Some experiences (Continuous updates) and experiences are continuously updated

Some experiences (Continuous updates) and experiences are continuously updated

Database Query:

Database queries have a great impact on both concurrent processing and transaction processing;

The better the query performance, the less impact the performance will be caused by the increase in concurrency;

The better the query performance, the less affected by the exclusive lock of other transactions;

In short, it is the most fundamental way to solve timeout and deadlock issues, optimize query performance, and shorten the transaction time (it is best not to use transactions ).

 

(Non-original, reference from the network) use two examples to see the differences between distributed and cluster:

Example 1:

If a task is composed of 10 subtasks, it takes one hour for each subtask to be executed separately, it takes 10 hours for the task to be executed on a server. The Distributed Solution Provides 10 servers. Each server is only responsible for processing one subtask, regardless of the dependency between subtasks. It takes only one hour to complete this task. (A typical example of this working mode is Hadoop's Map/Reduce distributed computing model.) The cluster solution also provides 10 servers, and each server can process this task independently. Assume that 10 tasks arrive at the same time, and 10 servers will work at the same time. After 10 hours, 10 tasks will be completed at the same time. In this way, you can complete a task within one hour!

Example 2:

If a bunch of machines are piled up here and have the same functional attributes, it can be called cluster distribution. It should be a type of system with the distributed characteristics.

Cluster role: Server Load balancer, error recovery

Cluster-based resource integration. Distributed Computing Capability Integration

Cluster = aggregation, distributed = combination

 

Reduce thread switching:

Almost all IO operations (including disks and networks) generate thread switching Mutex, AutoResetEvent, Sleep, and other methods, which will generate thread switching in the era of application servitization, network service interfaces are often used as local methods. When such interfaces are put into loops and recursion, the time consumed by thread switching often exceeds the logic to be executed + the overhead of network transmission. How can this problem be avoided: All IO operations are in asynchronous mode (this is difficult) use special caution to encapsulate IO interfaces for interfaces that will generate IO, and provide batch and single interfaces as appropriate when introducing cache design to reduce IO times ...... Reduce nested loops: optimize nesting by means of hash tables, dictionaries, and groups, and change multi-layer loops into multiple loops (complexity changes from high power to one party) to prevent brute force cracking: MD5 (x + MD5 (y); y. length> = 8
My experience in using dabao is continuously updated. Welcome to refer

Yulan oil Yafang is far less cost-effective than dabao !! A lot of big treasures are very gentle to use. I have used two Yulan oil and Yafang products very exciting !!

Persistent feelings

As long as you stick to it, it will succeed in the modern society, and it is no longer that great, but it cannot be done without effort, you need to make yourself feel that you have no regrets in the future, and you do not have to force everything!