&= nbsp; By the close of the 20th century, advances in data storage capacity, internet availability, bandwidth, as well as increased transmission speeds = allowable through fiber optics, microwave, satellite, and WiFi availability caused ma= ny analysts to reconsider the distributed data storage. At once, issues such as timeliness, data availability, mass data storage, and security allowed organizations for the first time to consider centralization of data as reasonable and appropriate.
&= nbsp; Yet even with the availability of high-speed internet services, and private wide area networks (WANs), there remains an arena where distributing data between locations, organization, or sales staff may make sense.
&= nbsp; This essay will discuss some of the traditional reasons for considering a distributed data storage strategy, and will then examine recent technologic= al advances which could have a direct impact on the continuation of that strat= egy.
&= nbsp; In large organizations, especially in organizations in which facilities are geographically disparate, distributing data between the facilities made sen= se for a lot of reasons. First, distributing data allowed the remote locations= to function operationally. The information that they needed to perform their t= asks was local and available. Second, giving the locals control over their own d= ata also made them responsible for their own accuracy and correctness (two different issues).
&= nbsp; At the onset, it is important to make the distinction between distributed syst= ems and distributed data. Whether by virtue of the internet, intranet, or extra= net, the most common system organization is distributed. The entire design of a distributed system is to put as much processing power as close to the user = as possible. Fitzgerald & Dennis (2002) discuss multiple approaches to distributed systems architectures such as:
Distributed data, on the other hand, is a perspective on where the storage of data should occur in an organization. The question in distributi= ng data is simply: should the data be stored in a central location or disburse= d to the local regions? When data is centralized, it is typically easier to cont= rol, maintain, and protect. Simple activities like backups can be done routinely= and consistently under the auspices of a limited number of people. When data is distributed, local data management can happen quickly and the local users w= ould by default only have access to data relevant to their facilities. They woul= d be responsible for their own data accuracy and then be required to transmit updates or new data back to the central database via transactions.
With certain advances in Internet and intranet technologies, especi= ally the speeds which can be obtained in T-1 and T-3 technology as well as fiber optic FDDI, and faster database structures operating on multiple terabyte f= ile systems, all but data in the largest files can be accessed in less that a f= ew seconds to less than a second. At these speeds, local groups can access the= ir specific information directly from a central facility with the appearance of the network being their own.
&= nbsp; Even with new technologies and faster transmission speeds available in a variety= of medium, there exists still an easy approach on how and where to distribute data. Limoncelli & Hogan (2002) describe a push and pull concept. Data control should either be enforced on the locals at the direction of the cen= tral authority (push), or be demanded from the central by the local authorities (pull).
&= nbsp; In a push situation, the central authority allows users to extract a copy of t= he data relevant to them which is updatable, as well as have read-only access = to various other corporate data with restrictions which would allow a particul= ar local facility to make intelligent decisions. For example, customer lists or particular account data but not the operational data from other facilities. The locals would = then be able to make updates to the information which would then be made availab= le back to the central data base asynchronously.
&= nbsp; In a pull situation, it is the local facilities which are driving the demand f= or the distribution of data. For several reasons, the locals have decided that they need asynchronous access to their specific data. An example of this mi= ght be a local sales staff needing a copy of customer information which making sales calls. Customer data, orders, changes, and requirements may be changed while the salesperson is at the customer site. In this case, network access= may not be immediately possible or reasonable. The salesperson would then return with the data changes (either physically or virtually) later to update the central data base.
&= nbsp; In a world of multiple hundreds of megabyte transfer rates and multiple hundre= ds of gigabyte personal storage devices, it is no longer a question of “= can data be distributed” but rather a question of “to what extent c= an data be distributed”? That is, the distribution of the data in and of itself is an aspect important to consider. Subramanian (2000) notes that an “important aspect of a distributed computing environment is the abili= ty of the processors attached to the LANs to do multiple functions and return = the results.” While data integrity and control is a critical aspect of any system, the levels of processing efficiency which can be attained in a distributed environment provide significant benefits to the organization as= a whole.
Fitzgerald, J. & Denniz, A. (2002). Business data communications and networking. New York, NY: J. Wilery & Sons. 7d.
Limoncelli, T.A. &am= p; Hogan, C. (2002). The practice of system and network administration. Boston, MA: Addison-Wesley.
Subramanian, M. (200= 0). Network management principles and practice. Reading, MA: Addison Wesley Longman. 12.
Wayne Machuca &= nbsp; &nbs= p; &= nbsp; &nbs= p; &= nbsp; Guidelines for Data Distribution - 1
CMD890B – Doctoral
Comprehensive Review
11/1/2005