At present, there are not many real intelligent operation and maintenance practices in the industry. Most of them are still in the stage of automation or even artificial operation. However, intelligent operation and maintenance is the general trend. How does Ali respond? Please see the speech “New Operation and Maintenance in the Intelligent Era” by Bi Xuan, head of Alibaba r&d efficiency team and ali researcher.
What are the responsibilities of Ali’s operation and maintenance system?
Ali operation and maintenance system introduction
Ali’s operation and maintenance team mainly covers five levels.
one Resource planning and disbursement are the cornerstone of operation and maintenance
The entire O&M team is responsible for resource planning and delivery.
Quota management: For example, we will do some budget management with the business team, for each business team needs to have a budget first. As long as you have the budget, the operations team will give you the resources, without the budget, nothing.
Planning: For example, for alibaba’s annual Double 11 transaction, the business team has to give the volume of the next year’s transaction, but the business team doesn’t care about the amount of machines that need to be added behind it. Therefore, the operation and maintenance team is required to do the transformation and planning from business requirements to resources, which is very important for the company, because it means how much money I will ultimately invest in infrastructure, as well as the control of the pace.
Procurement: When the scale is large, how to reasonably plan the quantity and delivery rhythm of resources is very important. For example, purchasing this batch of machines in May and purchasing this batch of machines in June are completely different concepts. It also needs the procurement of resources, such as the shortage of SSD procurement and insufficient supply. Usually larger companies have more access to better supply, and smaller companies struggle. How to do supply chain control is very important.
Resource scheduling: For the resource team, scheduling is also very important. How to hand over the machines we hand over, how to ensure availability and stability, and how to Bootstrap, etc. Each business has its own planning and how to hand over the entire business environment to the business side according to business requirements. Ali is currently facing great challenges. For example, in the international expansion, we may need to set up a point here this month and another place next month. How to quickly complete the delivery of the whole resource, not only of machine resources, but also of software resources, is very important. Now we are expanding our business in Southeast Asia. How to quickly complete the delivery of the entire software resources in Southeast Asia is very important for our competition.
two Change is the bane of operations
For the operation and maintenance team, change is also a part of frequent work, such as collecting change information, making application layer change, basic network IDC and so on.
3. Monitor to predict potential failures
Monitoring for Ali is mainly divided into basic, business, link, on the basis of monitoring to do some alarm.
Four. Stability is the goal pursued by many enterprises
Stability is a concept that we used to think was targeted at big companies, because it might affect the lives of the general public, it would be sensitive. But now new Internet companies, such as take-out, Ofo, Mobike, etc., have higher stability requirements than many entrepreneurial companies before, because it must be used at that point, if it cannot be used, it will have a direct impact on users. Therefore, stability may get more and more attention in the entire operation and maintenance industry, but for many small and medium-sized companies, stability investment is quite large.
Five. One key to build a site so that the scale of strong guarantee
In terms of stability, ali will mainly do multi-active system construction, and then fault repair, fault location, and then a set of full-link pressure measurement. Scale is a pain for a lot of operations teams. Maybe one year the machine is in this room, next year your infrastructure team might tell you, this room is not enough, we need to move to another room. Anyway, in Alibaba, a lot of operation and maintenance personnel have said that one of our annual work without writing is relocation. The infrastructure team will promise not to move again for three years, and then next year they’ll say, for some reason we’ll move again, and then they won’t let you move again for three years. But for the past three years, we’ve moved every year. In the future, we do believe that Alibaba may move less in the future. We believe that alibaba should not allow relocation to become the core competitiveness of its operation and maintenance team.
We have done a lot of things at the scale level, for example, we have done one-click website construction. For Ali, our requirements on the delivery time of machine resources will be higher and higher. For example, on Double Eleven, whether we deliver resources one month ahead of schedule, two months ahead of schedule or three months ahead of schedule, the money we pay is completely different and may be very different.
Therefore, the technical level can better shorten the time, is very important. Therefore, the important purpose of one-click website construction is this. Every year, we will expand many sites, and quickly complete the whole process through one-click website construction. Relocation is what I said, we have to move every year anyway, so we should do a better job of the relocation system. There are also moves, ali often need to do some business resources reuse, the best is to have a cabinet, at this time how to better complete the move process is also very troublesome.
We also need to make some unit adjustments, because there is a concept of unit in Alibaba’s transaction system, and how we can better control the ratio of machines in a unit is very important. The number of machines per unit may be fixed, so if the ratios are not well matched, it means that bottlenecks will be obvious.
These are the five areas covered by Alibaba’s operation and maintenance team. The evolution of the entire operation and maintenance system is almost from the earliest scripts to tools to automation to the future intelligence.
From tool to automation
From tool to automation, the process is not so easy, and for the whole industry, more work is still in the exploration of automation, how to make automation really be better.
I think the development of this industry is very different from other traditional software, standard software development industries. For example, in the process of Alibaba from instrumentalization to automation, we think instrumentalization is a relatively small challenge. Even traditional operation and maintenance personnel can easily write some tools, such as using Python to write more tool systems. However, if your tools become the most important to the stage of automation, it means that the requirements for tools will be higher and higher, such as the quality of tools, if you write tools often have problems, a large scale will not be able to carry, this time for everyone will slowly lose the sense of trust. In the end, it will be difficult to complete the process.
Operation and maintenance team transformation R&D team organizational ability is the biggest barrier
When Ali went down this path, we felt that the biggest challenge was the capacity of the organization. How can the operation and maintenance team better complete the transformation to the R&D team? This process is a huge challenge for many operation and maintenance teams. It’s also very important for an organization how it goes through this process.
I think a lot of teams have this feeling, it’s very easy to have some conflicts between the tool development team and the operation team and so on. Therefore, in this process, the core of Alibaba’s thinking is how to make an operation and maintenance team evolve into a better team we need from the organizational ability.
Ali went through four stages in this journey. It’s been a process of trial and error, and so far we think ali’s approach is relatively good. Like most companies, we started with a dedicated tool development team and a dedicated operation and maintenance team. The tool development team makes tools and makes them for the operations team. The most obvious problem with this process is when the tool is finished, the operation team says it is too difficult to use and does not meet the requirements. Or in the process of operation and maintenance team execution, there are often problems, and if there is a problem, we need to find the tool development team to help find the problem. It turns out that the tool team has to handle problems that can be solved by several lines of script. Slowly this situation becomes more and more difficult to break through, difficult to change.
So ali later made a try, since two team is hard to make a good combination, and that there is a way is a tool for research and development team with tools done, for example, made a release, after finish this function, the operational work thoroughly to the tool development team, don’t make operations teams, operations teams can do something else. This pattern appears to be a gradual takeover pattern that decouples the tool development team.
After doing this for some time, the biggest problem I encountered was organizational ability. For operation and maintenance tools, how to achieve high quality, operation and maintenance seems to be easy to do, but in fact, operation and maintenance tools are quite difficult to do, its complexity is greater than online business, but it is not logical complexity, more complexity is environmental level. Because it involves networks, it involves servers, it involves rooms, and so on, and it’s completely different from business. So after doing it for a while, we decided it was still a problem.
Integrate tool development and operation to break through organizational capabilities
Later, after finishing this round, we began to try in another direction, integrating the tool r&d team with the operation and maintenance team. The so-called convergence is to assign a lot of the tool development people to the operations team, to the operations team. We expect tool developers to drive the transformation of the operations team into a r&d team. That’s the idea.
It took about one and a half years for Alibaba to take the first three steps, which means that we have made three rounds of organizational restructuring. Because we think that all of this can only be done with organizational guarantees.
How does DevOps really land
Last June, we made the biggest organizational restructuring, handing over the daily operation and maintenance work to R&D. R&d will do all the daily operation and maintenance work by itself. But not all operations and maintenance work, there is still an operation and maintenance team, this operation and maintenance team is relatively different, very different from before.
We think this is where DevOps has really been thoroughly implemented. For this benefit is, the daily operations of the work to the research and development, operations teams into the r&d team this process is very difficult, in fact not entirely capability gap, the bigger reason is that operations teams to take a lot of daily chores, especially as a group company, whether it’s ali, tencent, baidu is the same, Conglomerates support numerous Bus. You support 20 BU by yourself, and if one person comes to you every day, you don’t need to do other work all day. You are constantly chatting with them and doing operations all day, calling for the team to upgrade, to upgrade the organization, to transform into a RESEARCH and development team. In fact, it is forcing others to go to a dead end.
So we think that what Google does, what Google mentions in the SRE book, is it forces the r&d team to spend 50% of their time on r&d. To be honest, it is difficult to implement this policy in most companies unless the operation and maintenance team and the R&D team have a very strong voice. But this is hard. Therefore, I think Ali’s approach is more thorough. Ali told the R&D team that the daily operation and maintenance work should be done by himself, not by the operation and maintenance team. This may be a bit rude, because the operation and maintenance system is not well prepared to do this thing, so relatively speaking, it also leads to problems later, such as operation and maintenance tools built everywhere, repeated construction and so on.
But on an organizational level, we were pleased to see that a year after this round of organizational changes, most of the operations team spent more time on r&d than on day-to-day chores. We saw the ability of a team to upgrade very well after this round of adjustment. And this is the biggest benefit for the organization. Therefore, we believe that this mode is the direction that Ali is most advocating and optimistic about, so that the whole operation and maintenance team will focus on the development and construction at the system level of the five parts I just mentioned, rather than the chores. This is ali from tool to automation, the most important is such a process.
Success rate is a key indicator of automated operation and maintenance
For automation, the most important problem is the success rate. For example, when we look at all operations and maintenance operations, the index we care about most is the success rate. The inside of the function, such as an operational system within a week, for example with hundreds of thousands of times, we focus on the success rate can be four more than 9, or calculate a normally singular will understand, how many people are there in the operations team to support this matter, these people don’t have the time to do the research and development work, and want to spend a lot of energy to do supporting work. Therefore, we need to ensure a very high success rate. The operation and maintenance system is the biggest challenge we have seen before. My previous background is all online business systems, such as taobao transactions.
One of the biggest differences we found in o&M systems is that o&M systems have a higher pursuit of success than online business systems. Online business systems, for example, if I have a problem accessing a later place, we will choose to fail the process as soon as possible, rather than continue to drag out time and trial and error. Online systems throw errors out faster. But for the operation and maintenance system, if we do the same, it means that the success rate is very difficult to guarantee. Therefore, the operation and maintenance system should have better thinking about how to ensure an operation and maintenance operation. There may be dozens of systems behind this operation, and most of them are written by countless teams. The situation that Ali met before was countless systems with poor quality and everything. How to ensure that in such a complicated environment, to ensure that the external level, the user level can achieve a high success rate. This is a big problem.
The challenges of scale are not to be underestimated
As the scale continues to grow, all open source o&M systems, as you scale up, as you scale up machines and so on, are generally going to be very challenging. For all of alibaba’s systems of this type, our argument is that it is more reliable to do it by ourselves. The biggest reason is scale, which will lead to many problems. Like code hosting, code compilation and other things, previously thought there would be no big problem, the fact that the scale of these are all problems. We also need to put a lot of effort into scale.
So I think, in the process of Alibaba moving from instrumentalization to more automation, the core issue we discussed was whether we could have a very good organization to complete the process. It allows the operations team to be more DevOps oriented. So we’ve been saying, we’ve been struggling with what the operations team should be called, and we’ve agreed that the operations r&d team, we don’t think it’s right, your main job is really r&d, not operations. But it’s kind of weird to call it R&D operations. And then Alibaba basically called the r&d team. Because we believe that there is no essential difference between the R&D team of operation and maintenance and the r&d team of online business. They are both engaged in R&D, just solving business problems in operation and maintenance. The five levels, the business issues in the field of operations and maintenance, are also business. There is no difference. Online business, like solving the problem of the transaction, solving other problems, it’s exactly the same. There is no essential difference between the two r&d teams.
So in this process, after the organizational adjustment of Alibaba in the past year, we see that Alibaba has made good progress in the whole automation level, but we still need to work harder to continue to evolve beyond our expectations.
Alibaba’s search in the field of intelligence
Now the topic of intelligence is very hot. For example, when the name AI rose, we suddenly found that all alibaba’s businesses talked about AI+ its own business, which was wildly criticized by everyone. We need to make it clear whether we have the premise of AIization, or even if we don’t have the premise, we continue to discuss the name. Because the industry is constantly heating up a lot of terms, let everyone to follow.
Automation is the premise of intelligence
For us, we think, for example, as I said to this team, my own team, I think the most important premise of intelligence is, one, automation. If your system has not completed the automation process, I think don’t do intelligent, you are still in the early stage. Most of the requirements for intelligentization are automation. If it is not automated enough, it means that a very good intelligent algorithm seems to have been made in the back, etc. Tell others that I can give you a great help, but it turns out that the previous automation process has not been completed.
One of the most typical cases is that Alibaba has always said that we think we can do better in the collocation of resources. For example, if you have a small amount of traffic in the middle of the night and a large amount of traffic in the daytime, can you better make some flexibility and release resources to do something else, and then make up for it during the day? It’s not that complicated on an algorithmic level, it’s easy to do a simple upgrade on an algorithmic level.
So we had a bunch of teams that built something that could do that. The result is that when it hits the ground, the business cannot scale automatically. If you want to, such as some machines particularly high load above, some machines especially low, can we hope to load more balanced, more stabilization online business, do a algorithm, such as backpack, better to do, the result is this thing done, recommends the best application is given to the machine, the application to the machine. After giving the business team a look, we do not do, because all these work to do by hand, you also give me advice every day, let alone do, every day to adjust the machine.
So first of all, you have to figure out your premise, automation, do you have the ability to automate, if not, there is no need to do too much investment in this aspect.
Data structuring is the source of intelligence
AI field are basically is to rely on violence and brute force, there may be other future direction, but the AI is basically rely on the accumulation of a large amount of data to look for a thing, so it must need to have a large amount of data accumulation, the data include a lot of things, for operations, may base level data, the data of the machine, operational changes, It also has some scenarioized data, such as if you are troubleshooting and have better structured data collection, which is very important. What is difficult about data is that at the very beginning, most companies’ operation and maintenance data are not structured enough. Structured data will not work well. Of course, there will be structured data, but the structured factors will not be good enough.
Just like Alibaba said, we are ai-oriented in the field of e-commerce. Our biggest advantage is that we constantly tell the outside world that what we have is structured commodity data, and other companies can only steal structured commodity data from us. You have to do your own analysis and make adjustments to the product structure, which is very difficult. But Alibaba is natural, everyone will help you make the structure very good. Therefore, it is the same for operation and maintenance. If you want to make more breakthroughs in intelligence, how to structure data better is a very big challenge. It’s hard to think straight. Those are the first two things I think we need to figure out.
Intelligent operation and maintenance (O&M) scenarios
At present, for operation and maintenance scenarios, there are two kinds of problems that intelligence is particularly suitable for solving, which seem to be similar for all industries. The first is scale, and the second is complexity. Scale means that I have a lot of machines, and in a lot of machines I have to find a problem with one machine, which is very difficult to solve in a traditional way because it’s so large. Or you have to put in a lot of manpower and so forth, and it’s a little bit more than it costs. How to solve the problem of scale better after scaling up, intelligence will bring some help.
The second is complexity. For example, if your application changes from one application to thousands, tens of thousands, hundreds of thousands of applications, then you have to find out which one of them is a very complicated problem. So the problem of complexity is something that is very difficult for humans to deduce with human brains, but relatively easy for machines to do. This is the direction that some teams of Ali hope to try to be intelligent. Usually, we will see whether the above prerequisites are met. If you have it all, explore it. Therefore, I say that Ali is actually in the exploration stage of the whole intelligent operation and maintenance, rather than the full development stage.
Alibaba intelligent operation and maintenance of five steps
Just to talk a little bit about the various areas that we’re in right now in the smart area, in the five areas of operations, for us, smart, some of the possibilities that we see, including what we’re doing.
one The focus of resources is cost
1. Infrastructure selection
For resources, the biggest concern of the whole company is cost. If you don’t deliver the resource at the lowest cost, this intelligence can really help a lot. The first point, for example, how to better plan the whole data center model, network, and the company, why this is, by means of the intelligent a data center location from a lot of factors, in addition to government level of policy factors, there are many other factors to consider, such as climate and so on various factors, all need to consider at this stage.
You need to analyze through the accumulation of a large amount of data, for example, in China and overseas, which places are most suitable for your business development strategy, and where is it? This needs to determine a scope, and on the basis of a scope, it is necessary to establish further people. In terms of network and models, we think the best thing to do at present is that ali’s model is different from that of some companies. More machines of Ali come from the same department, and basically the same department teaches all machines of Alibaba. That’s a huge benefit, because it’s all on the same team. For example, Alibaba started to build a unified dispatching system last year, which brings even greater benefits. Since all resources come from the same place, this place collects all the resource demands and data of alibaba, and all the data are in its hands.
If you combine this data with its actual operation, you can better deduce, for example, what is the most suitable model for Alibaba, which has been trying to do this in the last year or so. In all the processes before last year, Alibaba, for example, next year’s model of my server, the so-called model, the meaning of the model here is mainly a matter of ratio, not what kind of CPU to choose the next generation, that is determined by the development of hardware. But the ratio factor, before we are more brain beat, human intelligence. There is a certain stage in which human intelligence is of a higher order, after which it is no longer possible for a human to compete with a machine. The team said that the configuration of the model we would buy next year would be like this. After a calculation, we could shoot it. Last year we introduced a system that analyzes all the data and the money, the most important thing is the money, and then analyzes the process and deducts what’s best for us. So what exactly is the right model?
If you have a very good inference system to deduce how your model, network and IDC should be planned in the future, it will be a great help to the cost area. Take network for example, the current development, 10 gigabytes, 25 GIGABytes, 45 gigabytes, 100 gigabytes, what do you think is the most appropriate for your company? Most companies are likely to be brainstruck, but that may not be the case.
2. DC brain makes control more intelligent
DC brain, this is quite popular now, this field is very hot, the main reason for the hot may be because of an article last year by Google, Google published an article last year, there is a news that they through better intelligence, to control the intelligence of the whole machine room and so on. For example, controlling the outlet of the air conditioner, which way the wind blows, controlling that, and saving Google a lot of money, a lot of money. So a lot of data center teams are working in this area right now. Because it’s so cheap.
And then we did an analogy, and we said actually for most people, you might have a hard time feeling the data center, but the place you feel the most is the other place, your office. For example, we used to say that when alibaba comes to summer, the office is too cold, much colder than outside. If we can control the temperature better, it will be a great help for us, and it may save more money for the company. So how to do this is very important.
3. The premise of elastic expansion is to achieve automation
The r&d team said, the business team said, I want a hundred machines, you can’t argue with him, finally put a hundred machines online, you find that he is enough with ten. But it’s hard to argue with him, as if countless operations teams are trying to flex. But as I said, the biggest premise for elastic scaling is automation, and it doesn’t make much sense without automation.
4. Resource portraits make resources match better
How to better match resources, Alibaba is trying to do the portrait of resources. For all online businesses, it’s pretty predictable. Most online businesses, but a few online businesses are not. Most online businesses are a model. If the prediction is very good and the resources are reasonably matched, it will be of great help to the company’s resources.
two Reduces failures caused by changes by 30%
In this area of change we feel that the first is efficiency. Alibaba now has tens of thousands of r&d personnel, and we have transferred the operation and maintenance work to R&D. How to make r&d more efficient and insensitive in the process of change is a focus of Alibaba’s pursuit now. This focus we think, intelligence can play a huge help. The first example above is about intelligent flow control during file distribution. For example, if a release takes an hour, that means most development takes an hour. He doesn’t have to watch it all the time, but he does have to watch it after the release, which takes a lot of energy.
Another direction is now the industry is very hot unattended, how to do in the release process, for research and development is the best no sense, I set a day to send, as long as the test passed I can automatically complete the process, there is a problem with a little control, there is no problem as this thing did not happen. It’s even more helpful to have multiple r&d teams, or of course, if you have an operations team doing it, which means that a lot of operations people can take a big chunk out of the work.
So, to change the field, what we want to do is move in that direction. At present, we can see from Alibaba’s attempt that the failure rate caused by change is the highest. In this field, it can be reduced by 30% because of the failure caused by change. Interception is mainly used to intercept problems.
3. To monitor AI,
Intelligent alarm
This is the hottest area of AI coming into operations right now, all companies are doing it. The first one is ali, ali is no exception, we are doing the same. The first one is intelligence. We all know, for example, those who do operation and maintenance, that when you finish writing a business, you need to set the threshold of monitoring alarm, such as how much CPU should alarm, and then how much response time should alarm. One of the directions That Ali is trying to take is to let you not go to the match. Ali decides what situation to call the police according to the analysis, which is a huge help for research and development.
Anomaly detection directly affects efficiency
The second is anomaly detection, which many companies do. The biggest reason why anomaly detection is done is because of efficiency. If it is not done, it is ok, but it takes a lot of manpower. Let’s say the trade goes down, what is it, let’s say for us, the trade goes down, and when it goes down we need to analyze what the factors are. And it’s very possible that, in the end, it’s not us, it’s external, it’s national holidays, it’s all sorts of things. Especially small-scale business, such as our overseas business, fluctuates greatly. If a fluctuation is regarded as a problem, it will have a huge impact on the efficiency of the whole company.
So we think that if anomaly detection is done very well, it will be very helpful to our efficiency. Generally speaking, anomaly detection, operation and maintenance data are all time-series. There are various algorithms based on time series. The algorithms commonly used in the industry are listed above. The algorithm in the top left corner is the algorithm studied by Alibaba. From our current test, we can see that the accuracy of the algorithm studied by Alibaba is much higher than the industry. I won’t go into the details, but the most important reason is that this thing is going to be presented in a paper at some conference, and you’ll see it later.
Four. Stability is based on the principle of efficiency
Repair faults accurately and quickly
The most important thing about stability for us is efficiency. The first is the repair of faults. Faults occur in larger companies, larger scale and more complex business scenarios. It is inevitable that faults will occur, and the key is how to repair the faults as soon as possible. In this area of troubleshooting, Alibaba has been trying a lot of solutions for many years. Many cases, this process needs the accumulation of slowly, when failure occurs because of trust, many of the team we all said that the company is in a state of high tension, throws a system at this time, most such systems are now throw three decision, give you three Suggestions, and then you to choose. Sometimes an experienced troubleshooter will see that all three of your suggestions don’t make sense. When eight out of ten failures, not eight, if four or five failures are like this, then everyone will not look at the system, it’s too unreliable, it’s better for people to judge. The difficulty of this system is very high, and the whole company needs to move in this direction firmly, and accumulate a lot of data better.
Troubleshooting, ali now only try some very simple case, for ali, such as a room is out of order, because the entire alibaba trading system architecture is to support more, for us if in some cases, we can judge a room out of order, we can do some automatic flow switch, etc. But Ali now also believes that intelligence in the stability, especially in the repair of such actions, or very careful, in case nothing cuts a problem, this has a greater impact.
Use intelligence to locate faults
We used to think that positioning is not a big problem. If I can fix it quickly, you can fix it slowly. I don’t care if it takes two days. But the reason ali pays special attention to it now is that fault location costs us a lot of manpower and a lot of team strength. So we think there needs to be a smarter way to locate failures so that the r&d team can focus on other things. For example, now a fault out, RESEARCH and development checked along while, a look, with it all have no relationship. So they wasted a lot of, this picture is a system we are doing now, from an exception, there to mark the 12345, when there is an exception, the first step, found that the second step of analysis, position until the final exactly is which place out of the question, our goal is the final location as much as possible to the problem on the level of code, Or network or infrastructure or whatever.
Five. Do a good job of large-scale operation and maintenance while pressing
At present, the most important issue for Ali is efficiency. Such as we are ready to put A capacity every year, many people know all link pressure measurement, ali are one of the most important purpose is to adjust capacity, how to adjust the capacity of A computer into A ratio is the most appropriate, for example application is likely to be A bottleneck, but in fact if collocation good, applications will no longer be A bottleneck. So how to make the best match for a fixed machine? We used to press one time to adjust, and then press another time to adjust again, which consumed a lot of people’s energy all night. We think this process needs to be improved. Now we change it to a very simple mode. After the flow comes, the capacity proportion will be automatically adjusted continuously. I believe many operation and maintenance students have done this, because the business side gives you an index, you need to calculate, and it is difficult to calculate very accurate. Pressing and flicking means you don’t have to be precise, you can just do a rough number, and the system will automatically balance you.
The defense line to be broken in the future operation and maintenance field
Unpersonified dreams come true
I think the biggest challenge in operations is still whether you can really go unmanned, where there’s no one at all.
From the present point of view, the most important thing to achieve unmanned is the quality problem, the quality is not good enough to unmanned. In addition, if something goes wrong, can it be automatically repaired and so on? Therefore, we believe that unmanned is the biggest challenge in the field of operation and maintenance. Can we turn this into reality, laying the foundation of intelligence. If all the intelligent actions need to be involved, it is basically unnecessary to do.
Intelligence brings qualitative changes in efficiency
In terms of intelligence, the first point is the validity of the problem. If the intelligence is less than human intelligence, gradually no one will believe in it. So how to improve effectiveness, and the most important thing is to see that intelligence brings qualitative changes in efficiency to the field of operation and maintenance. Intelligence is a huge investment, a lot of collection, a lot of analysis. So it is better to bring qualitative change rather than quantitative change, if only quantitative change may not get back the investment. Fewer people and lower costs are very important for all companies. People are better off focusing on more important things like research and development.
Originally published: 2017-11-23
Author: Bi Xuan
This article is from the cloud community partner “Ali Technology”. For relevant information, you can pay attention to the wechat public account of “Ali Technology”