Database Machine Taxonomy

Introduction

In the traditional approach of the database system, the data is used to be stored in secondary storage devices. Due to this, the ability to perform any arithmetical or logical computing operation is limited to the central processor and the data has to be moved to the main memory from the secondary storage devices, attached to the central processor. Once the data is transferred, the processor can access it. A large amount of redundant data is also retrieved and processed. It has been determined that on average, only 10 % of data is relevant. To reduce this wasteful movement the index approach is used. However, it also takes considerable storage space and generates traffic on the input/output channels, and generates a heavy processing load on the processor.

Database Machine Taxonomy

In database machines, the approach is taken to get rid of the database management function to the special processor and at some level of computing capability. This extra level of computing capability makes the system feasible to decide that a given set of data is useful in the evaluation of a query without transferring the data to the CPU. 

A number of approaches have been proposed to move the computing power closer to data. These approaches are:

• Backend software approach 
• Backend database computer approach
• A processor associated with memory/Intelligent memory approach
• Special hardware approach

Backend Software Approach 

It is also called the backend computer approach. In this approach, the host computer (where the applications are located) is attached to a conventional database management system and a dedicated general-purpose computer. This dedicated computer is responsible for locating and retrieving required data by carrying out the database functions. It also ensures security, enforces consistency, and provides recovery operations. A single backend computer is connected to a single or a number of hosts (not necessarily identical).

The backend system can be a conventional computer system or a system consisting of one or more specialized processors which use traditional secondary storage devices or associative memories. Associative memory has logic related to each bit or each word of the memory. 

Dedicating a separate system for the database has certain advantages and disadvantages:

• We can attain higher performance with such a system.
• The response time in the backend approach for a single query is likely to be worse.
• In this approach the data security is enhanced because the data is under the control of a dedicated system and any user does not have access to the system.
• This approach is cost-effective.
• The reliability of the database is improved.
• Since the system has more components the probability of failure is increased. Failure of this system would lead the operations to a halt.

An alternative solution to recreate and increase the reliability of the system is to incorporate multiple machines in the database functions.

However, with this scheme, some problems encountered are as follows:

• The need to decide the distribution of data on multiple backend computers.
• The maintenance of the directory containing the information about the data distribution.
• Consistency enforcement if the data is replicated.

Backend Database Computer Approach

Intelligent Memory Approach 

In this approach, adequate processing logic is linked with the secondary memory so that data can process before being terminated to the host processor. This approach determines the usefulness of data by intercepting them from the secondary storage devices. Moving redundant data to the host system which is running the database management system is not needed. The host could be a conventional system. 

The VLSI-based microprocessor provides the processing capability associated with the secondary memory and thus it is cost-effective. The processing capability may be associated with one of the following manners,

Processor per track

It is also known as a cellular logic device since a processor is attached to each track, and logic is linked with each cell of memory. In this approach the entire content can be processed in one pass, indexes are not required. The disadvantage of this scheme is that the data from all tracks is not necessarily required and the processing of irrelevant data is unproductive, hence the cost of this type of device is high.

Processor per surface

It process per pass the same amount of data as the processor per track approach. It is an attempt to attach the processing power with every read/write head of moving head type secondary storage devices. However, processing of all the data takes in passes or revolutions where m stands for the number of tracks per surface. In case of mechanical devices like disks and drums. The movement of the head takes a finite amount of time but in nonmechanical the switching can be done much faster. Indexes are necessary for this approach to reduce the number of passes.

Processor per device

This approach is a venture to reduce the number of processing elements and the cost. There is a single processor attached to each storage device in this scheme. The processor acts as a filter between the device and the host computer. Indexes are necessary to reduce the number of passes and amount of actually processed data.

Multiprocessor and Cache scheme

It is used to optimize cost performance. It allows the filter processors to process any one data from a number of storage devices or from a number of cells or tracks of a single device. The data is placed in n high-speed memory caches and there are m filter processors. The caches and processors are connected through an interconnection network. The data in caches could be from the same device or distinct devices.

Special Hardware Approach

In this approach, a specially designed computer (database machine) is used in hardware, software, and firm to perform the database function instead of a conventional computer. Thus the functions are performed through dedicated components, the performance is enhanced.

Summary

Database machine taxonomy is used to reduce the wastage of data and also helps to get reader processing redundant data because of the secondary storage devices. It helps to enhance the performance of database and make it cost effective and reliable.