Discussion Replies: Business Information

The student must then post a reply of at least 400 words by 11:59 p.m. (ET) on Sunday of the assigned Module: Each reply must incorporate at least 1 scholarly citation in APA format. Any sources cited must have been published within the last five years.

The reply can be from any one of the paragraphs listed below but 400 words with 1 scholarly citation

Shawn Watson

A distributed database is a collection of several different databases distributed among multiple computers. Discuss why a company would want a distributed database.

            The modern world runs on technology and being interconnected. Few organizations operate within a silo without some sort of distributed digital presence and connectivity to their customers, vendors, and stakeholders. In this global market data must be accessible to all those that require access to it. Centralized databases and Database Management Systems (DBMS) are effective in providing secure storage of data but they are not as proficient at delivering data access to geographically disbursed partners. Distributed Database Management Systems (DDBMS) provide many advantages to reaching potential customers through marketing applications, suppliers and other vendors through Enterprise Resource Planning (ERP) systems, and clients through point-of-sale or other means. When databases are distributed strategically across a geographically disbursed area of operations strategies can be applied to increase the efficiency of accessing that data. Reducing overall response times is to the user is the goal. The DDBMS can be configured to fully replicate the database to every location which is higher in cost but the most efficient flow of data to every user. Alternatively, it can be configured to selectively replicate the data tables that are in high demand to a particular subset of the organization. This approach reduces the cost compared to full replication everywhere and is better for the security of data by reducing the data footprint, while still delivering increased performance to the end user. The differences in these DDBMS strategies are transparent to the user.  Distributing the databases of the organization also increases the availability and reliability of the data. In centralized DBMSs, an outage of the system will affect the organization globally. In DDBMS environments, an outage will only affect the subset population that subscribes to that instance of the database. With proper failover techniques the transition to another distributed location will be mostly transparent to the user with only the likely factor of reduced system performance (Gupta, et al., 2011).

            Cloud computing has become one of the most common methods of implementing DDBMS for organizations. The cost is lower than on premises data centers while providing scalability that does not hinder growth with long procurement delays. Due to the versatile utility of DDBMS methods of increasing efficiencies have consistently been revolutionized. Abdel Raouf, et al. (2018) propose a method referred to as Cluster Based Distributed and Parallel Database System (CB-DDBS). This method is a three-tiered architecture consisting of the application client, DDBSM, and the distributed database cluster layer. They used an algorithm known as Optimal Fragment Reallocation and Replication (OFRAR). OFRAR uses the observed changes in data access patterns, time constraints, site constraints, volume of data, site overloads, and the value threshold. This technique has proven to decrease execution time within DDBSM environments. These techniques provide added value to organizations as they work to decrease data delivery times and increase overall reliability while maintaining security.

Productivity increases as rapid response times are achieved. Discuss what is considered an acceptable system response time for interactive applications.

            Response times for interactive applications will vary based on many factors. The response times that are acceptable for a given system will also vary. Mobile applications for mobility on demand (MOD) such as Lyft and Uber are great examples of an application that has an extremely short tolerance for response times. In an example transaction a user will send a query for a particular start point and destination. The timely response to this query is a critical factor in the success of the system. Traffic conditions, new queries, requests all impact the response to a particular query. In this case, the user’s response time is also an important factor. Once they decide they want to lock in a rate and driver, they will submit that back into the system. If they have spent a significant amount of time debating on which offer to choose or whether they want the ride, the time interval may have permitted a significant shift from the original query response. The driver may no longer be available, or they may be further away. For this reason, user response times is also an important factor to consider in complex interactive applications. With the delays in these systems being variable based on the user and the operator’s responses, the technology response times must be extremely efficient to reduce user and operator confusion (Yu & Hyland, 2020).

            In their research, Hoxmeier and DiCesare (2000), found that user satisfaction peaked in application response times between zero and two seconds. Satisfaction fell in response times over three seconds but did not dramatically decrease until they increased over nine seconds. Users generally offered that they would continue to use an application if the response times stayed under nine seconds. Users found a negligible difference in the ease of use of the application based on response times. They found that user satisfaction was not only linked to the response times, but also to the experience level of the user and the expectations the user had of the system. They theorized that the more experienced employees would naturally be more tolerant of slower response times but were surprised to find that experience did not play a factor in the satisfaction of the software. They did find that considering the software easy to use was directly correlated with their satisfaction with the system.

A fully centralized data processing facility is centralized in many senses of the word. Discuss the key characteristics of centralized data processing facilities.

            Centralized data processing takes place in a single point in which all data is brought to and managed by a single computer. This computer is usually extremely robust with storage, memory, and redundant processors. While centralized data processing facilities converge on central controllers and are housed in a single facility, there are many devices involved in making the architecture of the facility run. These can be setup in a client/server model, three-tier or n-tier models, clusters, or peer-to-peer architecture. Centralized processing is an effective method for organizations with single geographical locations where distributed networks would slow the convergence of data unnecessarily. Organizations where data is collected and stored in the same location are good candidates for a centralized processing facility. Examples of this include a boutique store with a point-of-sale system or a non-chain hotel. Both organizations keep track of pricing information, sales and purchasing data, and a myriad of other processes. A large hotel may have a robust network with multiple servers, offices, data storage, web services, and more. If they are a single location organization, their processing should stay local to reduce cost and maintain efficiencies (Krishnan, 2013).

The three-layer architecture is historically the most common architecture of data processing centers. In three-layer architecture consists of a core, aggregation, and edge, all operating on layer two switching to tie the servers together. There are several other established topologies common to data centers. The Fat-tree topology is like three-layer in the use of server, edge, and aggregation switches but differing in the establishment of pods. Fat-tree separates the server groups into pods which then connect the pods back to the core. BCube and DCell topologies break the datacenter down into BCubes or DCells which are numbered and configured to interlace in a modular fashion. BCube architecture uses a higher-level switch to redundantly connect each of the BCubes to one another while the DCells are directly interconnected in a mesh from server to server to tie the DCells together. These topologies are commonly found in both centralized and decentralized data processing facilities (Souza Couto, et al., 2015).

Equipment and communication redundancies are common in today's data centers. Discuss the major types of equipment and communication redundancies found in today's data centers.

            Modern data centers often have service level agreements that require availability rates as near to 100% as possible. The term six-nines meaning 99.9999% uptime has been coined to describe these stringent standards. Uptime agreements cannot be possible in an environment with single points of failure. This is what drives the use of redundant equipment throughout datacenters. Every piece of equipment from power distribution and cooling all the way to servers, cables, and switches are duplicated in these environments to ensure near perfection. Servers have adopted this model with at least two servers being provisioned in for each function, usually designating an active and a standby for failover purposes. Data is replicated constantly between the active and standby, so they are essentially mirror images of each other; this means that if one becomes corrupt and a failover occurs, the transition is transparent to the user. In a Fat-tree topology, the edge and aggregation switches are deployed in redundant pairs with each connected to each other and to their failover. The Fat-tree pods are then able to provide additional redundancy by duplicating independent instances across the datacenter. Finally, distributed data centers ensure that one geographical disaster or ISP outage will not produce an outage to the end user (Guo & Yang  , 2015).

            Cui et al. (2013) propose a unique architecture which is designed to target the redundant traffic that is found within datacenter networks. With all the redundant servers and switches it is common for redundant data to congest the network within the datacenter. While some of this redundant transportation is a requirement of the fault-tolerant architecture, much of it is truly redundant and wasteful. Proxy servers are a common method to reduce redundant traffic on the network by caching commonly accessed data. The uncommon architecture recommended by their team is the use of wireless network cards in the servers to tie servers into the network mesh. Additionally, they utilize router caching of data units to reduce the repetitive data transmissions from server to server. The centralized nature of datacenter networks allows routers to interface wirelessly with many servers and build the cache relationships necessary to reduce traffic on the network. This helps to avoid congestion while still providing redundant interfaces and routes to ensure transparent fault tolerance to the user. Their design has proven in simulations to provide excellent performance while maintaining redundancy and centralized control (Cui et al., 2013).