Avoid blindly following the trend, Wanli CTO Luo Shilong will help you restore the essence of new technology

[51CTO.com original article] "Cloud computing, big data, artificial intelligence, blockchain..." These words have been hot in 2018. Driven by application demand, the development of new technologies has been accelerated, giving rise to more new industries and new formats. We have seen that these new technologies have been warmly welcomed by Chinese companies, quickly popularized, and then become a hot spot; but on the other hand, blindly following the trend and slow implementation of technology are also problems we are facing now.

So, in the face of increasingly complex and diverse new technologies, how do Internet companies at the forefront of the industry respond? As an Internet company, Wannianli CTO Luo Shilong shared Wannianli's journey to the cloud and its attitude towards new technologies such as artificial intelligence and blockchain in an interview with reporters.

[[258477]]

Perpetual Calendar CTO Luo Shilong

Facing traffic peaks, Wannianli's journey to the cloud

As an Internet company, the biggest challenges faced by IT architecture are high traffic and high concurrency. During the peak period of user usage, Wannianli has to process more than 5,000 requests per second. How to use technical means to enable the system to process more requests in parallel, reduce latency and response time, and improve performance and user experience has become a difficult problem that every company must solve.

Wannianli initially built its own computer room. In the face of traffic peaks, it increased the bandwidth of the computer room, added resources, and added servers. However, in the later stages of development, this simple and crude approach could not meet the needs of the rapidly growing user base, so Wannianli migrated all servers to the public cloud platform.

On the one hand, by expanding capacity in multiples, a programmatic automatic judgment can be made based on the existing traffic. If the number exceeds the capacity of the existing operating equipment or backend server, the computing resources will be analyzed to achieve automatic expansion;

On the other hand, with differential requests, each request does not directly return all the data, but compares the newness of the data. If the data has not changed, it will be returned directly, thus shortening the request time and reducing the number of requests.

In addition, differential requests are implemented. Each time a user starts a client, many interfaces are called, but the order or time of interface calls is not fixed. In this way, the concentration of requests will be dispersed to a great extent, which will bring a very obvious concentration effect during user peak hours.

As for the choice between building a private cloud and using a public cloud, Luo Shilong believes that if the cost is taken into consideration, building a private cloud is not recommended. The main reason is that scale generates profits. For public cloud service providers such as Alibaba Cloud, Tencent Cloud, and AWS, they can centrally purchase a large number of servers and bandwidths. The cost allocated to each unit will definitely be much lower than the price of purchasing a small number of servers and bandwidths by themselves. Even if the profit of the platform is added, it is still lower than the cost of building a computer room by the enterprise.

Currently, Wannianli is also considering a multi-cloud disaster recovery solution, with one public cloud platform as the main battlefield and several others as disaster recovery, to avoid being completely dependent on a single cloud service provider. Regarding data security issues, Luo Shilong believes that if cloud service providers can retrieve or read customer data at will, it will be a devastating blow to the entire public cloud industry. Therefore, as a customer, they are more inclined to trust the professional ethics and contractual spirit of the cloud platform.

[[258478]]

Restore the essence of new technologies and avoid blindly following trends

Since last year, two words have become more and more well-known and have become synonymous with the current state of technology: "artificial intelligence" and "blockchain". How to view new technologies, Luo Shilong pointed out that we should see more of the essence of these technologies and avoid blindly following the trend. For example, what value can blockchain bring to existing businesses? It is not just buying a few servers, making a few nodes, and putting some content on them to make a blockchain project.

In fact, the essence of blockchain is distributed storage. When it comes to specific applications, it needs to return to the essence of distributed storage, such as its advantages of not relying on a single node, being tamper-proof, and having stronger fault tolerance and disaster tolerance than data centers. Then, when enterprises combine new technologies with their own businesses, they need to consider whether there is a real point of integration. The existing data of the perpetual calendar is stored in the main database. After thinking and verification, it is believed that there is no need to use blockchain technology to make a distributed database, and it will not be of essential help to the entire business.

In terms of AI, AI and big data complement each other. The essence of AI is statistics. First, a statistical model is established based on past behavioral data, and a large amount of sample data is used to train the model. After entering new parameters, the model outputs the prediction results. Through AI technology, the most direct application in the perpetual calendar business system is the recommendation of information and advertisements. It can also perform image recognition and voice recognition. These applications are all based on information distribution and circulation realized by feature statistics.

AI also plays a vital role in the construction of the entire commercial BI system of the Perpetual Calendar and the distribution of content. The Perpetual Calendar will record all user behavior operations in the central database, including all click behaviors and content tags of each user in the entire App system, which will be uploaded to the statistical server. After cleaning and calculating these behavioral data, the Perpetual Calendar will build a portrait of each user, and at the same time derive the user's contribution to the product throughout the life cycle, such as whether the user has paid, which ads have been clicked at which locations, and other data. In this way, it can be calculated whether the promotion channel is worthwhile for product promotion.

At the same time, for self-built advertising platforms, the Perpetual Calendar can also push information based on user portraits and tags to help improve the efficiency of advertising and traffic utilization. In the past, the Perpetual Calendar could only be fully delivered, and one ad slot could only receive one advertiser's order in a unit of time. Now, more refined delivery can be carried out based on different user preferences. Therefore, in the same location and at the same time, the Perpetual Calendar can serve more advertisers and provide users with the information they need. Based on these two aspects, the Perpetual Calendar improves the efficiency of content distribution through AI technology.

[51CTO original article, please indicate the original author and source as 51CTO.com when reprinting on partner sites]