Grid Computing Standards - Essay Example

Literature Review GRID (Clustered) SAN computing standards Storage Area Network, or SAN, is, in simple terms, a network designed to attach different storage devices. These devices include disk array controllers, tape libraries and servers. Enterprise storage is common in enterprise storage configurations. Two variations of SANs are: 1. A network whose primary purpose is only to transfer data between the computer systems, such as a personal computer, an automatic teller machine (ATM) and storage elements. A SAN in this variation consists of an infrastructure that employs physical connections and a management later. The management layer organises the actual physical implements of the network such as the physical connections, storage elements and the actual computer systems in order to ensure data transfer is both strong and secure. 2. A storage system which consists of network elements such as storage devices, computer systems, servers, control software (such as server administration and site server) that will communicate over the network. Storage networks are distinguished from other forms of network storage devices simply by their low-level access method they use and is very similar to such network devices as ATA, disk drives and SCSI hardware. Inside a storage network, the server will issue a request for specific blocks of data and this device on the storage network will then send requests across the network. In the clustered GRID infrastructure would be a series of SAN devices that would then integrate as part of the larger network in order to share the storage capabilities inside the GRID. As such, each device would then be added to the larger network that can be used by other networked devices such as a computer workstation. An example of this computing infrastructure, courtesy of Force10, shows how this cluster/GRID computing network is positioned to utilized these storage devices The Differences between GRID and SAN GRID computing is the transformation of a computer infrastructure into an integrated Virtual Organisation that allows for dynamic collaboration and the ability to share resources from anywhere in the world. This sharing provides users with an unprecedented amount of computing power, especially for those in the field of scientific investigation and collaboration in which the needs of the computer power cannot necessarily be handled by one such computer. Through integration inside the GRID of such supercomputers will enable users to access power without the need to purchase larger systems. GRID computing is based on three concepts as outlined by Reddy (2004) "as: Virtualization: severing the hard-coded association of resources to systems Resource Allocation and Management: dynamically allocating resources on demand, and managing them and finally, Provisioning: configuring resources whenever and wherever needed." (Reddy, 2004) Kalzar Amin, Gregor von Laszewski and Armin R. Mikler Kalzar, et al describe the term Grid computing as commonly referred "to a distributed infrastructure that promotes large scale resource sharing in a dynamic institutional "virtual organisation" (VO). A computational Grid forms a closed network of a large number of pooled resources providing standardized, reliable, specialized and pervasive access to high-end computational resources." Typically, in order to establish a computational Grid, several institutions pool their resources such as computational cycles, specialized software, database servers, network bandwidth, and people. As a result of this "pooling" global policies will be set for the virtual organisation which will in an essence identify each of the participating entities' roles and responsibilities, much like in a LAN server networking scenario. Each of the site institution administrators, who are generally trained as network administrators will then enforce these policies at the domain level. The GRID administrators will then provide each of the GRID users their appropriate credentials and through these credentials will the users access the allocated GRID resources no matter where they are geographically located. Although the GRID community seems the most viable in network resource sharing, there are several technical challenges that are important to be addressed before the GRID can be adopted. The largest challenge in any P2P GRIDS is the enforcement of both dynamic and adaptive security models that will overlay a secure framework over an insecure network. "In a Grid approach, the notion of trust and reputation is implicit within a closed VO formation. Every collaborating entity trusts other entities. Grid resources are not anonymous; they are accountable for any misconduct. Further Grid resources adhere to well-established mechanisms for authorization and authentication, restricting the use of resources by non-collaborating or rogue users." (Kazar et al, n.d.) Normally, ad hoc and federated Grids require an adaptive security model that will incrementally build a secure Grid community based on the notion of trust and reputation. David De Roure, Mark A. Baker, Nicholas R. Jennings and Nigel R. Shadbold The vision of GRID computing as outlined by DeRoure, et al (2006) needs to address "three main issues: heterogeneity: A Grid involves a multiplicity of resources that are heterogenous in nature and might span numerous administrative domains across a potentially global expanse. As any cluster manager knows, their only truly homogeneous cluster is their first one scalability: a Grid might grow from a few resources to millions. This raises the problem of potential performance degradation as the size of a Grid increases. Consequently, applications that require a large number geographically located resources must be designed to be latency tolerant and exploit the locality of accessed resources. Furthermore, increasing scale also involves crossing an increasing number of organizational boundaries, which emphasises heterogeneity and the need to address authentication and trust issues. Larger scale applications may also result from composition of other applications, which increase the 'intellectual complexity' of systems. Adaptability: in a Grid, a resource failure is the rule, not the exception. In fact, with so many resources in a grid, the probability of some resource failing is naturally high. Resource managers or applications must tailor their behavior dynamically so that they can extract the maximum performance from the available resources and services." (DeRoure, et al, (2006) A key to also tackling heterogeneity is through both setting and using a standard system of such things as system APIs. "Systems use varying standards and system APIs, resulting in the need to port services and applications to the plethora of computer systems uses in a grid environment. As a general principle, agreed interchange formats help reduce complexity, because n converters are needed to enable n components to interoperate via one standard, as opposed to n2 converters for them to interoperate with each other." (DeRoure, et al, 2006) DeRoure, et al (2006) state that the data infrastructure "itself can consist of all manner of networked resources ranging from computers and mass storage devices to databases and special scientific instruments." There are other computational resources that are available for the infrastructure which include supercomputers and cluster resources. Grid applications are generally characterised as having a huge scale of data and computations that require the use of a high density infrastructure. The main design features required at the data and computational fabric of the Grid are: Administrative hierarchy Communication Services Information Services Naming Services Distributed File Systems and Caching Security and Authorisation System Status and Fault Tolerance Resource Management and Scheduling User and Administrative GUI IBM Redbook. http://www.redbooks.ibm.com/redbooks/pdfs/sg246650.pdf IBM Redbook provides a very descriptive adaptation of the GRID infrastructure and advise that there are a fundamental grid models based on the type of basic services provided. Resources can basically be computing power, provided by servers or individual computers, data storage capacity, provided by information and data repositories, or network bandwidth, provided by networked infrastructures. "These grids expand the principles of the basic models, leveraging and combining them in innovative ways to provide more advanced services." (IBM Redbook, 2006) The Redbook explains the models of the GRID as follows: computational grid: is an infrastructure that allows resources to donate computing power to the grid whenever the workload demands. This infrastructure is suitable for applications that demand: as much processing power as possible or additional processing power during certain periods of time and a single machine cannot provide it or at too much cost it is usually associated with resource scavengingi n desktops machines and underused servers. Such a grid takes advantage of the idle resources in the virtual organization. Just think about the cycles of CPU that are unused when a typical user browses on the Internet, reads an e-mail, or creates office documents like presentations or word processing files. Examples of these grids include server-oriented grids, desktop-oriented grids data grid: within a data grid infrastructure are the components used to provide grid capabilities to the data and information virtualization disciplines. It provides the ability to supply homogeneous access to heterogeneous repositories of data. It allows the data consumes to see a unified image of the respective information or data spread across different resources, potentially based on different technologies. Information infrastructure: information is usually defined as "meaningful data". Meaningful data is often associated with the unit of information that means something useful to the end user. Information grid is intended to integrate different sources of information in a comprehensive way. It allows applications and users to see a single database that hides the complexity of accessing multiple databases. This grid should implement connectors to the final databases, which surely are heterogeneous and geographically distributed. It is desirable as well to have replication and caching mechanisms that make the management and usage of the IT infrastructure more efficient. Examples of these grids include file system and block data infrastructure. network grid: in a typical corporate network, computers are very often permanently connected to it while using only a portion of its bandwidth. Every machine, server and desktop, has underused network bandwidth, which can be considered as an idle resource. When a given user or machine requires more resources from the network, a bottleneck is reached. This type of grid is sometimes referred to as "distributed peer to peer" grid, network grid communication grid or "grid delivery". (IBM Redbook, 2006) Economic Drivers: Many enterprises and institutions over the past decade or so have relied heavily on investing in their IT infrastructures without really looking at the larger picture of whether the infrastructure is satisfying the needs of the customers and users or is it simply a highly distributed infrastructure. According to economists in many IT sectors, a "sizeable portion of all enterprise computing has been estimated to run at only 20% to 40% of its total capacity. This costly condition - a paid-for, yet idle, computing resource has often resulted from three factors: Only one application deployed per operating environment Systems were sized with performance head room in mind, as forecasted from more demanding workloads and The duration of peak loads was often short. (Reddy, 2004) GRID computing provides a more efficient means of dispursing IT services and managing them at a reduced cost as through introducing simpler technologies, greater bandwidth capabilities, automation of processes and configurable infrastructures that are not decidedly too complex, the impact of GRID computing can provide nearly 100% uptime at a fraction of the cost computed into billable server usage times. By introducing a GRID computing infrastructure, resources are evenly distributed throughout the enterprise and as the cost of managing today's more static and fixed environments, the benefits reaped will be exponential. Peer-to-Peer Networking Saad N. Ahmad. Business Models of P2P Companies Ross lee Graham defines P2P through three key requirements: a) operational computer of server quality b) an addressing system independent of DNS, and c) able to cope with variable connectivity The P2P architecture allows anonymous peers to share their resources, with or without a limited interaction with a centralized server. In its purest form there is no concept of server and all participants are 'equal" peers. A peer gives some resources and in return obtains other resources that are essential for the functioning of the system and benefit all peers. P2P peers play the role of client and server at the same time. P2P model became popular with the rapid spread of music exchange software "Napster" and is increasingly finding applications in other areas like grid computing. Characteristics of the P2P architecture include: cost sharing, higher reliability and scalability, privacy and anonymity, resource aggregation and synergies, greater autonomy, dynamic environment and ad hoc collaboration. Inherent in the privacy and anonymity of P2P environmental is the issue of lack of responsibility. A lack of central authority makes it difficult to enforce contracts. In addition, there is no true P2P payment mechanism so far, which facilitates exchanges without depending on some sort of intermediate authority and without infringing the privacy concept. Conclusion Based on the two architectural models and application usability, the use of a network GRID system would be the most beneficial in replacing large file server hardware and software applications. As all computers in the GRID share resources and hard drive space, each user can be responsible for sharing applications and resources evenly distributed. By having applications shared equally it can then replace the need to have a file server and with different computers having a copy of the applications on their computers, there is no possible way to encounter server downtime. References Bell, G. and Gray, J. (2002) "What's next in high-performance computing"Communicaitons of the ACM, vol. 45, no. 2, pp 91-95, Fe b 2002 [Online] Available at http://doi.acrn.org/10.1145/503123.503129 D. De Roure, M. Baker, N.R. Jennings, and N.R. Shadbolt. "The Evolution of the Grid" [Dissertation]. Retrieved April 15, 2006 from http://www.semanticgrid.org/documents/evolution/ evolution.pdf "Doe Science Grid". Web Page [Online] Available at http://www.doesciencegrid.org/ Foster, I and Kesselman, C, Eds. (1998) "The Grid: Blueprint for a New Computing Infrastructure" Morgan Kaufmann Publishers, July 1998. Foster, I., Kesselman, C., and Tuecke, S. (2002) "The Anatomy of the Grid: Enabling Scalable Virtual Organizations". International Journal of Supercomputing Applications. Vol 15, no. 3. [Online] Available at http://www.globus.org/research/papers/anatomy.pdf Foster, I., Kesselman, C., Nick J. and Tuecke, S. (2002) "The Physiology of the Grid: An Open Grid Services Architecture for Distributed Systems Integration." Web Page. Jan. 2002. [Online] Available at http://www.globus.org/research/papers/ogsa.pdf "Gnutella Homepage". Web Page [Online]. Available at: http://www.gnutella.wego.com "grid.org Homepage". Web Page [Online]. Available at: http://www.grid.org Grimshaw, A.S. and Wulf, W.A. (1997) "The Legion Vision of a Worldwide Virtual Computer". Communications of the ACM, vol. 40, no. 1. pp. 39-45. [Online] Available at http://legion.virginia.edu./copy-cacm.html GriPhyN - Grid Physics Network". Web Page. [Online]. Available at http://www.griphyn.org/index.php Johnston, W.E., Gannon, D. and Nitzberg, B. (1999) "Grids as Production Computing Environments: The Engineering Aspects of NASA's Information Power Grid". Keahey, K. Fredian, T. Peng, Q, Schissel, D.P., Thompson, M. Foster, I., Greenwald, M and McCune, D (2002) "Computational Grids in Action: The National Fusion Collaboratory". Argonne National Laboratory. Tech. Rep., 2002. [Online]. Available at: http://www-unix.mcs.anl.gov/keahey/papers/FusionPaperSubmitted.pdf "Napster Homepage". Web Page [Online]. Available at: http://www.napster.com "National Research Grid Initiative". Web Page. [Online]. Available at http://www.naregi.org S. N. Ahmad. (2003). Business Models of P2PCompanies. [Dissertation] Reddy, S. (2004). Grid Computing: Crossing the Chasm. [Online]. Available at http://sciencecareers.sciencemag.org/career_development/previous_issues/articles/3220/grid_computing_crossing_the_chasm/(parent)/12095 Thain, D., Tannenbaum, T., and Linvy, M. (2003) "Grid Computing: Making the Global Infrastructure a Reality." "The Global Forum of Grid Computing" Web Page. [Online] http://www.gridforum.org "The DataGrid Project". (2000) [Online]. Available at http://www.en-datagrid.org/ "TeraGrid" (2001). Web Page. [Online]. Available at http://www.teragrid.org Read More