Jame 'sreading09-10

Scalability, that is, Scalability t. experiences and Lessons from cnz8wbRUsReddit system expansion: 1. automation as much as possible, 2. there is no need to build a scalable architecture at the beginning. 3. do not build an SOA architecture at the beginning. 4. the key to scalability is to solve the scalability problem before the user perceives the bottleneck. 5. consider SSD as a cheap memory, rather than an ang

Scalability, that is, the extended http://t.cn/z8wbRUs Reddit system expansion experience and lessons, 1. automation as much as possible, 2. there is no need to build a scalable architecture at the beginning. 3. do not build an SOA architecture at the beginning. 4. the key to scalability is to solve the scalability problem before the user perceives the bottleneck. 5. consider SSD as a cheap memory, rather than an ang

Scalability, that is, Scalability

Http://t.cn/z8wbRUs Reddit system expansion Experience and Lessons 1. automation as much as possible, 2. there is no need to build a scalable architecture at the beginning. 3. do not build an SOA architecture at the beginning. 4. the key to scalability is to solve the scalability problem before the user perceives the bottleneck. 5. consider SSD as a cheap memory instead of an expensive hard disk. 6. each tool has its own scenario, and suitable tools are used in appropriate places.

Http://t.cn/z8wheDy scalability problems. this article explains the concept of extensibility from the perspective of concurrent connections. For a single request that takes a short time but requires a large number of Long-connected businesses, the system based on the process/thread model (Apache, Oracle Dedicated mode) when processing a large number of requests, the scalability is poor, and the Event-driven system (Nginx/Oracle MTS/MySQL) performance will be much better.

The scalability and performance of http://t.cn/zTc4oZu are not a good choice. at this stage, the focus on scalability far exceeds the focus on performance. Many people think that it is okay to solve the problem of scalability, that is, to split key shared resources. theo refuted this statement. at present, it is not difficult for many systems to double performance. 2. by saving half of the servers, you can save a lot of hidden costs.

Http://t.cn/z8Zefs6 2x2x2 Requirements for Database Scalability, a brief introduction to Database Scalability, 2*2*2 is more interesting, the first 2: horizontal/vertical scaling, horizontal scaling is implemented through replication and splitting, and vertical scaling is implemented through Memory-based databases (In Memory DB) and fast persistence (RDMA network or NVM ).

Http://t.cn/zQ0L5q1 20 Obstacles to Scalability [By Sean Hull] to achieve system expansion (Scalability) of 20 Obstacles, the content of this buddy has been repeated many times, this time to do an integration, is relatively complete, it is also worth your reference. The only difference is that using at all costs for multiple times is irrational, but it must be carefully considered.

Http://t.cn/z8iuiqJ Scaling out and creating fault tolerant systems with MySQL replication, how to implement a Scale out fault-tolerant system through MySQL replication, the name is very scary, the main three points: 1. master/Slave topology management, automatic Master/Slave Failover (MHA), 2. server Load balancer and connection pool management for requests, that is, the main tasks (such as TDDL) performed by most database middleware)

Solutions and application cases

Http://t.cn/zY8W7V3 acceptable response time, 1. no industry standard for page response time; 2. there is a cognitive scientific conclusion on the latency of human-computer interaction. The difference between human perception and human perception is less than 1/4 seconds. 3. the 8-second response rule is too simplified. 4. the user no longer accepts the so-called average response latency, 5. the "4-second rule" may not describe the user experience. 6. the relationship between expectation and experience determines user satisfaction.

Http://t.cn/z861NpO performance reverse mode, 1. the performance problem is fixed at the end of the project. measure and Fix Error indicators. ignore the role of algorithms. 4. focus on what you can see, rather than the underlying problems. 5. too many software layers, 6. excessive threads, 7. hardware usage is not balanced, 8. invalid Data Copying between CPUs ,...

Http://t.cn/zO3hEPD Godaddy uses Cassandra as its Session Store, 1. use cluster A/cluster B to upgrade the cluster. collaborative development of ASP/ASP.net protocol access, provide connection pool, compression, configuration management and other features, 3. before Cassandra supports the TTL function, the manual secondary index is set to clean up the Expire Session.

Http://t.cn/z8Mo36p why Cassandra does not need (use) Vector clocks By Jonathan Ellis, after reading, I feel that Jonathan basically understands the problem of Vector clock, 1. performance issues (read, deserialization, serialization, and write), 2. multi-version of data, 3. you can help find the inconsistency, and you have to solve it yourself. However, I don't understand why Cassandra doesn't need it.

Http://t.cn/zQHfmu8 NUMA (Non-Uniform Memory Access): An Overview, Numa concept Overview, 1. numa is only good for memory access-intensive businesses. the difference between node local and interleave. how does Linux handle Numa memory allocation of processes, and how does CPU scheduling optimize Numa-related processes? 4. commonly used Numa tools: numactl/numastat/numa_maps

Http://t.cn/z8UpsOZ for time series data storage, TokuDB is significantly better than InnoDB, regardless of the data loading time, data compression efficiency, there are several times the advantage.

Http://t.cn/z8w7lkU LinkedIn has recently opened up its Stream Processing System in Apache. Apache address: http://t.cn/zQDVp4j

Http://t.cn/zQS7y8k oratop from MOS (post), MOS (Oracle Support) Address: oratop-utility for near real-time monitoring of databases, RAC and Single Instance (Doc ID 1500864.1)

Performance Optimization and other

Http://t.cn/zQa8Mz1 distributed consistency algorithm and Raft, Riak people introduced the basic distributed consistency, several possibilities of Data Loss, and strong synchronization, the balance between consistency and availability, finally introduced Raft algorithm, and their implementation based on the Erlang language.

Why is the queue where the http://t.cn/z8L4VyC is in the queue slower? This article explains from the psychological factors such as cognitive prejudice.

Books on http://t.cn/z8ij2nc Performance. this article collects a large number of performance-related books in various fields, and naturally contains many books I like. neil Gunther and Daniel. the Capacity Planning series written by Menasce, Connie U. performance Engineering books written by Smith, Brendan Gregg Performance Analysis Series, Martin L. A series of books on architecture design principles written by Abbott.

How the http://t.cn/zQDqCyQ improves query efficiency by 100 times by modifying a row in an SQL statement. I entered Oracle Database in the early days because of similar cases. The first execution of a poor SQL statement took 13 hours. After optimization, it took about 23 seconds to run.

# Oracle tips # How to effectively kill a session, alter system disconnect session 'sid, serial # '(immediate | POST_TRANSACTION ). find a time to verify that due to the problem of kill session, we still have a big fault. The session kill, but the process has not been killed, resulting in an error in max process.

Http://t.cn/zQl4dM4 The top 5 proactive measures to minimize MySQL downtime 1. maintain verified backups. build redundant components to cope with spof. all changes must be verified and have a rollback plan. 4. verify the validity of the copy/copy. do a good job of monitoring and trend analysis. The several monitoring items mentioned are more practical.

Domas mituzas, MySQL engineer at http://t.cn/zQYU2X1 Facebook, explores Durablity. In general, Durability is about the art of trade-offs. in MySQL, the cost is quite high to achieve persistence. Three FSync operations can ensure the Durability of a transaction. The Reliability Design of MySQL Master/Slave is not secure enough (Crash-Sage), and 5.6 has been greatly improved.

Http://t.cn/zQa8KMi OF The Antifragile Organization, Netflix's Ariel Tseitlin describes their high fault tolerance system design at Netflix, their attitudes to Failure, and how to deal with Failure, how to Design and avoid the problems caused by Failure. the following describes how to design Fault Tolerance at the system level from various possible Failure dimensions.
The concept of anti-vulnerability is why the new book "anti-vulnerability" of Taleb. Netflix introduced their Simian Army series's monkey series (Chaos Monkey, Latency Monkey .....), The basic design concept is as follows: 1. design applications with fault tolerance capability through redundancy; 2. Reduce uncertainty by constructing faults.

Several key questions about fault Review: 1. What is the fault? 2. How can I quickly find this fault? 3. How to avoid this fault from happening again? 4. How can we avoid this type of fault from happening again? 5. How can we improve the efficiency of troubleshooting such faults next time? By Jeremy Edberg, Reliability effecect, Netflix

Social Science Library Reading Abstract

In a market-free society, competition is also endless, but the forms of competition are different. competition is the form of competition, power struggle is the form of competition. the reason is clear: competition must exist if more than one person needs the same economic goods. from Zhang Wuchang's economic explanation: Demand in science, Chapter 2 ends.

Although retaliation can cause a lot of harm (people who have been resentful or have had a hard divorce should understand what I am talking about ), I would like to say that the threat of revenge-even if people have to pay a huge price-can become an effective force mechanism to maintain and support social order. I do not advocate "eye-to-eye and tooth-to-tooth", but I guess that the threat of revenge is always effective. from Dan. iResearch chapter 5th of "weird behavior.

The most skeptical review of all the hypotheses present before us, while maximizing openness to new ideas. If you... Without a trace of doubt, it is impossible to distinguish useful ideas from those without value. if all ideas are equally effective, then you lose yourself, because that .. there is no effective idea at all. this article is excerpted from 54 pp. "misunderstanding: six mistakes often made in thinking.

1. our efforts to make a thing not only bring changes to it, but also change our comments on it. 2. the more you pay, the more you love. 3. we have a high valuation of our work. This prejudice goes deep into the bone marrow and mistakenly thinks that others share the same view with us. 4. if great efforts are still unsuccessful, we will not feel too much attachment. from Dan. herry's weird behavior

If you set up some obstacles on the road for your favorite person to pursue you, they will cherish you more. on the other hand, if you keep rejecting them from forcing them into desperate circumstances, you should not expect to say "we are friends ". from Dan. herry's weird behavior.

Efforts are made to increase the technical content of "Made in China", which is currently the main direction of our public policy. Without severely compressing the proportion of rent-seeking in the rent value, I think this transformation is impossible. In this sense, political system reform is far more urgent than economic system reform. In other words, without changing the way political operations are conducted, any significant and effective economic system reform is no longer possible. Ding ding, lecture on behavioral economics, page 97

"The institutional feature of a great society is that he encourages all individuals to explore in all possible directions, because we do not know the potential disaster of the future, the exploration in which direction can save us all; the hero who saves us must be anonymous because of the uncertainty of nature and our living environment-so we can only encourage exploration in all directions. "-Wang Dingding, lecture on behavioral economics, page 93

If you have common sense of life, you will continue to return to your common sense when you study the theory. You will find problems-either the problems of the theory itself or the expressions of common sense. The life experience is true, and the tree of life is evergreen. Therefore, knowledge is always a process of knowledge, not a pile of static concepts. The process of knowledge is entangled with your life experience. The experience is rich and profound, make your knowledge rich and profound.

Henri Bergson believes that the concept is used to cover the life experience. If you use a concept without a life experience, it is pale to you and meaningless, you have never been touched by any part of it. from Wang Dingding's lecture on behavioral economics-evolutionary perspective

Social development depends on individual creativity, Because creation is an individual thing. the creative spirit of an individual requires the tolerance of the society. The tolerance is an integral nature. Without tolerance, the individual freedom will disappear. therefore, freedom is the whole, and creation is the individual. excerpt from: Wang Dingding's lecture on behavioral economics, page 102nd.

Although gratitude and gratitude are all emotions produced by people who have received some help, their nature is not the same. After a person accepts help or gifts, he feels that being in debt is a kind of feeling. The Helper or donor also hopes that he will keep reminding him of this feeling, he has repeatedly asked him to be "grateful" or "grateful". Otherwise, he would be "grateful", "unconscientious", or "forgot ". By Xu Yu

The biggest feature of http://t.cn/z8iFq1R's authoritarian fear technology is the use of uncertainty. Only such uncertainty can lead to the expansion of the fear effect. At the same time, only uncertainty can form your inner self-discipline and self-review. The condition that uncertainty works is the general and arbitrary nature of power. Arbitrary can cause uncertainty, while general can expand the arbitrary space.

Related posts:

1. Jame's Reading 10-14
2. Jame's Reading 07-25
3. Jame's Reading 06-28

Original article address: Jame's Reading 09-10. Thank you for sharing it with me.