ZTE Assists China CITIC Bank in Building a Data Bank

OTT TV Becomes Essential for Operator Development

China CITIC Bank was one of the first commercial banks established during China’s reform and opening up. After nearly three decades, the bank is one of the best capitalized in China and is rapidly growing.

Cloud computing and big data have created new opportunities for innovation and service modes in the financial industry. However, in the banking industry, the requirements for security, stability and timeliness put added requirements on cloud computing and big data. In recent years, internet companies and terminal vendors have been changing traditional ways of thinking. WeChat, Weibo, and third-party payment services are already in full swing and are impacting traditional industries, including the financial industry. Combining internet and finance is a growing trend.

China CITIC Bank proposed to rebuild their online banking by launching internet-based financial strategies. CITIC will create new operation and profit models through  payment, data mining, and financial management based on big data and other new technologies. CITIC established an IT team to research and discuss the data bank project. They decided to build an internet-oriented big-data platform that could carry major existing and future banking services.

The data bank deployment project was initiated in 2013. The vision of CITIC is to implement a new big-data platform with basic cloud computing architecture that will drive service development. After eight months of technical exchange and proof-of-concept testing for the big-data platform, ZTE won the bid for the project in March 2014.

Sophisticated Architecture for Stable System Running

China CITIC Bank’s big-data platform runs on the X86 server and is capable of high performance and mass storage through the software system. The platform is highly reliable, expandable, efficient, and fault-tolerant. It reduces hardware fault rate through redundant, distributed data and services; allocates data between available PC server clusters that can be expanded to thousands of nodes to fulfill computing tasks, and moves rapidly data between nodes while keeping dynamic balance among them. It can also automatically save multiple copies of data and re-allocate failed tasks.

The big-data platform comprises hardware resource layer, Hadoop big-data processing software layer, and service application layer. At the hardware resource layer, unified x86 storage servers are deployed for the cloud storage system. The Hadoop big-data processing software layer comprises software resources used for the cloud storage system, such as the unstructured data storage engine HDFS, parallel computing engine MapReduce, NoSQL data storage engine HBase, and structured data storage engine HIVE. These engines manage user and data access and provide services through the HDFS, CMD Line, REST, MR, FTP, JDBC/HQL, and NoSQL interfaces. The service application layer stores and processes WAP gateway logs, click-streams, CDRs, and signaling data.

Hadoop and HBase Based Big-Data Platform

In traditional IT deployment modes, each service platform that comprises server, storage, and network resources is constructed independently. Repetitive deployment of these resources on different service platforms leads to high capex. To meet service requirements, different hardware devices are used. This greatly increases opex. Unbalanced load between different servers also leads to low hardware utilization, and many computing, storage, and network devices are idle. This increases equipment room rental, power consumption, and cooling costs. Moreover, a long cycle for deploying new services also leads to high opex.

Internationally recognized, open-source Hadoop and HBase are used to handle mass data. ZTE has built a big-data platform based on Hadoop and HBase that has distinct advantages over traditional platforms.

●    Concurrent Read/Write and High Performance

The big-data platform has optimized algorithms based on the underlying Hadoop layer. This improves application efficiency and enables balanced computing and storage. The distributed Hadoop architecture is used to provide highly concurrent services through multiple nodes. The platform also has the same features as the NoSQL data platform. These features include mass storage, linear expansion, and highly concurrent read/write. The platform is low cost and can easily exchange data with other components in the Hadoop ecosystem.

●    Efficient and Secure Data Isolation

The big-data platform has an enhanced Kerbos security mechanism that allows only authorized users to access data and services. Data is also isolated at the server to ensure different users have different data access rights.

●    Load Balance and Data Cache Mechanism

The big-data platform provides REST services and sends user requests evenly to multiple REST services through the built-in load-balancing service. The load-balancing service provides an effective data cache mechanism that can use the cache to directly return the requests with duplicate content instead of submitting many times the same computing requests to the Hadoop cluster.

●    Dynamic Online Expansion

The big-data storage platform serves as a centralized data storage platform at the back end and can provide dynamic expandability to meet the requirements of front-end systems for continuous expansion and expandable storage. The platform also supports dynamic online expansion. Storage and CPU capacity can be increased without interrupting existing services. This avoids the risks inherent in traditional expansion, and storage capacity and system throughput increases almost linearly. 

●    Rapid Deployment and Easy Expansion

The big-data platform provides an automatic system installation program that can calculate configuration parameters automatically and is applicable for most scenarios. It also provides efficient OAM with high automation and intelligence. Multiple versions for maintenance are supported, and online capacity expansion can be completed in several minutes.

●    High Availability and Reliability

The big-data platform is a distributed system that involves several peers processing nodes for computation redundancy. The platform provides high availability and reliability. Multiple copies of data are stored. When a node fails, other nodes can automatically take over its functions without affecting existing services.

●    High Cost-Performance Ratio

The big-data platform uses low-cost universal storage servers. Full hardware redundancy is implemented through software. This improves system reliability and reduces storage costs. The platform is highly automated and easy to maintain. Mass data on the order of petabytes can be stored. The platform can also reduce computing load according to dynamic traffic load and put idle disks to sleep to reduce power consumption and extend equipment life.