“Can you tell me the story of SOFA?”
“The story of SOFA is very special and very common. Tell you the story of SOFA. Or maybe it will be your story.”

The idea: Technology is supposed to solve real problems

Ant Financial does not make technology for its own sake, but hopes to use technology to solve the current and future problems of society. If the social value of digital finance is depicted in a pyramid structure, at the top of the tower is that digital finance can bring more equal opportunities across the globe.

“For Ant Financial or Alibaba, first of all, we are very idealistic and vision driven, which must guide us when it comes to bringing more equal opportunities to the world,” said Li Cheng, ANT Financial’s CTO. However, we are also a very realistic company. When confronted with specific problems, we will look at how to bypass the current obstacles and go to the future. He does not take a rigid approach to specific practical problems. Specific issues will have to be analyzed on a case-by-case basis, but our vision will not change, nor will we turn so-called values into dogma. Sustainability in business is very important to us. If we are not sustainable in business, we are not going to have a future.”

Technological innovation is forced

The same is true of Ant Financial’s own SOFA. SOFA takes a different path of distributed architecture from the traditional financial industry. To achieve finance-level performance and reliability based on unreliable hardware systems, there are many difficult problems to be solved in dealing with super-scale Internet financial applications such as Alipay. Ant gold suit to build a set of processing level financial transactions of a distributed architecture and platform, the financial level of consistency and mass concurrent processing ability achieves a good balance, and the quick disaster recover, elastic scaling ability, ability to ensure high availability are much more live and on-demand supply fine resource scheduling ability precipitation has abundant practical experience.
Middleware, as one of the traditional basic software troika parallel with operating system and database, is also a highly difficult software engineering. The concept of traditional middleware was born in a previous era of “distributed” computing, server/client computing in small-scale lans, the “middle tier” software above the operating system and below the application software. The early emergence of middleware is to solve the challenge of supporting application software in heterogeneous hardware environments such as PC server, network and even computer rooms in different geographical locations. Unlike operating systems and databases, middleware does not have a clear definition, generally including messages, data, remote procedure calls, object request brokers, transactions, components and so on.

With the rapid development of the Internet, especially the vigorous development of cloud computing in the past decade, the IT environment of enterprises has undergone profound changes: From the past distributed computing environment based on LOCAL area network and metropolitan area network and single urban geographical scope (traditional enterprises), to the Internet and fiber network, national and even global geographical scope of the super-scale distributed computing environment evolution (Internet enterprises). In this process, software also evolves to large-scale Internet services and cloud services, both operating system and database have undergone profound changes, and middleware is also evolving and expanding its boundaries in this process.

The development of middleware represents the upgrade and change of technology architecture, which is closely related to enterprise organization model and business practice. In theory, middleware shields heterogeneous hardware, software, network and other computing resources downward and provides unified computing environment and management for the whole life cycle of application development, operation, and maintenance upward. It is an intermediate connection layer, which is of great value to enterprises. According to Conway’s law, the design of software and system architecture is closely related to the organizational structure, business process and communication mode of an enterprise. Therefore, with the super-large scale and rapid iterative development of enterprise business scale, the quality and capability of middleware directly determines the fate of enterprise technical architecture. Especially with the rise of digital business, middleware, which could not be perceived by business and could not bring direct value to end users in the past, has become a part of digital business.

Ant gold suit is a bring equal financial services to the world of science and technology enterprises, as a representative of the native digital enterprise and digital business, the ant gold obedience was established in 2004, pay treasure, in the past ten years out of a itself, for large scale application of the Internet financial system, financial level middleware technology. Double tenth one since 2008, in particular, the double tenth annual shearing section.at mass flow rate on the impact of the ant gold suit continuously break through the limits of existing technology, in the financial sector reached unprecedented technological achievement, especially for ten years since the research of middleware technology can meet the 2017 double tenth of a 256000 t/s peak, throughout the 1.48 billion’s payment, That compares with a peak of 20,000 transactions per minute and 12.8 million for the entire day in 2010. In the past few years, ant Financial’s self-developed middleware technology has increased the number of peak payments by 750 times, the number of daily payments by 115 times, and transactions by 225 countries and regions around the world.

Extreme business scenarios give rise to extreme IT systems. Zhao Zunkui (name: Miaocai), head of ant Financial’s financial Core Technology Department, said that he often receives heads of external financial institutions to visit and learn about Ant Financial’s IT system, and “they can’t imagine what they see”. Today, Ant Financial’s software engineering achievements have turned the Singles’ Day extreme challenge into the new normal, and this Architecture is called SOFA (Scalable Open Financial Architecture) to support ant’s distributed practice.

The birth of the SOFA

The key words of SOFA include: unlimited scalability, consistency, second-level disaster recovery and extreme low cost, thus defining a new “finance-level distributed architecture”.

That the origins of SOFA

Cheng Li, alias Lu Su, Capricorn, work number 3896. In 2004, Alipay just had its own independent system, and the basic platform had to rely on outsourcing teams to provide technical support. In February 2004, Cheng li was still studying for a PhD at Shanghai Jiao Tong University. By chance, he came into contact with Alibaba and assisted in the construction of alipay website as an outsourced architect. After one year of cooperation, Cheng decided to give up his PhD degree and officially joined Alipay in February 2005. Cheng Liyi is rigorous, pragmatic and logical, which is regarded by colleagues of ant technology team“Being like a god.”



Li Cheng, CTO of Ant Financial

As the chief architect of Alipay, the first generation architect of Alipay, and the firefighting leader of the biggest crisis in alipay’s history — the shutdown and release of 17 hours on January 1, 2008, it can be said that without Cheng Li, there would be no alipay now. The most thrilling 17 hours in The history of Alipay are recorded in Ant Financial’s introduction manual, Mindfulness: Announced New Year’s day 2008, pay treasure to 8 hours of downtime released three periods of “financial”, but frequent accidents, the “financial carry with it”, “wet dishcloth lead to server outage” rumours abound, no package to express little brother post pay treasure to hurry back, genuflect is begged Cheng Li knocked nearly two hours at a critical point of the code, The release ended a 17-hour outage.

Cheng Li tells the story of the history of the birth of SOFA: the earliest alipay system, is also unlikely to large system development personnel, like Cheng Ligang out from school do pay treasure to the first generation of the architecture, so the first generation system is very simple – is a simple application, in an application server, use a database, a big taobao customer service. A simple system supported the development of Alipay from 2004 to early 2006. Pay treasure early system structure is simple, but the benefits are particularly fast, the product manager wants how change, the implementation code can change in a minute, for example alipay, a red envelope, from demand for four days to live, but to the back, such a simple system cannot support more volume, also cannot support more complex business.

The development began at the end of 2006, when one of alipay’s biggest problems was that the business was becoming more complex and the number of engineers was increasing. The original system was called monolithic — which means a monolithic system. The system became unable to carry more and more complex business logic, nor could it allow so many engineers to work together in parallel. At the time, Alipay hoped that hundreds of projects could run in parallel, and that each engineer could work undisturbed, without the complexity of the system increasing exponentially as business logic increased.

Therefore, in 2006, the technical team of Alipay had to make a choice on the future technical structure. At that time, there were two schools of opinion: one was to learn from the bank’s big Brother, who had been going for more than ten years, so this road must be safe; The other was to take a new route, one that no one had yet taken, by using distributed architectures to support future payment systems. The distributed architecture here is not the small-scale enterprise-oriented distributed architecture in the client/server era, but the super-large-scale distributed architecture in the Internet era. After about a year of discussion and thinking, the Alipay team made a decision to take a road that no one had ever taken before, so they started the construction of the second-generation architecture of Alipay, namely, the servitization of alipay technology system. Since 2007, Alipay has started to transform its trading system, merchant system, membership system and payment and clearing system.

In that year, Alipay went to Dalian to recruit Hu Xi (aka Xi), who had worked on SOA and OSGi technologies at his previous company, and joined the Alipay team to help Cheng with the next generation architecture transformation. Hu recalled that when he joined the Alipay team in 2007, the developers were miserable. At the time, Alipay was using WebX, Alibaba’s unified technology framework. WebX is a set of general Web framework based on JavaServlet API developed by Alibaba. It is widely used in Alibaba Group and opened to the society at the end of 2010. WebX is more inclined to the architecture of front and back end integration, which can quickly build a website, but it does not take into account the complexity of business development to a certain extent, how to better build the background. For example, when paying treasure of a digital wallet system call iWallet, system startup is five or six minutes every time, developers take a cigarette, and come back if you find errors was modified to restart, the developer is not every day in the process of code compilation, is the process of restart, a system contains dozens of projects, a dozen team parallel development, Project concurrency also leads to a lot of merge conflicts and time consuming, and the overall r&d efficiency is low, making it difficult to proceed. Therefore, from that beginning to study to solve the change of the whole structure of Alipay.



Hu Xi, Vice President and Deputy CTO of Ant Financial

Cheng Li gave this distributed architecture a name “SOFA”, which has two meanings: First, according to the technical trend at that time, to do service-oriented Architecture, namely ServiceOriented Architecture, but added financial business, so it is ServiceOriented Fabric Architecture; The second is to be like a sofa, so that engineers can work very cool. At that time, SOFA was created out of such a simple consideration. The so-called SOA and servitization transformation is to reorganize the enterprise’S IT systems in the way of “services”, and then connect them through the “service bus” to form a pluggable enterprise IT architecture, which is SOA. IT is important to note that SOA is really a set of architectural ideas for traditional enterprise IT, and in the early days of SOA there were only theoretical frameworks rather than concrete successful practices.

The first generation of SOFA actually solved two problems: first, how to have a “glue” mechanism, or connector, that can connect the distributed system into a whole when the system is to be distributed; 2 it is to hope that every service itself is modular, so at that time in the first generation of SOFA USES OSGi (a Java modularity specification, allows the application to use refined, reusable, and build collaborative components), so that each engineer can focus on their respective components, then will be able to make these components are assembled together to become a “service”, The “services” are then pieced together into a grand system. This is the first generation SOFA framework.

With the first generation of SOFA technology architecture, the Alipay team began to do a very critical business service transformation. The first is to turn the core accounting system for all alipay users into a business service that can be integrated with other businesses. But once the accounting is unbundled, there is a more difficult problem: how to solve the problem of distributed service consistency, also known as distributed transactions. When solving this problem, the Alipay team took a big risk. At the beginning of the project, it was not clear how best to solve the problem. At that time, the industry technology trend could be referred to as SOA and several SOA frameworks proposed by the industry.

The SOA industry at that time proposed two standards for SOA transactions: one based on Atomic Transaction, called WS-Atomic Transaction; Another is a transaction based on business process choreography called WS-BusinessActivity; The open source community implements transactions under these two reference standards through JBoss’s TransactionServer. At that time, the technical team of Alipay wondered whether the JBoss open source technology and these two standards could be used to build the core transactions and accounting of Alipay. However, not long after the project started, about three months into the project, it was discovered that the two standard and open source implementations of the industry at the time could not possibly support a practical application.
The core of the reason is very simple, one of the most simple trading system and accounting system is the core, for the most simple of a transaction, also should pass a dozen times messaging, including any messaging if interrupted, then the transaction failed, and failed, the SOA standards in the industry at that time did not put forward how to restore the failed transactions, At the same time, if any transaction is passed more than ten times, the whole performance is very low. If such a system is finally released, in fact, it cannot support the transaction volume of Alipay at that time. So halfway through the project, the team abandoned the industry standard and its open source implementation and had to find its own way.

In those days, another approach to the consistency of distributed transactions was based on the two-phase transaction standard (Prepare and Commit) and the open source distributed implementation XA, but it was already known that PayPal had gone down this path, resulting in a system outage for a week and a complete system rollback.

So at that time, alipay’s technical team considered whether they could come up with their own standard, which would be simple: First is to solve the problem of distributed consistency, must want to commit protocol of distributed, implement the agreement if the database layer, efficiency is very low, because the database layer don’t understand any of the business logic, only to realize in a general way, leading to unable to optimize the upper level of the business logic layer, so think of to commit protocol in the service layer.
“At that time, our big idea was very simple. Since the Alipay system had been divided into very small scale services, we should let the service itself have the property of transaction, called transactional service. Such small transactional services are like small stones that can fit into a large cup, and then design a distributed commit protocol to bind these small transactions into a large business transaction. The service is not just a microservice, it’s a microtransaction, turning each service into a transaction, and then turning each transaction into a big whole service through a choreographed framework.” Cheng li used graphic language to explain the origin of eXtended Transaction Service (XTS), now known as Ant Financial Hack technology.
Had this idea, pay treasure technology team began to do at that time. The first difficulty to overcome was to transform existing transaction services and accounting services into transactional services. This difficulty was broken through at that time. The second challenge was to achieve a Scalable framework to orchestrate large volumes of transactions: XTS. SOFA project was put online in January 2008. Since its launch, it has been continuously polished and has been supporting the entire business transactions of Ant Financial until now.

Evolution of SOFA

From the first generation to the current fifth generation, the evolution process of SOFA is actually alipay from the earliest single system of a large business and IT intertwined together, while dismantling the financial business system (later business center), while dismantling the bottom IT system (later data center, computing center), In the process of splitting and solve new problems such as extensibility, consistency problems of, and continuously meet every year can limit the double tenth of a breakdown system, but also the data from the original system “can” get bit by bit into the new system, the management of new huge amounts of data at the same time, such an extremely complex process is how to? Interestingly, one of the added values of this process is that it is unintentionally done to “IOE”, since the separation from monolithic systems to Internet distributed systems is itself a process of replacing expensive IOE equipment with PC server rooms.
“It was kind of a wild ride.” Yang bing later recalled the whole process. “There are only so many ‘radishes’ and there are so many pits that it is impossible to fill in the state. Every middleware product can’t even do ‘one radish, one pit’. Many radishes are placed in two or three pits, which makes you think how challenging it is. Second, business targets double every year. State of the whole team is basically a year after the large presses, after the Spring Festival begins to dense prepares to promote the next year, blink of an eye the double tenth away a few months, many systems of technical innovation is likely to be ready in June and July, all upgrade up again, the business is changing, constantly had new ideas emerge, every year is such a state, Basically is the development of aircraft to upgrade the engine up.”



Yang Bing, middleware team leader of Ant Financial

Cheng Li memories, SOFA early development is completely against the project management logic, in the process of project to promote both research and development platform and research and development of the upper business system, is equivalent to a lot of risk guide do together in a project, the SOFA is the first generation of project by team work together, will face new problems every day, every day to solve various problems, But there was a sense of triumphalism behind it and a willingness to embrace change, to overturn all previous architectural decisions in the middle and later stages of the project and replace them with a new one. “So to the accounting of the third phase of release in 2008, the original release 8 hours, then we issued 17 hours, shows that in the process of project launch, still have a lot of problems not solve, but in the end we abruptly to release on this project, and succeeded, in retrospect, is actually a bit of a terror.” Cheng li said with a smile.

After 2008, alipay’s technical team began to define a principle that all releases should not be stopped, and project releases must be risk-free. Secondly, to end all research projects, technical research to solve the technical problems, and then apply to the commercial system. And since 2008, every new technology will not be used first in the core system, but will be fully tested in the relatively marginal business system before being used in the core system.

In the early stage of SOFA, we can see that in the two projects of trading and accounting, business system developers and technical platform developers were not separated. No matter Cheng Li or Hu Xi, they wrote the code of business transaction and the code of technical platform, and the engineer team did not strictly distinguish. At that time, Yang Bing joined the middleware team. Some people were assigned to study the underlying technology, while others were assigned to write the application system architecture. Gradually, the team became more and more formal, including the gray level and control of the on-line process of new technology.

Yang Bing recalled that he joined the Alipay team in 2009 as a graduate student, when the process of service separation is Cheng Li, Hu Xi personally involved, while the WebX system to do service separation, hu Xi wrote the prototype of SOFA framework, Yang Bing and the back to join the small partner to help constantly improve SOFA. “At that time, we were very young, and basically Cheng Li and Hu Xi took us to implement the whole set of ideas that they had conceived. While the concepts of servitization and SOA were hot at the time, the industry practice was not nearly as rich as it is today, and many implementation mechanisms were just feeling their way through the mud. At that time, Cheng Li, Hu Xi and Ni Junjun (alias Miao Renfeng, chief architect of The first generation of Alipay and president of Ant Financial Alipay Business Group) were mainly involved in the separation of business codes. These people both wrote framework codes and component codes and participated in the separation of business codes. At that time, all the businesses of Alipay were evolving. On the one hand, the Alipay architecture team got familiar with Alipay’s business in the process of business logic separation and technical architecture separation, and on the other hand, they considered how to abstract reusable codes, data and frameworks from the familiar business to better support future businesses. So it was a case of taking apart the business and the technology and moving the parts into a new framework. The whole process is not about designing and working on it, but working on it as we go along.”

XTS frameworks are written in that process. Because transaction consistency was not a problem in the original centralized architecture, it is a problem in the split. When a problem arises, solve it as you go along. Of course, the solution is not human intervention, but the idea of a technical scheme, or even precipitation out of a set of technology. At that time, The Oracle database in Alipay system was still in use, and high-end traditional equipment such as minicomputer was still in use. After alipay’s business system including membership and transaction was separated, it directly ran on the X86 architecture, which was not only the difference in physical form and deployment form. It is the evolution process from WebX to SOFA from the development mode of single application to the distributed development mode of SOA. The accounting system was the last system to be separated from Alipay. With the successful separation of the accounting system, IOE equipment was completely removed from the Alipay system.

Throughout the alipay architecture in the process of the reform and the development of the SOFA, on the basic composition of middleware and thought is the reference, such as message middleware, data middleware, transaction middleware, etc, but SOFA technical team to be geared to the needs of very large scale Internet distribution reform of financial transactions, and the black technology such as unitary, “We found some papers, some concepts and some similar practices. But at that time, alipay was already very large, and no one was sure that it could be done, especially in such a high-risk business scenario as finance.”
“To paraphrase Peng Lei, the former CEO of Ant Financial, she mentioned that a lot of things people do is how to turn Jack Ma’s decision into the right decision, and we have a similar situation throughout the implementation of middleware engineering. For example, in the era of SOFA3, when Hu Xi put forward this concept, there was a huge internal debate. Everyone thought this was unreliable, difficult and complicated, and there was great resistance. The hardest thing is to convince the team. But in the end, people were proud that they could do it and save the company so much money, “Yang recalled.
In order to solve the new challenges, Ant Financial’s middleware technology team came up with various solutions. Yang Bing for RPC calls of unitized structure design of the routing logic: for all kinds of business system, some can be upgraded, some can transform, some not, then when the RPC remote service invocation will have five or six kind of branch to decide whether to local priority, or across the room, or according to the business of depots do routing table and so on. This logic is extremely complex, because there are both isomorphic room and heterogeneous room, and the heterogeneous room needs to converge the communication to a proxy, so there will be agents, resulting in very complex. In order to manage a system that cannot be upgraded, even if the owner of the system is no longer there, a technical proxy like Service Mesh is used to “disguise” the Service. “The whole architecture is beautiful, but every detail of the engineering implementation is complex, and all the design is imbued with the wisdom of architectural balance. After we implement the whole process, we will slowly remove the three or four routing logic that is completely unnecessary and turn it into a more orderly mode. Basically is such process, because cannot pay treasure to stop.”

Wang Lei (alias: Wenruo), who was in charge of the SOFA system’s messaging middleware, recalled that Alibaba started to hold double 11 in 2008. In the first year, it was just a trial run, so there was no large-scale publicity, and even many people inside did not know about it. Since 2009, Alipay and Taobao began to participate in the Singles’ Day together. At that time, all the goods in Taobao mall were advertised at half price. However, since we did not have a clear understanding of the power of singles’ Day in 2008, various problems suddenly appeared in 2009. Wang lei was in charge of messaging middleware, known internally as Message Broker, which was a Message queue: messages were sent from upstream applications to downstream systems, including Oracle databases, which were still in use at the time. “Was watching the process of this activity, even we ourselves also when buying things, all of a sudden the DBA ran to get the message current limit, because immediately downstream of the database will not hold, the database of the log space is going to be swallowed up, if swallowed leads to a database outage, restart up to a few hours later. At that time, there were three members in our team, and we had never been able to quickly limit the flow of messages before. Finally, we had to manually log in to the server of the upstream application, and then type commands on the server to do flow control. Finally, we saved the downstream system from being washed down. At the time, it was a pity, because instead of relying on the messaging middleware to limit the flow, it actually killed the upstream messaging application. And then, after that, we decided to do something called one-click limiting, one-click limiting capability for message centers at the middleware level, and that was built from that day.” There are many more such stories. “The whole process is like leveling up, killing each one you see.” Wang Lei’s practice represents the working state of the whole SOFA team, and also represents the biggest difference between SOFA and other Internet distributed middleware — precipitation of alipay/Ant Financial service for more than ten years, the best shared practice of the whole business and IT system.

In this way, the evolution of SOFA with the evolution of alipay’s entire structure and development, Cheng Li recalled, the first generation of SOFA is relatively simple, just built a framework and model so that the system can operate, to the late system operation to do a lot of optimization, These include the communication performance, the most efficient disaster recovery, the construction of remote disaster recovery architecture, the transformation of the unit, LDC logical data center project, etc., all of which have been gradually settled in SOFA. SOFA has evolved from addressing distributed services and distributed transactions to a real infrastructure addressing financial level system building, so it’s been renamed, Scalable Open Financial Architecture Scalable Open Financial Architecture This framework can truly solve the disaster recovery and expansion problems of the financial level system, and SOFA’s scalable capacity is not only to process more transactions, but also to accommodate more business, can let thousands of engineers or even the future tens of thousands of engineers to work together with the scalable architecture; Open means that it hopes that this framework can make business applications relatively easy to use and organically integrate with the classical architecture system. In the future, SOFA framework can not only arrange the business logic written by Ant Financial engineers, but also arrange the business logic of partners to become a complete arrangement framework. Financial means that SOFA must have the finance-level attributes to truly achieve finance-level consistency, availability and stability.

SOFA design philosophy

The development of SOFA from the first generation to the fifth generation is an extremely complex and huge process. Cheng Li summed up the RESEARCH and development methodology and experience of SOFA and said: In terms of the whole research and development method and process, Ant Financial conducts finance in the way of the Internet, because Ant Financial itself is a financial system. In addition to strict requirements, Ant Financial also hopes to have the iteration speed of the Internet, and accumulated experience in this process.

First, focus on the governance of the architecture. Ant Financial has always had an architect team, which includes both the general architect team and the architects from different domains, so as to keep the architecture of Ant Financial on a single map. The iterative evolution of Ant Financial’s architecture is also very clear, from the first generation, second generation, third generation, fourth generation to the current fifth generation architecture, there are very clear evolution stages, which is the result of top-level design from the governance level.

Second, focus on the risks of technology.Ant Financial makes 10% more efforts than other Internet companies in developing any system to ensure the control of system risks. Therefore, the technical risk control of Ant Financial is embedded in the process and controlled in the whole research and development life cycle, thus achieving a very good control. Of course, ant Financial’s R&D efficiency team will also control the r&d personnel to make it as simple and intelligent as possible.

Third, system optimization is realized under the premise of highly distributed. Ant Financial is a rare company with thousands of people working in a business process. The core payment process is divided into many layers from the front end to the back end, and there are many functions in each layer. In this process, it is very critical to ensure excellent distribution and integration efficiency as well as quality. Therefore, from the requirements, analysis and decomposition of the whole project management, to the realization of each team, and finally to the efficient integration and iterative release, I have accumulated a lot of experience on the R&D deployment operation and maintenance platform. Ant Financial’s R&D, deployment, operation and maintenance platform has undergone many generations of changes, and for some time, it also introduced the whole set of methods and tools from IBM and other suppliers. After about two years of use, it was found that the working mode and speed of Ant Financial were completely unsuitable, so it turned to a self-developed platform and kept lightweight. But it is not an open source model, because it does not apply to Ant Financial.

Fourth, another consensus of ant Financial’s middleware team is Design for failure, which assumes that in any case there is a risk of unreliability at the bottom.For financial IT systems, any temporary failure of underlying hardware or network jitter can be magnified to the level of loss of capital or overall service stability. For such a large-scale Internet application as Alipay, the ability of high reliability, high availability and high performance should be transferred to the middleware layer and database layer to ensure from the beginning of design. The following IaaS must be simple and allow the underlying hardware to fail, leaving the middleware and database layers to take care of the failure. Why would you do that? It depends on the tolerance of the business, so it can’t sink down to the IaaS layer, but there’s no need for everyone who writes business code to write their own code, so it sinks to the middleware layer.
Middleware is all people, will affect all systems, like ants gold suit now have thousands of systems and tens of thousands of micro service, thousands of r&d staff, how can you make in the middleware layer to do every work can make the overall architecture can be a smooth upgrade, but don’t let the business system, how to establish a connection with other developers, how to balance and operational efficiency, This is the middleware challenge.
Asked what’s the difference between a SOFA with other micro service platform, Yang Bing an example “if there are two sets of architecture in the top-level design, a set of balance will tend to be the” cost “, a set of favoring a “minimal risk”, there will be one thousand times in the process of implementation design decisions based on the principles of the great balance “architecture”, At the end of the day, the architecture will be completely different. Just like the balance in CAP theory, what kind of business determines what kind of technology will be incubated, and the technology will ultimately serve the business.

conclusion

Innovation is forced out, ant Financial self-developed SOFA is the same. SOFA takes a different path of distributed architecture from the traditional financial industry. To achieve finance-level performance and reliability based on unreliable hardware systems, there are many difficult problems to be solved in dealing with super-scale Internet financial applications such as Alipay. Ant gold suit to build a set of processing level financial transactions of a distributed architecture and platform, the financial level of consistency and mass concurrent processing ability achieves a good balance, and the quick disaster recover, elastic scaling ability, ability to ensure high availability are much more live and on-demand supply fine resource scheduling ability precipitation has abundant practical experience.
With the launch of the science and technology opening strategy in 2015, ant Financial technology team is not only faced with internal business, but also with more complex and changeable external business scenes and technical challenges. Previously, the technical team was a support organization forced by the business, but now it has gradually changed into a learning organization and an innovative organization that drives the business. “The best performance yesterday is the minimum requirement today. As Double 11 has become a regular work technically, meeting the transaction business has become the most basic requirement, so we need to look further into the future and prepare to meet stronger challenges.” Yang bing’s words, from one side to explain ant Financial technical team to expand the boundaries of the broader digital financial world expectations.
“Thanks for sharing, will you always do technology?”
“It will.”

Click follow to get the latest distributed architecture

Welcome to jointly create SOFAStack https://github.com/alipay