Understanding Storage Redundancy, RAID Levels, And Pros And Cons

Types of Data Redundancy

Question:

Describe about the Storage Redundancy,Data computation-"Moore's Law" and Transport Layer Protocol?

Storage redundancy involves the duplicity in the storage of data. It is a condition that is created within a database in which the same is place in two separate spaces. Storage redundancy is commonly known as data redundancy. It can be two different fields in a single database, or in multiple software platforms. Whenever data is repeated, this is known as data redundancy. Storage redundancy may occur by accident, but sometime it is also done intentionally for backup and recovery purposes.

In a general manner data redundancy has different classifications that are based on what is most appropriate in database management, and what is wasteful or excessive. Wasteful storage redundancy normally occurs when a given data need not to be repeated, but it ends up being duplicated because of coding inefficiency or process complexity.

A positive approach of data redundancy works as a safeguard for data that promote consistency. Many web masters consider it acceptable to store data at multiple places. The key point is to have focused master space for this data to be stored, so that it gives a way to update all of the places wherever  data redundancy occur,  via  that central access point. On the other hand, data redundancy will be a big problem if there is no central access point, with data inconsistency, where update in one does not automatically update another. As a result, the data that are supposed to be same end up with different values. So storage redundancy is not perfect, and has its pros and cons.

 The most obvious virtue of storage redundancy is that it protects data in case of drive failure in real time. It means if the internal drives in a RAID fails, when you are working on a file but one of the storage device is working normally. It will indicate that one of your internal drives has encounter error or fails, and offers you to the chance to back up your important data and replace the drive. Then your device will blend the replaced drive itself to become part of the RAID. During that time, the storage device is still available. In short, storage redundancy gives you the facility of data protection and though internal part will fail another will be there to back up the data.

The first limitation of redundancy is the cost; you have to spend bugs on multiple drives. For example RAID 1 setup, basically requires spending just double on internal drives.

Another disadvantage of storage redundancy is that it doesn't provide protection to data against physical disaster, like fire or flood. Storage Redundancy also doesn't offer facility of versioning, which include the data saving in different versions.

RAID is a virtualization technology used to store data that combines the number of disk drives into a logical unit to improve the data redundancy and performance. RAID enables the information to access several disks. Many techniques such as striping (RAID Level 0), mirroring (RAID Level 1), and parity (RAID Level 5) are involved in RAID to achieve redundancy, increased bandwidth, lower latency, and maximized ability to recover.

Data Redundancy as a Safeguard for Data

RAID distributes pieces of data across every drive in the array. Then RAID breaks down the data into chunks. Then each chunk is written to the hard drive in the RAID in accordance to the employed RAID level. When the data is read, the process will be reversed, showing illusion that the number of drives is actually one large drive. Below are several levels choices that are most commonly used.

RAID 0— Striping

RAID 0 level consists of striping, without parity or mirroring. The capacity of a RAID 0 is the sum of the disk’s capacities in the set. There is no additional data redundancy to handle disk failures, just as a spanned volume. Therefore, failure of one disk causes the entire RAID 0 volumes lost. It reduces the possibilities of recovery of data while compared to a broken volume. Data Striping distributes the file’s contents roughly among all disks, which makes concurrent write or read operations on the numbers of disks. The concurrent read or write operations make the profit of most write and read operations that is equal to the output of one disk multiplied by the multiple disks. Increased output is the big profit of RAID 0 i.e. Striping.

RAID 1—Mirroring

RAID 1 is also known as mirroring, without striping or parity. In this approach, data is written to two or more drives, producing a "mirrored set". Any read request will be serviced by any drive in the mirrored set. If a request is broadcast to each drive in the given set, this request can be serviced by the drive which first accesses the data, improving the performance. Sustained read output, approaches the sum of outputs of each drive, just as for RAID 0. Actual throughput of most RAID 1 is slower than the fastest drive.

RAID 2

RAID 2 includes a bit-level striping with Hamming-code parity. All the disk spindle rotation can be synchronized and the data is striped in such a way that each sequential bit is placed on a different drive. Hamming parity is evaluated across the corresponding bits and finally stored on one parity drive.

RAID 3

RAID 3 includes the byte-level striping with concerned parity. Rotation of all disks spindle is synchronized and the piece of data is striped in such a way that each byte is on a different drive. The parity is evaluated across corresponding bytes and finally stored on a dedicated parity drive.

RAID 4

RAID 4 includes the block-level striping with parity. This striping was previously used by Net App, but has now replaced by an implementation of RAID 4 with two parity disks.

RAID 5

RAID 5 includes the block-level striping with dedicated parity. In this approach, parity information is divided among the distinct drives. It requires that all the disks drives but one to operate. On the failure of single drive, subsequent reads can be evaluated from the distributed parity in such a way that no data is lost. It requires at least three disks and seriously affected by trends and chance of failure during rebuild.

The Advantages and Disadvantages of Storage Redundancy

RAID 6

RAID 6 includes the block-level striping with double parity. Double distributed parity gives fault tolerance for two failed drives. It makes larger RAID groups especially for high-availability systems. The large-capacity drives take longer to restore. In single drive failure reduce the performance of entire array. The larger the size of drive capacities and the larger the size of array results to choose RAID 6 instead of RAID 5.

Moore’s Law is a computing term which came into existence in around 1970. The most simplified version of Moore’s law states that the overall processor speeds, or power for computers will be double in every two years.  The technicians from different companies of computer prove that the Moore’s law is not very popular but the formula is still accepted.

To divide the law even further, it is specifically said that the number of transistors that are embedded on an affordable CPU would get double in every two years but ‘more transistors’ is more accurate.

In fact, the next-generation of processors, have to contend with a memory bottleneck which is known as the bandwidth between the CPU and a computer’s memory channels or the instruction level parallelism (ILP) or the availability of discrete parallel instructions and the chip’s temperature and power consumption.

The power wall now defines the limit of the speed of the modern CPU. As the processors have become more capable, the heat production as well as energy consumption has grown rapidly. Due to this problem, chip manufacturers have been forced to create “systems on a chip”– smaller, specialized processors.

What is Interrupt Request?

An interrupt request is a signal that is sent from a device to a processor.  It indicates that attention is required in order to process a request. A hardware interrupt request is induced by hardware peripheral while a software interrupt request is induced by software instruction. The result of both IRQ in processor status savings, and revert IRQ serving use an interrupt handler routine.

Interrupts are used for computing multitasking and eliminate the requirement for the processor to sample the lines when waiting for external events.

An interrupt request is made to the processor by using programmable interrupt controllers (PICs) that manage the interrupts. It can be level-triggered interrupt or edge-triggered interrupt. Level-triggered indicate that the line is held via the device at active level, interrupt triggering until it is served. Edge-triggered indicate that the device interrupt the line from level 1 to 0. Interrupt request levels are assigned to devices so that it indicate their identities. IRQ can be categorized into the following types:

• Maskable interrupt (IRQ): A maskable interrupt is a hardware interrupt that can be ignored through setting bit in an interrupt mask register's (IMR).
• Non-maskable interrupts (NMI):  It is hardware interrupt that deficient an associated bit-mask. Non Maskable Interrupts are used for the maximum priority tasks like timers.
• Inter-processor interrupts (IPI):  It is a special interrupt case that is generated by one processor to another processor in multiprocessor system.
• Software interrupt: It is an interrupt that is generated within a processor by instruction. Software interrupts are used for system implementation because they give subroutine call with a CPU.
• Spurious interrupt: It is an unwanted hardware interrupt that are typically generated by conditions of system like electrical interference through incorrectly designed hardware.

Processors have an internal interrupt mask that allows software to ignore external hardware interrupts. Setting this mask can be faster than interrupt mask register (IMR) access in a PIC. It also disables interrupts in the device itself. Interrupts disabling and enabling act as a memory barrier that may actually be slower. An interrupt which leaves the well defined machine is called a precise interrupt. These interrupt has four properties:

• The Program Counter (PC) is saved in a known place.
• The instruction pointed to execution state by the PC is known.
• No more than one pointed instruction has been executed by the PC.
• All instructions by the PC have fully executed.

Understanding RAID Technology

Data transferred on internet sometimes become changed between the receiver and sender’s time. This may be occurs for a various reasons, but modern network protocols now support "error checking and correction”. The numeric values of all the data bytes in the packet are added up, the output  is then used to know whether the received packet was correct or not. This process is known as checksum. If  receiving node detects wrong  packet means if the received checksum doesn't match the sum of received bytes , the request message is sent for the another copy of packet.

The TCP/IP Transport Layer suite has many protocols that operate at the Transport level. Transport Layer is responsible for establishing the connection. It manages connection oriented streams, reliable transport, flow control and multiplexing. Following are the two most important protocols for the transport layer.

• Transmission Control Protocol (TCP)— It handles the transmission of the data packets from the Application Layer. If a large file needs to be received, TCP manages the IP packets from the below layer. TCP arrange the packets into sequential order. It also takes smaller IP packets and merges them into the originally sent file. If a data packet is not received, TCP send an acknowledgment message for packet retransmission.
• User Datagram Protocol (UDP) — UDP is similar to TCP, but there is no guarantee of delivery in this protocol. UDP has a minimum overhead as it does not send the acknowledgments of packet receipt or order of receiving or retransmission of data. If the packets are out of order or lost, then theses packets are skipped.

What is the difference between Client-Server and Peer to Peer Network in distributed computer system?

Peer-to-peer (P2P) network is a decentralized communications model in which either party can initiate a communication session.

The client/server model is a centralized communication model in which the client makes a request and the server will fulfills their request.

The main difference between peer to peer and client-server systems is that in the client-server model, there are clients that request for services and the servers provide services to the clients, but in peer to peer systems, peers act as both consumer as well as service providers. Further, client-server systems has a central file server and its implementation is expensive than peer to peer model. While in the client-server model, a dedicated file server provides an access level to the clients, by providing better security.  But in peer to peer systems, the security is handled by end users. Moreover, peer to peer networks also suffer in performance as the nodes increases, but client-server systems are stable than P2P and could be scaled as per the need. Therefore, selecting one is dependent on the environment that of implementation.

References

1. Jacobson, Michael B., et al (1995). "Memory systems with data storage redundancy management." U.S. Patent No. 5,392,244. 21 Feb. 1995.

2. Kimmel, Jeffrey S., Sunitha S. Sankar, and Tomislav Grcanac (2009). "Lost write detection in a storage redundancy layer of a storage server."S. Patent No. 7,549,089.

3. Vachuška, J., and M. VoboÃ…â„¢il (1971). "Kinetic data computation from non-isothermal thermogravimetric curves of non-uniform heating rate." Thermochimica Acta 2.5: 379-392.

4. Yan, Zhixian, et al (2013). "Semantic trajectories: Mobility data computation and annotation." ACM Transactions on Intelligent Systems and Technology (TIST) 4.3: 49.

5 Singh, Hartej, et al (2000). "MorphoSys: an integrated reconfigurable system for data-parallel and computation-intensive applications." Computers, IEEE Transactions on 49.5: 465-481.

6. Comer, Douglas E (2000). The Internet book: everything you need to know about computer networking and how the Internet works. Prentice-Hall, Inc.

7. Perkins, Charles E (1998). "Mobile networking in the Internet." Mobile Networks and Applications4: 319-334.

8. Goleniewski, Lillian (2002). Telecommunications essentials: the complete global source for communications fundamentals, data networking and the Internet, and next-generation networks. Addison-Wesley Professional.