Company introduction

Zero-krypton Technology, as the world’s leading artificial intelligence and medical big data platform, has the largest and largest medical big data resource base and the most advantageous technical support service system in China. Over the years, zerokrypton Technology, relying on the core technical advantages in medical big data integration, processing and analysis, relying on advanced artificial intelligence technology, is committed to providing high-quality medical big data overall solutions for the society and industry, government departments, medical institutions at all levels, domestic and foreign medical device manufacturers, pharmaceutical enterprises, etc. And artificial intelligence assisted decision-making system (assist management decision-making, assist clinical research, AI intelligent diagnosis and treatment), patient process management, hospital public opinion monitoring and brand building, pharmaceutical research and development, insurance cost control and other integrated services.

Main application scenarios of LinkDoc

LinkDoc applies patients’ real case data and algorithm model to tumor treatment, constructs accurate diagnosis and treatment model and provides data support, so as to assist hospital management decision-making, scientific research and clinical diagnosis and treatment. At present, the “prediction of the risk of lymph node jumping and metastasis of lung cancer” module of the Hubble system can avoid the recurrence of lung cancer patients 8-10 months in advance due to misjudgment, and prolong the lives of nearly 20,000 patients 8-10 months every year. The “AI-Intelligent Diagnosis of Pulmonary Nodules” module of the Hubble system automatically identifies all nodules in CT images, and the recognition rate reaches 91.5%. LinkDoc hopes to improve the global medical standards of Chinese physicians by virtue of its core technical advantages in medical big data integration, processing and analysis, and its innovative research and development on Internet artificial intelligence. By supporting the development of drug research and development and medical insurance industry, LinkDoc can provide universal and accurate medical services for every patient.

The underlying database platform supporting LinkDoc’s business is also facing technical and business challenges in the new field of medical industry, such as the rapid growth of data volume (100 million level), data erasers of cleaning logic under large data volume, and the pressure of analytical things on database reading, all require us to re-explore in the database platform. Choose a database solution suitable for medical big data business.

Choose TiDB

  1. Business pain points

  • Large amount of data, complex capacity expansion operation of single-instance MySQL;

  • Large amount of data is written, and the master-slave delay is high. As services require low data delay, the traditional master-slave architecture of MySQL cannot meet the requirements under this project. A large amount of data is written into the master library, which becomes a performance bottleneck.

  • With the increasing amount of data, the speed of some statistics query is slow.

  • The development and maintenance cost of sub-database and sub-table is high.

  1. demand

  • High reliability & stability;

  • Scalability, storage & request processing capacity can be rapidly improved with the increase of data & request volume;

  • Lower latency.

  1. Project research

Before we choose TiDB, we have investigated middleware solutions such as MyCAT, Cobar and Atlas. These middleware generally make users feel “frustrated” and are unsatisfactory in terms of community support, MySQL functional compatibility and system stability, requiring a lot of business transformation. Switching costs are too high for fast-growing companies.

Under the recommendation of Wang Xiaozhe, chief architect of LinkDoc, we investigated TiDB. The following features of TiDB impressed us:

  • Compatibility with most SQL functions (meaning that businesses can be easily modified and smoothly migrated to TiDB);

  • Horizontal expansion ability;

  • Distributed transaction;

  • Quick fault recovery ability;

  • Monitoring index coverage.

Online TiDB

  1. Compatibility test

After the compatibility test, we made the following simple changes to the business:

  • Blob data is migrated to HBase as key-value storage.

  • Batch delete Deletes 1000 items in a Batch.

  1. Gray line

Due to the high demand for master/slave synchronization delay, we switched our first application with the service dual-write scheme. In the first stage of gray scale, the business simultaneously writes MySQL and TiDB, reads MySQL, and verifies data consistency. After 2 weeks of verification, we perform gray scale in the second stage. In the second stage of gray scale, TiDB and MySQL are written in pairs, and TiDB is read in pairs. After a month of business verification, we completely dropped MySQL.

  1. System architecture

There was also a small problem in the launching process. The aliyun common instance + SSD cloud disk used to run TiDB was often met with performance jitter under this configuration. Under the advice of PingCAP, we replaced aliyun local SSD model, and the system is running well now.

System configuration & architecture is as follows:

Production cluster deployment (machines based on Ali Cloud) :

Current status and next step planning

At present, TiDB has carried the two businesses with the largest amount of data in LinkDoc. Normal QPS 6K, peak 12K.

In the future, TiDB will be used to carry more business libraries with large data volumes, and TiSpark will be investigated. A unified database HTAP platform compatible with analytical and transactional database is constructed by TiDB.

To PingCAP

Thank you very much for the strong support from PingCAP partners. From hardware selection, business optimization, system training to online support, PingCAP has shown warm service attitude and professional technical ability, helping LinkDoc to successfully launch TiDB, solving system problems and supporting the rapid development of business. I believe that with the efforts of such a group of small partners, TiDB will become more and more mature, carry more business scenarios, and create miracles with technology.

Yang Hao is currently in charge of operation and maintenance & Security of Zero-Krypton Technology. He used to work in Alibaba – Technology Security Department -CDN. Focus on CDN, security, automated operation and maintenance, big data and other fields.