously on different processors. An overall control/coordination mechanism is employed [13].
Figure 1. Serial computing (a) vs Parallel computing (b) [13] a single processor (Figure 1). While in_ parallel computing the problem is broken into discrete parts that can be solved concurrently. Instructions from each part execute simultaneously on different processors. An overall control/coordination mechanism is employed [13].
Figure 2. The need for distributed computing In fact, for decades, the Von Neumann architecture [19] has known many improvements like: increasing Clock Frequency, Memory Hierarchy / Cache, Parallelizing ALU structure, Pipelining, Very-long Instruction Words (VLIW), Superscalar processors, Vector data types, Multithreaded concurrent programming, instruction-level parallelism (ILP), multicore processors, etc.
Figure 2. The need for distributed computing In fact, for decades, the Von Neumann architecture [19] has known many improvements like: increasing Clock Frequency, Memory Hierarchy / Cache, Parallelizing ALU structure, Pipelining, Very-long Instruction Words (VLIW), Superscalar processors, Vector data types, Multithreaded concurrent programming, instruction-level parallelism (ILP), multicore processors, etc.
Figure 3. Principal distributed computing paradigms In this section, we are going to present the principal distributed computing paradigms (Figure 3): the idea behind them, their goals, problems, and limits. Then in the next section we are going to compare them.
Figure 3. Principal distributed computing paradigms In this section, we are going to present the principal distributed computing paradigms (Figure 3): the idea behind them, their goals, problems, and limits. Then in the next section we are going to compare them.
In Peer-to-peer computing [25], in contrast to Client-Server systems, every node acts as both a client and a server providing part of the system resources. All client machines act autonomously to join or leave the system freely. This implies that no master-slave relationship exists among the peers. No central coordination or no central database is needed. In other words, no peer machine has a global view of the entire P2P system. The system is self-organizing with distributed control without the need for centralized coordination by central server [1]. Figure 4. Peer-to-Peer network
In Peer-to-peer computing [25], in contrast to Client-Server systems, every node acts as both a client and a server providing part of the system resources. All client machines act autonomously to join or leave the system freely. This implies that no master-slave relationship exists among the peers. No central coordination or no central database is needed. In other words, no peer machine has a global view of the entire P2P system. The system is self-organizing with distributed control without the need for centralized coordination by central server [1]. Figure 4. Peer-to-Peer network
Figure 5. Volunteer Computing Volunteer computing is a type of distributed computing that allows people donating their computing resources. The idle time of PC’s is used to do research and scientific projects. The system employs unused CPU cycles to fulfill the work. By breaking a monolithic job into a large number of units and distributing into different desktop computers called workers, it can perform jobs efficiently via massive parallelism [30].
Figure 5. Volunteer Computing Volunteer computing is a type of distributed computing that allows people donating their computing resources. The idle time of PC’s is used to do research and scientific projects. The system employs unused CPU cycles to fulfill the work. By breaking a monolithic job into a large number of units and distributing into different desktop computers called workers, it can perform jobs efficiently via massive parallelism [30].
Figure 6. Jungle computing system [55] stand-alone machines, clusters, desktop grids, mobile devices, and supercomputers, possibly with accelerators such as GPUs [56] (Figure 6). Jungle Computing Systems are Multi-Model / Multi-Kernel [56].
Figure 6. Jungle computing system [55] stand-alone machines, clusters, desktop grids, mobile devices, and supercomputers, possibly with accelerators such as GPUs [56] (Figure 6). Jungle Computing Systems are Multi-Model / Multi-Kernel [56].
Figure 7. Geographical distribution of the resources The emergence of high-speed networks at low costs and the availability of Internet to the large public have encouraged the geographical distribution of the resources at large scale (Figure 7).
Figure 7. Geographical distribution of the resources The emergence of high-speed networks at low costs and the availability of Internet to the large public have encouraged the geographical distribution of the resources at large scale (Figure 7).
Figure 8. Resources classification based on ownership It is also possible to have hybrid private and public resources combined together to form a large scale system. Or a community owned resources shared in the purpose of a specific common project between two or more teams and/or labs.
Figure 8. Resources classification based on ownership It is also possible to have hybrid private and public resources combined together to form a large scale system. Or a community owned resources shared in the purpose of a specific common project between two or more teams and/or labs.
Figure 9. Resources Availability First, as shown in Figure 9, the availability level can be directly seen in relationship with the resources ownership. In fact, volunteer computing provides the ess level since the resources are not dedicated and they are being used by their authentic owners for their purpose too. Second, it has also a direct relationship with the geographical distribution of the resources. The more the resources are not located in one central location, the ess probability of system’s unavailability and/or failure 1S.
Figure 9. Resources Availability First, as shown in Figure 9, the availability level can be directly seen in relationship with the resources ownership. In fact, volunteer computing provides the ess level since the resources are not dedicated and they are being used by their authentic owners for their purpose too. Second, it has also a direct relationship with the geographical distribution of the resources. The more the resources are not located in one central location, the ess probability of system’s unavailability and/or failure 1S.
TABLE IL. “ABLE II. GRID VS. CLUSTER COMPUTING
TABLE IL. “ABLE II. GRID VS. CLUSTER COMPUTING
As shown in the Figure 10, the distributed computing paradigms evolved from the cluster integrating tens to thousands of computers to Peer to Peer, volunteer, utility and now the mobile and pervasive computing integrating hundreds of thousands to millions of different types of computing resources like supercomputers, PCs, Mobile phones, etc. Figure 10. System’s scalability
As shown in the Figure 10, the distributed computing paradigms evolved from the cluster integrating tens to thousands of computers to Peer to Peer, volunteer, utility and now the mobile and pervasive computing integrating hundreds of thousands to millions of different types of computing resources like supercomputers, PCs, Mobile phones, etc. Figure 10. System’s scalability
Figure 11. Utility computing paradigm When the system runs out of resources, using the utility model, it acquires those resources from another source (pay-as-you-use model) from a_ resources provider [49]. Therefore, the ability of utility computing to expand and contract ensures that an application can ride the hype curve to success [54]. Users (consumers) pay providers for using computing power only when they need to. All grid/cloud platforms are regarded as utility service providers. However, cloud computing offers a broader concept than utility computing [1].
Figure 11. Utility computing paradigm When the system runs out of resources, using the utility model, it acquires those resources from another source (pay-as-you-use model) from a_ resources provider [49]. Therefore, the ability of utility computing to expand and contract ensures that an application can ride the hype curve to success [54]. Users (consumers) pay providers for using computing power only when they need to. All grid/cloud platforms are regarded as utility service providers. However, cloud computing offers a broader concept than utility computing [1].
H. Barkallah et al.: Evolution of the Distributed Computing Paradigms: a Brief Road Map
H. Barkallah et al.: Evolution of the Distributed Computing Paradigms: a Brief Road Map
Figure 12. Comparing the utility computing paradigms As a conclusion, the Figure 12 summarizes the most important characteristics of the utility computing paradigms: service, grid, and cloud.
Figure 12. Comparing the utility computing paradigms As a conclusion, the Figure 12 summarizes the most important characteristics of the utility computing paradigms: service, grid, and cloud.
Related topics:
Computer Science
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved