Several 9 questions about the reliability of software system

On the way to work today, read an article about an architect who speaks about his refactoring system experience, and sees a noun, reliability 3 9, 4 9. What Ah, did not hear, the Internet search, the general system, like the traditional power system is also reliable 3 9, 4 9 said, can be said that this is the problem of system reliability, search several blog posts, keep a bit ~ ~

------------------------------------------------------I'm a gorgeous split-line-------------------------------------------------------

http://blog.csdn.net/alonesword/article/details/21555745

The high reliability of the software system (also known as usability, described in English as Ha,high Available) has a standard for measuring its reliability--x 9, the X is a representation of numbers. X 9 means that in the software system 1 years of use, the system can be normal use of time and total time (1 years) ratio, we have the following calculation to feel the next X 9 at different levels of reliability differences. 3 9: (1-99.9%) *365*24=8.76 hours, indicating that the software system in continuous operation for 1 years time the maximum possible business interruption time is 8.76 hours. 4 9: (1-99.99%) *365*24=0.876 hours =52.6 minutes, indicating that the software system in continuous operation for 1 years time the maximum possible business interruption time is 52.6 minutes. 5 9: (1-99.999%) *365*24*60=5.26 minutes, indicating that the software system in continuous operation for 1 years time the maximum possible business interruption time is 5.26 minutes.

Then x 9 x only represents the number of five, why no two, nor more than 6 of it. We then calculate: a 9: (1-90%) *365=36.5 days 2 9: (1-99%) *365=3.65 days 6 9: (1-99.9999%) *365*24*60*60=31 seconds

You can see that 1 9 and 2 9 represent the time of day when the business may be interrupted for 36.5 days, 3.65 days, and the reliability of this level may not be enough to use the word "reliability", while 6 9 means that the maximum business interruption time is 31 seconds within a year, then the reliability of this level is not realized. But to achieve from the 5 9 "6 9 of the reliability of the increase, the latter need to pay more than the former several times the cost, so in the enterprise everyone only talk about (a) 9.

See----------------------------------Split line every day------------------------------------------

Http://www.yunweipai.com/archives/1284.html

High availability is an important area of the IT world, but it is not necessary to have a high availability of up to 99.999% for large IT companies or organizations.

Whether a company or organization needs 99.999% reliability depends on a number of factors: software costs, whether the company can afford to pay.
"When your device handles human life, or the business breaks for a minute and loses millions of dollars, you can consider 99.99% reliability. "Robertson (Linux high-availability project developer) said.

Weigh the high available pay and get
The cost of achieving a highly available system, whether 99.99% or 99.999%, includes software costs, hardware costs, labor costs, and training expenses. Before a company intends to make a highly available system, it is best to weigh the cost of building a highly available system, and the loss of the system's business disruption, which pays more. If you don't want to spend too much money on building high-availability systems, then you need to analyze your system, when business outages occur, control the loss of business disruption, and build the ability to quickly recover.
In fact, a system that has millions of knives per minute, or 5 9 of highly available systems, can control the planned disruption of the business (or disrupt business control), such as when the business does not disrupt business during trading.
According to Emerson's statistical data on 41 data centers, "unplanned business disruption is a major factor in the loss of company performance". The report also points out:"The fault location after business interruption, analysis and resolution, also need to spend a lotof money."

If the reliability of a company's business commitment is 99.99%, it allows only 8.76 hours of business interruption per year. If the loss caused by a business interruption is 1 million per hour, then the loss for one year is 8.76 million.
However, if the company is committed to the reliability of 99.999%, then only allow business interruption of less than 1 hours per year, if the business interruption caused by the loss of 1 million per hour, then the loss of less than 1 million a year.
We can count that the reliability is increased from 99.99% to 99.999%, which can recover about 7 million of the loss. Well, it's worth it in this case.
But if you increase the reliability to 99.999% above the cost, more than 7 million, it is a choice to maintain 99.99% reliability.

Any product needs to take into account input and output ratio, then you allow the business interruption of the amount of time.

99.99% of the reliability is enough for most businesses, and for some offline businesses, 99.9% reliability is sufficient for most scenarios.

Achieve 99.99% Reliability
In some areas, 99.999% reliability is a must, such as the transportation industry. If the train control system fails, then thousands of passengers may be stranded. (We all know about the 7.23 accidents in the Celestial Kingdom.)
In this scenario, the implementation of 99.999% reliability typically uses fault-tolerant processing, which includes software fault tolerance and hardware fault tolerance. Common software fault tolerance is load balanced, that is, multiple instances provide services at the same time. When the software fault tolerance does not work, it is necessary to use hardware fault tolerance, hardware fault-tolerant half is to do hardware redundancy, such as two servers to prepare.

--------------------------------------------I'm me, not the same split line-----------------------------------------------------

Http://server.it168.com/a2013/0220/1455/000001455094.shtml