ThoughtWorks global CTO: build the architecture according to needs, and excessive engineering will only "waste people and money"

author | Wintersweet
Interviewing guests | Rebecca Parsons,Thoughtworks The global CTO

Thoughtworks Founded on 1993 year , At present, it is near 20 Countries have offices , Have 1 More than 10,000 employees . these years ,Thoughtworks Has been developing for agile software 、 Continuous integration 、 Continuous delivery 、 Microservices 、 Provide services in terms of evolution architecture and data grid , Is one of the first companies to develop agile for complex software projects .

these years ,Thoughtworks The technology radar summit has always been well-known in the industry “ Technology weathervane ”. The summit was founded by a technical advisory committee composed of a group of senior technical leaders , They meet regularly to discuss Thoughtworks Global technology strategy and technology trends that have a significant impact on the industry .

On the occasion of the technology radar summit ,InfoQ I have the honor to interview Thoughtworks The global CTO Rebecca Parsons, Ask her to talk with us about the release of technical radar for so many years , What value do you hope to bring to you ？ What technology trends deserve attention in the future , And how a technician should keep technology forward-looking .

stay 1999 To join in Thoughtworks Before ,Rebecca Is a computer science researcher and university lecturer . After completing the master's and doctoral degrees , She is mainly engaged in compilers 、 Program optimization 、 Distributed computing 、 Programming language 、 Computational theory 、 Research on machine learning and Computational Biology .

2007 year , She became Thoughtworks Chief technology officer . In addition to depth technology , She also advocates diversity and inclusiveness in the industry , Especially add coding and STEM Number of women in the field .

The following is the transcript of the interview ,InfoQ Edited without changing the original intention .

InfoQ： I'm glad to have the opportunity to interview you , You can go to InfoQ Would you like to introduce yourself to your readers ？

Rebecca： I am a Thoughtworks Chief technology officer of the company , Although I don't want to admit , But I have been in the computing technology industry for a long time . But I have always enjoyed this process , I feel like a typical geek .

My technical focus is very broad , The doctoral thesis studies programming language and compiler optimization , In concurrent and distributed systems and Evolutionary Computing 、 In particular, genetic algorithm has also done a lot of work . There is also Computational Biology , In short, I have a broad interest in Computing .

I am a 1999 year 12 Month join Thoughtworks Of , from 2000 In, I officially started my work . lately , I participated in writing a book named 《Building Evolutionary Architectures》 New book , The second edition is currently being revised .

InfoQ： The technology radar summit has been released for many years , What value do you hope technical radar will bring to you ？

Rebecca： In fact, technical radar is the result of our cooperation . At first, an internal salesperson asked us , Which technology is the coolest at present . We had a serious discussion on this topic . actually , At that time, he was the technical assistant of the technical advisory committee Darren Smith It should be “ Technical radar ” The originator of this concept . But we won't divide it into so many details , It is basically the result of collective wisdom and effort . At that time, a member of the committee was talking to a British friend , The other side said “ I'm curious Thoughtworks Staring at something in the radar ”, This sentence gives us the greatest inspiration .

So technical radar was born . I think the value it brings to you is mainly divided into three points . First , Technical radar covers the world , So here comes India 、 China 、 Brazil 、 Technical information in North America and even Europe , Technicians all over the world have the opportunity to contribute . But we also found that , Sometimes people have different opinions about the same technology . I remember , Once a British team wanted to stop a certain technology , But another Indian group felt that it should be vigorously promoted . After a thorough study , We found that the customers of both sides are completely different , Therefore, the use of technology varies greatly . That's why I think , We must always maintain such a global perspective .

Second point , I think it's solid practical experience . This kind of experience cannot be gained by chatting with suppliers or analysts alone . Of course, what is presented in the description document is also experience , It also works well , But it hasn't really turned into your own wealth 、 You may not be able to use it easily .

The third is , We will clearly express our attitude . Overall , Technology radar is more supportive of open source technology 、 Relatively weakening commercial products , And they all speak with real security and data . After all, if there is no good privacy protection means and clear architecture system design ideas , It is impossible to build an excellent architectural scheme .

This is our consistent insistence in technical radar .

InfoQ： You are from 1999 I joined in Thoughtworks, After so many years of work experience , What is the deepest technological change you feel ？

Rebecca： I feel the deepest technological changes twice , And it's very different .

When I first started working in business technology , I found that without the help of database , We can't even write any data into the disk . at that time , Everyone feels that all data are related to each other , So I will put all the data “ Stuff in ” relational database . Later, those Internet giants , Google 、Netflix La , Their appearance really broke this mentality , Make people believe , Not all data are related .

For example , We can also store the graph in the traditional database , But this does not reflect the graph properties . And in new technologies like product catalogs , The properties contained are much richer , It can also be easily put into the document database . So for me , This marks a fundamental shift in our view of Computing . That is to say, after having a relational database , We have a new problem ： What kind of data do we need to keep ？ How to query these data ？ Which Persistence technology is more suitable ？ Put forward these questions , Represents a real change in thinking .

As for the second change , When we have the foundation of continuous delivery , Will be based on the entire deployment pipeline 、 Test automation system attempts to promote automation . This also allows us to break a new , The experimental construction road that was unimaginable in the past . before , No one dares to imagine something like microservice architecture —— Manually enter the deployment scripts of dozens of micro Services , It's impossible and unsafe . Too many services mean that we introduce too many risks into the architecture , So no one did it . It is also the difficulty of system deployment , That makes it impossible for us to explore the system organization .

But now these restrictions have been broken , Just follow the principle of continuous delivery , You can split the system into modules . Even in the face of hundreds of modules , Automation technology can also maintain the deployment difficulty at a very low level . After all, computers don't care what they have to do , Or a hundred things . So , We no longer need to consider constraints , You can choose the most suitable architectural solution purely from the perspective of deployment and operation .

So for me , These are the two fundamental changes in the field of computing technology . With these qualitative changes , Other quantitative changes will follow .

Build architecture on demand , Over engineering will only “ harass the people and waste money ”

InfoQ： In the context of digital transformation , In addition to some hot topics about basic software , People also pay more attention to evolvable architecture , You think build “ Evolvable ”、 Can face the future （ Decade ） What is the key to the development of the architecture ？

Rebecca： Say evolvable architecture , I think the most important point is that , Focus on what you want to achieve through architecture . Concrete implementation is always for the result .

As an enterprise architect , My job is to determine the need for scalability 、 What degree of security or flexibility . I will not specify which load balancer must be used , Only reasonable load intensity targets will be set . So if you can turn your attention from concrete realization to normative goals , The final result will be better , Because in this way, the system construction team has flexible space 、 There is a clear direction , They know how to choose by themselves . Elastic 、 Scalability , These are objective measures .

With such guidance , The system team can weigh the impact of different architecture choices on the results . And in the face of new technologies , We will still push back the plan based on the results — Meet architecture requirements , Then use ; If it doesn't meet , No matter how new the technology is . So our architecture will be more flexible , We can focus on what we need to achieve , No longer obsessed with a specific time 、 What specific solutions should be used in a specific location .

This is when building an evolvable architecture , The mindset change that must be completed —— Result oriented . I don't care what you do , But it must be able to achieve a specific uptime or scalability . After the requirements are clear , The team on the front line knows more about how to make choices . They are free to choose technology , Decide how to make your goals a reality .

InfoQ： So you mean , We must prevent over engineering , Right ？

Rebecca： you 're right , But it's a pity. , I often see over engineering 、 Over designed system . Not every system needs 99.999% The reliability of the , Not every system needs to cope with the load scale of Google . It is enough that the system we build can meet our own needs and reasonable expectations , Why do you have to follow Netflix As exaggerated ？

InfoQ： Can you share some over engineering 、 Cases with low output ？

Rebecca： In the speech on evolvable architecture , My favorite example , It's the story of our opening an office in Britain . We had a British client , They run a sandwich takeout system . Every day in the morning , Diners can place orders on the system , Then they are responsible for delivery , Customers' credit card information will not be stored in their systems .

Our office is in central London , So whenever something goes wrong with this system , Let's go downstairs and have a look at those sandwich shops at the door , Can immediately find clues . This does not involve confidentiality at all . I order a sandwich and eggs every morning , There is no personal identity information in the order 、 There is no credit card information , So there is no need to encrypt . Even if the system is shut down for maintenance , I just need to go downstairs and go to the store myself , It doesn't matter at all . For such a simple system , use rails Such development is enough . Orders created by customers will remain for a period of time , And then it's the weekend to delete . All in all , This is naturally the simplest system , There is no need to force some complex functions into it .

I often mention another case . At that time, we built a trading system for our customers —— On hearing the trading system , Many people first think of “ High frequency trading ”, That is, the frequency is very high 、 The delay is very low . But most trading systems are not that complicated , It only needs to be handled every day 100 Tracking transactions for about times . So performance is not a big problem , Instead, the success rate of the transaction is the key , After all, each transaction is worth billions of dollars .

So the situation is completely different from the first impression , This kind of trading system is not in the same dimension as the high-frequency trading system . So we still need to return to the current reality , Think about how to solve 、 What features are introduced into the system , That's the essence .

InfoQ： Facing the current trend of digital transformation , Traditional and financial enterprises are often afraid of distributed architecture or micro service architecture in the process of digitalization , What advice did you give them ？

Rebecca： I often quote Martin Fowler An article written , Talking about the applicable threshold of micro service architecture . Yes , The microservice architecture is quite complex , There are applicable thresholds .

I am also in an article , Explained why we will never use microservices on technical radar . Although microservices do solve some specific problems , But the complexity of the architecture itself has greatly increased . As you mentioned, distributed architecture , They are really complicated . If other high-quality models with simpler structure are adopted , Many faults will not happen at all .

So organizations should first clarify what benefits they can get from the microservice architecture . These distributed architectures or micro service architectures are certainly valuable , But you have to prove that the extra complexity is worth it . Each microservice can be extended independently 、 Flexible function , These are real advantages , But it will also give IT The Department brings greater operational burden .

I've mentioned it before , Continuous delivery is the real core , So if you can't guarantee this , Micro services are meaningless . So if IT The Department is very mature , Then you can safely and boldly use the microservice architecture ; But if the original model is completely enough , Why bother yourself ？ In case IT The Department can't decide the micro service mode at all , That will only make a mess of the good business .

So think about the benefits of microservice architecture , Whether it can play irreplaceable value in the organization . Weigh the benefits against the costs , I believe you have your own answers .

It is urgent to build responsible technology

InfoQ： You talked about building responsible technology at the technology radar summit , How to understand your speech theme ？“ Dangerous technology ” It doesn't appear in recent years , Before that , So why do you choose to talk about such a topic at such a time ？

Rebecca： For me, , The greatest significance of technology lies in , What changes can it bring to people's lives . I'm not tired of Technology , At most, I buy things according to the recommendations on Amazon , And I still like those recommended . There's really no big problem . But there are other key areas , For example, medical decision-making 、 The criminal justice system 、 Credit score 、 Fraud detection 、 Face recognition and so on , Errors related to this can have serious consequences , Naturally, we cannot take the corresponding technology lightly .

As a technician , We need to think about once we make a mistake , What kind of impact will it eventually cause . Can we bear the consequences ？ Do we really know all kinds of data used in system training ？ About prejudice , I often give such an example . Although prejudice is not necessarily malicious , But the impact still cannot be ignored . A hospital plans to train a set of models , Guide medical students how to design patients' postoperative recovery plan , In particular, is it necessary to send patients to the intensive care unit （ICU）. result , The materials in the data set are biased against asthma patients , So it will automatically send all asthma patients to ICU.

The fact proved that , There is no information about asthma patients in this data set , These patients are treated under different medical systems . But the hospital did not take this into account when training the model , So the model knows nothing about asthma symptoms . Over time , If a hospital really uses this model directly , Then it will definitely cause a lot “ There is something wrong with the medicine ” The problem of . This model certainly has no malicious bias against asthma patients , But the hospital did not carefully screen the data set used in model training . We are in the actual work , I have been exposed to many systems with similar problems , That's why I think responsible technology is a very important topic .

Fortunately, we still have the collective memory of rational decision-making , It is not too late to correct it in time .

InfoQ： Under this topic , What technologies with negative effects do you think should be focused on and discussed ？

Rebecca： In fact, everything has negative effects . I think the biggest negative effect , In fact, it comes from the wrong decisions made because of the lack of understanding of the underlying data . It's not just making wrong decisions , It will even reinforce these mistakes 、 Make it more likely to make a wrong decision next time . After all, this is the basic principle of reinforcement learning .

But from another aspect , With the higher and higher level of Automation , The speed of problems in many systems is also getting faster . If there are human elements in the whole cycle , At least wait for them to press “ confirm ” Button 、 Or enter some instructions . But if it's machine to machine interaction , These delays will no longer exist . In this way, we get a feedback loop of rapid upgrading , Mistakes can make trading companies declare bankruptcy within minutes . This is the harm caused by algorithm errors , The whole combination will instantly disappear . And if someone is involved , These problems will not happen .

So we must establish such a more reliable automatic cycle , Otherwise, something unexpected will happen sooner or later 、 In a situation where we simply cannot predict how the automated system will react . Autonomous driving is a typical example . The goal seems simple ： Let the car drive along the highway , Avoid other cars and keep in the middle of the driveway . Is it difficult ？ However, some unexpected situations often occur during driving , For example, a deer suddenly jumped onto the road , Or someone broke the driving sign 、 The system cannot recognize it normally . For this kind of fully automated system , We must think comprehensively about the possibility of their mistakes , And set circuit breakers in the system , Ensure that it can be shut down before the unexpected consequences occur .

InfoQ： The negative effects of technology are often unpredictable , What management measures or frameworks do you think can be adopted for general enterprises 、 Tools make technology more responsible ？

Rebecca： exactly , That's why we released 《 Responsible technical manual 》, You can see it on our website . It includes a series of promotion technologies , It can help the development team effectively break their own inherent views , Look at the problem from the perspective of other different technicians . Yes , When discussing responsible technology , We will start from the problem .

We will find out what the problem is first , Let's see who the problem will affect , That is, the object we need help . We will also think from the standpoint of the affected , For example, what they may need ？ What concerns may they have ？ The methods in this manual are not all our own ideas , We just collected these ideas and integrated them , At the same time, indicate the corresponding source . With such rich promotion technology , We can look at our technology choices more comprehensively 、 Consider the possible impact of these choices . After all, technology itself has no position , Therefore, it is impossible to predict the unintended consequences after implementation .

This is the significance of the concept of responsible technology . As a technician , At least we have the responsibility to think further , Predict the unintended consequences that technology implementation may cause . There may be no consequences , Maybe it can improve the products and expand the target market , These are good results . But some unintended consequences may also hurt others , Maybe there are ways to prevent or alleviate it in advance , This is our responsibility as technicians .

“ Love of Technology , Let me come to today ”

InfoQ： I see that you have been exposed to programming since high school , What is the earliest opportunity to contact programming ？

Rebecca： I was studying algebra at that time , Our high school math teacher happens to be learning programming in a local university . He bought a textbook , I remember it was 1974 year , I happen to be in high school . He also gave me a textbook . At that time, there was a keyboard puncher in the school . do not know why , Our school has never opened this course , But I can write a program with this machine and input it into the punch card . after , The algebra teacher will take my card to the University , Try to run in class , So I learned to program . Rely on the textbook of algebra teacher to learn programming , I came into contact with programming for the first time .

InfoQ： How did you feel when you first programmed , Do you like it? ？

Rebecca： Yeah,I like it . I seldom recall the past , But I still remember the feeling of being able to solve problems with computers . This is a completely different way of thinking , It's really fascinating . Follow this way , We get the program , I especially like this feeling .

InfoQ： At that time , Do you think data will change the world ？

Rebecca： It's not so much data , Rather, technology can change the world , At that time, I believed that programming and technology could change the world .

Because these are real jobs , Can make things better . We can program accounts 、 Accounts receivable 、 Payroll and other jobs . For these simple structured tasks , At that time, people couldn't write formulas directly to solve them . But the program can , So this is a revolutionary leap 、 Can make the world a better place . It sounds a little naive , But we did think so at that time .

InfoQ： The technology category you are concerned about is also very comprehensive , How do you maintain your enthusiasm for learning and maintain your skills “ Forward looking ” Of ？

Rebecca： Mainly because I always have a group of partners who also love technology . Now I also host a podcast program , I especially like the process of recording podcasts , Because everyone can sit together , Discuss some technical topics of mutual concern .

In this hour , I will talk as the host 、 Feel others' excitement about some technical details . That's why we hold the technical radar conference twice a year , Here you can access a wider range of technical tools 、 Platform and method . This is how I keep my passion for Technology .

InfoQ： Keep the technology forward-looking , I guess the only trick is to keep learning 、 Never stop , Am I right? ？

Rebecca： you 're right .

InfoQ： As Thoughtworks Of CTO, Where are the challenges facing this role for you ？

Rebecca： When I first started working in technology ,CTO My identity is somewhat like a reporter , Just in the operating system 、 compiler 、 Just choose between databases and Applications . But now the technology field is too extensive , So reporter type CTO It's no longer true , Any specific type of technology will cause corresponding types of problems .

for instance , When we face the Internet of things , Then some problems can be directly ignored ; But when using cloud servers , The same problem has become crucial . The whole technological ecosystem is much more complex than it was . So I will start from the train of thought 、 Look at these technologies from multiple perspectives, such as effectiveness , Understand its connotation .

Thoughtworks The size of the company is not too large , There are 11000 Employees , But it is no longer small . So we must make our own choices 、 Maintain a rigorous and critical attitude towards technologies that may have a significant impact on customers . I've been worrying about whether some of my technical judgments will go wrong . I'm most afraid that the decisions I make will cause customers to complain , There must be no such thing .

All in all , As the technological environment becomes more complex , Customers use technology more and more widely . Over the years , We are used to deploying sensor networks in specific types of application scenarios , But in the past, these systems were only applicable to certain specific problems , Now we need to develop in the direction of universal application . in the capacity of CTO, I think it's really more and more difficult to choose technical solutions now .

InfoQ： Most developers / After a few years as an engineer , Will face the choice of whether to switch to technology management . Many people will feel confused about this , Their problems may include “ Do you want to transfer to technical management ”“ Whether I am suitable for technical management ” or “ If you don't manage , Is there a future to continue to develop in the front line ” wait . How do developers make the right choices , What should be paid attention to in different choices , Based on your experience （ Including your own experience and the experience of the team members you manage ）, Give you some advice ？

Rebecca： This should be divided into several points . First , Some people are more like lone Rangers , I did it my way 、 Don't like dealing with teams . But some people especially like working in teams , But unwilling to undertake management affairs . It's all normal , But sometimes it will limit one's development space .

This is also the biggest obstacle for everyone to engage in technical management . As a developer , We are facing problems that can be quickly fed back ： Did the code compile successfully ？ Did you pass the test ？ A very specific index can clearly measure the effectiveness of the work , Let us feel our progress . If the test fails , We'll go back and fix it , It doesn't matter. . But as a technology manager , This more immediate success began to fade , We may not be sure whether what we do is right for a long time . This is the most difficult for many people to adapt , Because they are used to doing the right thing 、 Instead of instructing others to do the right thing .

I think it really needs a major psychological change , But back to the question of maintaining technical enthusiasm you mentioned before . I think even if you are in the technical management position , We still need to maintain the technical level , This is more helpful for work . It's the only way , Technology managers can distinguish between good and bad technology , Understand the shortcomings of team members 、 What kind of problems they face 、 How to help .

So in my opinion , Whether it's pure technology or management , Maintaining enthusiasm for technology is the most important prerequisite .

InfoQ： In fact, male practitioners are in the majority in the developer industry , But you did CTO, There are also many works , Built extraordinary achievements , these years , How do you balance your career and family ？ Can you give us some guidance or experience on women's growth ？

Rebecca： The development path of women in the technology industry is still very difficult . I'm a hard core geek , So I never thought of quitting , Weizhi runs all the way . But for many other women , These obstacles do affect their feelings 、 Even force them to quit . If you want to stick to it , The most important thing is the strong expectation of continuous learning of new things . in addition , Cultivate self-worth 、 The ability to ignore noise is also helpful .

What really keeps me going , Or the love of Technology . I like learning technology , Explore how technology is used .

I like learning new technology 、 Like to work with everyone . For what you love , We also have to make a trade-off , Figure out what is most important . What I regret is , Many male technical workers can stick to it out of generous pay and interesting work , At the same time, I hope there will be more room for improvement ; But for women , Their criteria have changed again , It seems that a good job suddenly becomes not so good . So I see women in technical occasions , Many people's first reaction will feel that the other party is a secretary ,“ Give me a cup of coffee. ” such . I have encountered similar situations , They feel that middle-aged women who appear in the office , It's definitely impossible to engage in technology .

So if everyone is as enthusiastic about technology as I am , Then please hold on , It's all worth it . But if in your eyes , This is simply a job , It's better to find a position with similar salary like project manager . exactly , In the technology industry, women will still be somewhat discriminated against and offended , There is .

All in all , Now the whole industry has done better and better in diversity and inclusiveness . But if all female friends are not enthusiastic about technology 、 Just want to find a good job , That may not be suitable .

InfoQ： Thank you very much for accepting our interview ！

Rebecca： thank you .

---

Link to the original text ：https://www.infoq.cn/article/weyXNTEtOFXa1sF6dqe5