Hardware system optimization and IO Optimization core optimization principle detailed

650) this.width=650; "src=" Http://s2.51cto.com/wyfs02/M00/7C/88/wKiom1bSpO6zW1vJAABl9zwJkkw393.png "title=" Qq20160228153017.png "alt=" Wkiom1bspo6zw1vjaabl9zwjkkw393.png "/>

From the graph, the above is the CPU level, the following is the resource level, the more slowly down. Running speed.

From the capacity, the farther down the bigger.

Processed in memory before the hard drive.

CPU and memory processing speed ratio is also very big difference, 1000 times times the gap

So the light memory is also slow, the general CPU also has a cache, Level 1 cache, Level 2 cache. CACHE

The interactive process may be placed in Level 1 cache, level two cache, for gentle operation, CPU and memory.

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M02/7C/88/wKiom1bSpSWiyhzDAADPIdpH9NQ501.png "title=" Qq20160228154105.png "alt=" Wkiom1bspswiyhzdaadpidph9nq501.png "/>

The speed of the computer has what to decide and you then optimize what place.

For example, the bottleneck is the hard drive, my CPU becomes dual core, the hard drive is not fast, should be the bottleneck is not the CPU, is the disk.

You'd better have a balance between the various parts of the computer. A good, can not improve the overall performance of the computer.

NS: nanoseconds

US: microseconds

MS: MS

Nic and SSD almost, ordinary disk is the bottleneck, with ordinary hard disk do not use a piece, 3, 5 pieces

Do site architecture: can use Cpucache without memory, a step closer, can use memory, do not use SSD can use SSD without hard disk.

This is a server optimization that will rise to the optimization of the entire cluster in the future.

Cluster optimization is the same, the data thrown into the memory to provide services, future data will definitely priority in memory.

In some cases, data may be lost, first on the hard disk, and in memory. Just read from memory.

Almost all Web sites are viewed by many people, with fewer people to write.

650) this.width=650; "src=" Http://s2.51cto.com/wyfs02/M02/7C/87/wKioL1bSq4rD9WREAAA3wR9cz_U238.png "title=" Qq20160228160813.png "alt=" Wkiol1bsq4rd9wreaaa3wr9cz_u238.png "/>

Also want to write to the memory, write to memory once the downtime?

Downtime data is lost and downtime is not necessarily the crash.

Downtime is not able to provide services, do not say the operation of the crash, said downtime.

Data lost or your business accepts data loss, you can't accept this plan.

I can tolerate 1 seconds of data.

1 seconds in memory write 10,000 times, 1 seconds after the disk.

Many times faster than a direct write disk. The fewer times you read a drive, the more efficient it is.

To meet the application premise, swipe inside the hard drive.

650) this.width=650; "src=" Http://s2.51cto.com/wyfs02/M01/7C/87/wKioL1bSroyT-bAJAAAzKKqM18Y860.png "title=" Qq20160228162103.png "alt=" Wkiol1bsroyt-bajaaazkkqm18y860.png "/>

For most Web sites, read more, and for Web sites, the stress of reading is the first.

The general practice for a Web site is to write to the disk, read it into memory, and then read the memory when the user accesses it.

It's fast to read, and the disk can withstand it.

For data can not be lost, write pressure is not big, read the pressure, first write disk, read to memory, so provide services.

Summary: The CPU is fast, so there are L1, l2cache,cpu after memory, memory is faster, the memory is SSD SSD, then the ordinary hard disk.

Performance sequencing of server hardware.

In the future to do operational optimization site, the key to the server: try to use memory.

1, to solve the pressure to browse the Web, but also read the pressure.

Due to write pressure, write to the hard disk, read into the memory, and then provide access to the customer.

2, the solution of such applications, such as micro-bo, reading pressure is very large, writing pressure is relatively large.

Write stressed business, we generally, can consider writing to memory, and then according to the enterprise can accept the degree of data loss, select a frequency to write to

Disk. Read the same as above. This ensures that you receive greater write concurrency, and that the amount of data loss I can tolerate.

There is no absolute good plan on the Internet.

Higher requirements, and write memory, and do not want data to be lost.

The core reason for downtime is that there is no power.

1, with UPS uninterruptible power supply,

2, the motherboard installed flush battery role: When the server is not power, the use of battery power, the memory data brush to disk.

You understand some of the things in life, and you will.

This article from the "9525154" blog, reproduced please contact the author!

Hardware system optimization and IO Optimization core optimization principle detailed