Recently read a book, called “enterprise IT architecture transformation way: alibaba China strategic thought and the architecture of actual combat, in writing this article haven’t finish reading the book, for fear that if after to read all the books to write this article, a lot of wonderful content may have forgotten, so the way to write an article to share with everyone.
Alibaba in Taiwan strategic thought and architecture combat tells the history of Alibaba’s technology development, but also an Internet technology architecture practice and development history.
directory
Part ONE introduction
Chapter 1 Reflections on The Strategy of Alibaba Group in Taiwan
1.1 History of Alibaba Sharing Business Division
1.2 The crux of enterprise information center development
Chapter 2 constructs the base sharing service system of business center
2.1 Return to the essence of SOA: service reuse
2.2 Services need continuous business nourishment
2.3 Shared service system is the soil to foster business innovation
2.4 Empower the business with rapid innovation and trial-and-error capabilities
2.5 Reserve for the power of big data
2.6 Changing organizational formation will improve organizational effectiveness
The second part is the construction of shared service system
Chapter 3 the choice of distributed service framework
3.1 The “servitization” process of Taobao platform
3.2 Comparison between “centralized” and “decentralized” service frameworks
3.3 Alibaba distributed Service framework HSF
3.4 About micro services
Chapter 4 Construction principle of shared Service center
4.1 General picture of Taobao’s sharing service Center
4.2 What is a Service Center
4.3 Division principles of service centers
Chapter 5 Data split to achieve linear expansion of database capabilities
5.1 Database bottlenecks Hinder sustainable development of services
5.2 Practice of database subdivision and table
Chapter 6 principles of asynchrony and caching
6.1 Service Process Asynchronization
6.2 Database Transaction asynchronization
6.3 Transactions and flexible Transactions
6.4 SEC activity promotes high use of caching technology
Chapter 7 Build digital operation capability
7.1 Problems caused by Service servitization
7.2 Hawk-Eye Platform architecture
7.3 Burying Points and Outputting Logs
7.4 Distributed processing platform for massive logs
7.5 Log Collection Control
7.6 Typical Service Scenario
Chapter 8 Building platform stability capability
8.1 Traffic Limiting and Degradation
8.2 Traffic Scheduling
8.3 Service Switch
8.4 Capacity measurement and evaluation planning
8.5 Full-link pressure measurement platform
8.6 Business-induced platform
Chapter 9 Sharing Service Center internal and external cooperation sharing
9.1 Problems caused by the savage development of servitization construction
9.2 Construction idea of shared service platform
9.3 Collaboration between Shared Service Platform and Service Providers
9.4 Cooperation between the SERVICE CENTER and front-end Applications
9.5 Performance appraisal of business in Taiwan
9.6 Capacity opening is the foundation of ecological construction
The third part is alibaba’s ability output and cases
Chapter 10 Internet transformation of large central enterprises
10.1 Project Background
10.2 Project Implementation
10.3 Customer Revenue
10.4 Author’s Impressions
10.5 Project Postscript
Chapter 11 Internet transformation of fashion industry brand companies
11.1 Project Background
11.2 Transformation of supply chain
11.3 All-channel integrated marketing based on SCRM
Due to copywriting restrictions, Alibaba zhongtai actual combat books can not be fully displayed, friends who want this book forward + follow the public account below, I can send you a complete book PDF document.
China strategy
Alibaba set up its Taobao Affairs department in 2003, as shown in picture 1.
In 2008, when B2C business was booming, Alibaba set up Tmall, which was initially called Taobao Mall and operated as a department of Taobao Business Division at that time, as shown in Figure 2.
With the increasing number of B2C business, Tmall began to independent, alibaba established Tmall division alone, with taobao’s affairs, as shown in figure 3, this time taobao technology part and at the same time support the two divisions, the organizational structure determines the technical team would be preferred to meet business requirements from taobao, serious impact on the development of Tmall business. Those who have used Tmall and Taobao can see that e-commerce platforms like Tmall and Taobao both contain the functions of commodities, trading, evaluation, payment, logistics and so on.
In 2009, the Business Division of Sharing emerged at the historic moment. Major members came to The Technical team of Taobao, and in terms of organizational structure, it became a business division of the same level as Taobao and Tmall, as shown in Figure 4. In this way, the Group hopes that the technical team can support both Tmall and Taobao businesses, while depositing public and general businesses and making more reasonable use of resources.
But at the time actually Shared business division is “obey” Tmall and taobao, Shared business department to meet the needs of the Tmall and taobao, team members often work overtime may reach Tmall and taobao’s demand, thus causing Tmall and taobao’s business units to not satisfied with the Shared business division, At the same time, colleagues in the sharing business division can only suffer untold sufferings.
Group-buying business deals appeared in 2010, the flow of bargain, has a strong ability to attract, so Tmall, taobao, 1688 want to bargain, docking platform to expand their traffic, bargain, suddenly faced with such a big butt is inundated with demand, at this time if the group requires three big electric business platform to bargain, docking platform, must by sharing business unit! Just because of this policy, sharing business Division has a strong business grip, which has pulled the imbalance of discourse power with the three e-commerce platforms to a relatively fair level. Thus, the sharing business Division that you see today has become the core business platform of Alibaba Group, as shown in the figure below:
The figure above clearly describes alibaba’s “thick platform, thin application” architecture, and sharing business Division is the true embodiment of “thick platform”, “thick platform” provides the most professional and stable business services for alibaba’s various front-end businesses, which is China Platform. We can find that the Zhongtai strategy is not accomplished overnight. When the sharing business Division was established in 2009, it has laid a certain foundation for the Zhongtai strategy, but it also needs the strong support of the group to build the Zhongtai strategy. Once zhongtai takes shape, it will lay a solid foundation for the takeoff of the business.
Chimney structure
In 2008, the technical team of Taobao supported both Taobao and Tmall, and 1688 had its own technical team, with the structure as shown below:
This architecture is a chimney-like architecture, in which each business department and its corresponding business department stand like a chimney, and according to this architecture, when the enterprise needs to expand new business, there will be a new business department and corresponding new technology department, that is, an extra chimney. So far, there are still many enterprises like this architecture, and this architecture must have its advantages:
- Enterprises consider different business models, so independent construction
- The new business team decided that there would be too much technical and business historical baggage to rebuild from the previous business
It’s just that the disadvantages of this architecture far outweigh its advantages:
- Repetitive function construction and maintenance leads to repetitive work and investment. Repeated construction can reduce risks for enterprises, but will increase the cost of repeated.
- The cost of collaboration between “smokestack” systems is high if they interact.
- It is not conducive to business precipitation and sustainable development. After a chimney goes online, it enters the operation and maintenance stage. At this time, it will take a long time to modify the business on this basis and release the business.
In the Internet era, it is inevitable to better integrate the internal resources of enterprises, reduce the cost of enterprises, and realize the interaction between various systems. In 2004, SOA concepts were proposed to solve the problem of “smokestack” system interactions.
SOA
Core SOA function points:
- Service-oriented distributed computing
- Loose coupling between services
- Support service encapsulation
- Service registration and automatic discovery
- Define service interaction in terms of service contracts
Centralized SOA
Many enterprises implement SOA through an ESB, which is a centralized SOA.
ESB is an enterprise service bus. As its name implies, THE ESB system can manage various services in an enterprise in a unified manner. The ARCHITECTURE of ESB shields the impact of service interface changes on service consumers and is a good architecture for realizing interconnection between different systems, as shown in the following figure:
In 2004, many large software companies have found that more and more enterprises in the process of construction of IT for many years, gradually built more and more of IT systems, these IT systems are adopted chimney type model of system construction, make the business enterprise inside the system with numerous and complicated, these systems have a plenty of buy business suite, have a plenty of independent research and development, Some of them are developed by outsourcing companies, and the final result is that the technology platform, framework and language used by each system are different. So software companies have developed ESB systems to help these enterprises solve these problems.
The service provider needs only to define the interface and the access path of the interface on the ESB system, it doesn’t care who consumes the service, and changes to the service require only one adjustment in the ESB, thus isolating the impact of changes to the service interface. ESB reduces the coupling between systems, facilitates and efficiently realizes system integration, and provides more professional capabilities in terms of service load balancing and service management and control than point-to-point mode.
ESB provides functions such as adaptive access to various technical interfaces (HTTP, Socket, JMS, JDBC, etc.), data format conversion, data tailoring, and service request routing. The purpose is to enable enterprise customers to improve development efficiency and implement projects faster based on these functions.
Therefore, the ESB approach has become the mainstream of SOA implementation in this era, which well solves the interaction between heterogeneous systems.
Decentralized SOA
“Decentralized SOA” is brought by the Internet industry, because in the Internet industry, the user group is the whole Internet public, so the system architecture designers should first solve the problem of system scalability, in order to carry out faster business response, better support business innovation, etc.
So “decentralization” not only meets the core function points of SOA, but also avoids the difficult scalability problems and potential “avalanche” impacts of “decentralization”.
“Decentralized SOA” is a “point-to-point” architecture that has no center, as shown below:
It may be argued that SOA was created to solve the problems of smokestack architecture, and that calls between smokestack systems are “point-to-point”, so isn’t that a step backwards? In the Internet industry, decentralized service framework is to run in the enterprise internal, rarely appears across the service interaction between inner and outer net, also took the form of contract ahead of the service interface functions, to some extent is good enough to safeguard the stability of the service interface, and decentralized services framework with support for many versions, load balancing, and other functions, Essentially masking the various system instability problems of the previous “point-to-point” mode.
In a centralized architecture, the ESB is at the center of the entire architecture, and all service invocations and returns go through the ESB, so the service caller has a lot more network overhead when calling a service than in a decentralized architecture.
In addition, all service invocations go through the ESB, so it is necessary for the ESB to be clustered, and the server configuration or network configuration for deploying the ESB system is better to ensure that the ESB is not a problem, which can lead to significant increases in software and hardware costs if the enterprise wants to scale up the ESB.
In addition, even if ESB systems are clustered to ensure high availability, there can be an “avalanche” effect that causes all services in the enterprise to become unavailable.
An avalanche
We assume that the ESB for each server in the cluster’s biggest concurrency value was $100, assuming that now has 10 server in the cluster, smooth in daily amount of user requests, each server after load balancing the concurrency value of an average of 80, but if a server in the cluster suddenly fails, another 9 units is required to take at this time before the concurrency value, The amount of concurrency on the remaining nine servers increases, and it is quite possible that one of the nine servers will be overwhelmed, which will cause the remaining eight servers to be overwhelmed in succession, an “avalanche”. Once the “avalanche” failure occurs, even if you go to restart the server is very difficult to solve, because it is likely that the server just started the completion of the traffic was overwhelmed, so this time you can only prohibit the outside traffic into your system, and so on all servers are successfully started after the traffic in. And when that happens, you may not even have time to locate the problem, and the restarted cluster is actually still in a “fragile” state.
This means that a “centralized” architecture is not a good solution to the scalability problem, but a “decentralized” architecture is better, because even if the above situation occurs, it does not affect all services. So this is why the Internet industry chooses a “decentralized” architecture.
Here we introduce alibaba distributed service framework HSF, and I will continue after reading it… Ha ha.
There is a pain point innovation, a technology is certainly to solve a pain point just appeared.
Due to copywriting restrictions, Alibaba in Taiwan actual combat books can not be fully displayed, want to this book friends forward + follow the following public account can get a complete book PDF document for free.
Part of the algorithm screenshots, directly on the directory (internal PPT and implementation manual)
Large set of machine algorithms
Due to copywriting restrictions, Alibaba in Taiwan actual combat books can not be fully displayed, want to this book friends forward + pay attention to get a complete book PDF document, as well as the article Ali internal implementation manual and PPT
