Exclusive interview with Cheng Xianfeng: The popularity of APM reflects the pain of operation and management

Xianfeng is the Chief Operating Officer of BlueOcean Communications (including the SaaS-level OneAPM and enterprise-level blueware business brands). He has 10 years of experience in the IT field and is a well-known technical consultant. He is the initiator of the MongoDB Chinese community, the chief evangelist of AdMaster, and co-founded the Beijing R&D center. He was the vice president of JiMuBox Technology and holds a master's degree from the University of Sydney and a bachelor's degree from Harbin Institute of Technology.

[[123883]]

“I cover a wider range of fields, but I spend most of my time doing technical management work.”

I have known Xianfeng for a long time. I remember doing an interview with Xianfeng about agility when I was at AdMaster. At that time, he always emphasized the gap in management in China. There were few very experienced people to do engineering in this area. Perhaps due to long-term accumulation and perception, Xianfeng believes that technical management and communication are the core values, which can quickly improve the growth of junior programmers. So, Xianfeng has been responsible for technical management, team building and employee training at AdMaster. Later, after leaving AdMaster, he continued his work in technical management at JiMuBox and now Blue Ocean Communication.

This year is the first year of APM. What do you think is the main reason why APM has suddenly jumped into people's vision and attracted the attention of the industry in China? What changes will there be in management?

Cheng Xianfeng: Let's talk about what APM is. Now everyone is talking about application performance management, but in fact, the understanding of this in China is still very superficial. APM is the first year in China, but many Internet-based companies in the United States have been using it in 2008. For example, large companies such as IBM, Dell, and HP all have their own APM solutions. In fact, it is not a new concept. Why is it only recognized by everyone in China now? I think a big reason is that the software life cycle in China is short. A software abroad may have a life cycle of five, ten, or fifteen years, and its business system is also very stable. But in China, most software does not have that long. In this scenario, the motivation to adopt service-oriented software is obviously insufficient. For example, if you buy a car and plan to drive it for 30 years, you will definitely go for careful maintenance. But if you only plan to drive this car for one year, you may not pay too much attention to maintenance, or the maintenance is very poor. APM is like efficient maintenance in the automotive field. When there is a problem in the system, it can quickly detect and alarm.

People often ask, is APM the same as monitoring? In fact, APM and monitoring are different in nature. Monitoring can find problems but cannot diagnose them, but it does have the effect of monitoring to some extent. It's like a thermometer. It knows that you have a fever at 39 degrees, but it can't tell where your problem is. We can think of APM as a CT machine. It knows which part of your body has a problem and can clearly locate it.

Looking at operation and maintenance, there is a great lack of operation at present. Many IT companies tend to focus on R&D and neglect operation and maintenance, mainly because the system life cycle is very short, so the investment in operation and maintenance is insufficient.

The situation abroad is like an airplane, which can operate normally for more than ten years. It pays great attention to safety and stability, so it is carefully maintained and the annual cost is very high. It is also the same for large domestic enterprises, whose production systems have been in operation for a long time. In fact, domestic financial, telecommunications, and energy industry customers have applied APM for a long time. So the acceptance of APM by the industry this year is a result of Internetization, not a very early concept.

Why is APM becoming more and more important? This actually starts with the difficulties faced by IT system operation and maintenance:

First, systems are becoming more complex and distributed.

Second, there are many legacy systems, and we have to ensure the normal operation of these systems that were built five or ten years ago, which puts a lot of pressure on operation and maintenance.

Third, business is always changing dynamically.

As we all know, IT company operations are dealing with different changes every day. Traditional operations have many responsibilities, which can lead to being stuck in a quagmire and unable to extricate yourself from doing more value-added things. In the United States, operations are called IT Operation Analysis, also known as ITOA (IT Operations System Analysis), which is a holistic analysis of all indicators such as IT configuration, performance, data leakage, and compliance based on big data.

Understanding it from an analytical perspective is very good and can solve a lot of problems. The problem is that if you don't have APM, you won't have the data needed for analysis. Traditionally, data analysis comes from three aspects: network-based data, log-based data, and APM-based data. In a broad sense, these aspects are all APM industries, but if you understand this APM based on probe (Agent) technology in a narrow sense, the data it provides is almost incomparable to the network and logs.

***, let's talk about logs first. If you enable too many logs, you will be drowned in a large number of log events and noise, and it will be very difficult to extract. However, if the level is very low, you will find that the problem cannot be located at all. In fact, the intensity of logs is very difficult to control, and it is also a way of post-analysis, and the delay is more prominent and obvious.

Second, although the network has full data, ITOA runs in the more important real network structure diagram, and the most important indicator for diagnosing network failures is MTTR (mean time to recovery). However, network failures are not very helpful for the mean time to repair, and can only be located at the machine or HOST level. In fact, the help for diagnosing problems is not clear from the current complexity. Therefore, it is necessary to upgrade to the application level to solve the problem.

APM is based on the concept of ITOA, including feasibility analysis, performance and other dimensions, and is the most difficult of all ITOA components. A usability analysis monitoring can be done by yourself, but regarding the implementation of Performance. A third-party organization conducted an in-depth survey in the United States, and everyone unanimously recognized that Performance is the most technically demanding in ITOA. Therefore, when it is difficult to implement, I think you should find a partner, just like the most difficult chip in hardware. IT companies rarely develop chips independently, and everyone saves development costs and manpower investment through cooperation.

More often than not, what we need to change is our mindset so as to gain the overall value of IT operations, rather than saving costs in a certain dimension. We should analyze whether the investment is worthwhile from a long-term perspective.

Based on your many years of management experience, what abilities and characteristics do you think an excellent team should have?

Cheng Xianfeng: First of all, I think that as a technology company, technical capabilities are undoubtedly the most important capabilities of the company. What is completely different from other companies in terms of application performance management is that the business and its technology are completely integrated, so technology is our core business. Speaking of technical capabilities, we are proud of the fact that we can run stably in large-scale commercial systems and core systems in a very short time. We will not use some end-customer data to replace the data in our core system. At the same time, we will actively ask customers to do very strict POC tests, help them develop test standards and suggestions, and customers will also make open and fair comparisons. Therefore, we are also very confident in our technical performance.

Another is the ability to serve customers. Why is it difficult to use foreign APM in China? Network reasons are one aspect, but more importantly, I think it is the ability to serve customers and create value. We have a very professional consulting team that can help customers solve their problems very quickly. In the long run, I hope it is a team that is very self-reflective and has an endogenous motivation for improvement. Including our overall understanding of the APM market, and foresight is also very important.

Although this year is the first year for APM in China, I think our team has a very deep understanding of the development trend of APM. OneAPM did not emerge with the trend. After all, we have done a lot of application performance management for enterprises in BlueWare before and have accumulated many years of experience. So we have a deep understanding of this industry, which is also our core capability.

So in today's cloud computing context, what changes will occur in operation and maintenance and operating models?

Cheng Xianfeng: First of all, operations, this is a two-fold matter. For OneAPM, it is first of all a service in the SaaS model, and for us, it is a manifestation of the cloudification of traditional APM. On the other hand, for our customers, they also face many cloudification issues. Therefore, in the future, more and more enterprises and organizations will put their traditional businesses in the cloud.

Cloud computing is a huge opportunity for APM products. This is also an important reason why APM technology in the United States developed in 2008. The development of New Relic and AppDynamics has all relied on the trend of cloud computing. The reason is that after using the cloud, customers abandon many traditional operation and maintenance habits, and can better adapt to modern tools and new businesses. The cloud simplifies operation and maintenance, and strengthens its elasticity and management, which are all areas where APM is particularly suitable. Including cloud security strategies and the overall service concept brought by the cloud, the idea of ​​APM is naturally integrated with the cloud. Including the current mobile Internet, APM is a particularly suitable scenario. We also have this kind of cooperation with almost all cloud vendors in China, which is equivalent to closely combining our products with their customers and creating a very large value-added effect for their customers.

For operation and maintenance, the first thing is to have a very big change in thinking. Recently, Amazon held the Reinvent conference in the United States and announced that it deploys 50 million times a year. Although many people do not believe this number, 5 million times should be possible. If it is 5 million deployments, it is far beyond the imagination of many people and is a huge challenge for operation and maintenance personnel. Many operation and maintenance personnel cannot complete 5 million deployments even if they are given a year.

Where does the problem occur at this time? In fact, Amazon's operation and maintenance is completed by the technical team. The operation and maintenance team only provides some tools and is extremely automated. If everyone really uses the cloud, the role and mindset of the operation and maintenance team must undergo this kind of transformation. I think Amazon is a typical example.

Our domestic operation and maintenance personnel are doing routine online and fault diagnosis. Foreign operation and maintenance personnel do not do these things, at least many advanced companies' operation and maintenance personnel do not do these things. This is a trend. You are required to be able to develop tools and design the architecture very well. Therefore, this requires high standards for operation and maintenance personnel, and also requires them to consider issues.

Nowadays, many people are discussing topics such as DevOps and Operation. In fact, if you go deep into these topics, you will find that they are still back to the things that traditional operation and maintenance do. Some people have proposed a new operation and maintenance model, called ETA (Environment, Tools, Automation). Operation and maintenance will definitely have such a change in job responsibilities and thinking, and they will essentially embrace the new generation of technology. In fact, the technology of operation and maintenance has changed very quickly in recent years. In the past few years, we rarely heard of large-scale automated operation and maintenance.

Now the degree of automation is constantly improving, and the requirements for performance management are also constantly increasing. This is actually a challenge for operation and maintenance personnel, but also an opportunity.

Companies like HP and Dell are all doing monitoring and management, but they are mainly targeting large enterprises.

Cheng Xianfeng: Yes, but according to the Gartner report, it is bound to fail if they only target large enterprises. From the perspective of the Internet market, there are a few simple points. First, they cannot produce user-friendly products. Second, their deployment and delivery methods are completely unable to achieve the Internet. Third, they are too slow to adapt to the Internet, and their market share in this market will inevitably decrease, and then gradually lose market share.

The popularity of mobile Internet today has led to the emergence of entrepreneurial teams like mushrooms after a rain. Do you think there will be new competitors entering the APM field?

Cheng Xianfeng: First of all, APM itself is a niche market, and there are not many players. Even in a mature market like the United States, its players are only a few. In addition, the technical threshold of APM is relatively high, and it is a technology that needs to be verified by practice for a long time. Even if you develop it, many people will not believe that you can achieve a good effect. After all, you need to consider growth and integration, and there are many unfavorable factors for latecomers. However, the APM market is very huge, and we still hope that more competitors will enter this market.

The hottest words now are cloud computing and big data. Both of them are inseparable from performance. We have improved performance, which is an essential improvement. When the performance doubles, many things that were impossible before become possible. For example, sometimes there are often more complex advertising algorithms that cannot be calculated within the specified time. But after the performance is improved, it can be calculated.

On the surface, it is performance, but in fact, it has a huge impact on business. So I think the market prospects are still very broad. We hope that more players will enter this market and serve customers in a down-to-earth manner. After all, it is still in the early stages, and the market needs a process of mutual cultivation and maturity.

Currently, APM will face a huge market with the development of PaaS platform. What impact do you think it will have on business?

Cheng Xianfeng: I was one of the first people to come into contact with PaaS in China. At that time, the domestic PaaS environment was not yet mature. In addition, domestic cloud computing vendors did not provide corresponding PaaS platforms, so the industry now holds a relatively pessimistic view of the domestic PaaS platform. At least PaaS will not have a big move in China for about two to three years. If PaaS becomes popular, it can bring huge value to the APM product. Especially the typical strong combination of Heroku and New Relic abroad, there is no such platform in China.

Of course, if there are such platforms in China in the future, we are very willing to try to cooperate with them in depth. PaaS will definitely have a huge driving effect on APM products, but overall, China is still at the IaaS stage. PaaS is still a little far away from the market strategy.