Search
  • Huawei’s FusionStorage Helps China Merchants Bank Respond to Elastic Demands

Huawei’s FusionStorage distributed cloud storage delivers superior performance, achieving linear performance growth as the number of servers increases. Currently, FusionStorage provides higher reliability and performance than traditional storage. In the future, we intend to use FusionStorage at a larger scale and gradually expand the application of FusionStorage in non-critical services. We will use distributed storage to first move storage systems, and later the whole IT infrastructure, to the cloud.

Background

Banks at a Crossroads in the Digital Era

China Merchants Bank (CMB) has positioned itself as a lightweight fintech bank. To better serve its customers, CMB is poised to transform itself for the Bank 3.0 era. In contrast to Bank eras 1.0 and 2.0, represented by a wide proliferation of bricks-and-mortar bank branches and e-Banking services respectively, the Bank 3.0 paradigm requires building a fully mobile real-time services environment that can provide consumer-centric virtual services without relying on any physical infrastructure. Using mobile banking as a starting point, CMB aims to provide customers with more intelligent information services and real-time marketing decisions through big data. By running IT infrastructure on the cloud, it hopes to construct a more agile development and testing system to meet customers’ needs for diverse applications through various channels and terminals. To support the company’s strategic transformation, the IT department has identified the cloud and big data as two important development directions. It aims to reconstruct IT infrastructure and customer experience through cutting-edge cloud computing and big data technologies.

Challenges

To accommodate the rapid development of Internet financing services, CMB is strengthening conventional banking operations based on bricks-and-mortar branches while also developing new services, such as mobile payment and Internet finance, which help expand the end user base. However, the traditional siloed IT storage system and data processing platform work smoothly only when branch business makes up the bulk of systems requests, and e-Banking services account for a relatively small proportion of requests. In an Internet financing environment dominated by e-Banking services, traditional storage systems are inadequate due to rigid resource provisioning and limited performance expansion when handling hundreds of millions of active users. The current system can no longer meet the demands in the coming Bank 3.0 era.

Solution

Ongoing Innovation and Transformation

Huawei is helping CMB plan a next-generation IaaS framework on which it will build an open and flexible cloud platform. This will enable on-demand resource allocation and fast service deployment. The IaaS platform incorporates products from multiple vendors to avoid vendor lock-in. This requires a storage resource pool with wide scalability and compatibility.

CMB wants to apply storage resource pools to development and testing, channel access, report, and database systems as well as big data platforms. The demands of access modes, data read and write models, as well as access protocols for storage systems vary dramatically in different service scenarios. For example, a development and testing system requires mid-level access latency that can already be achieved by traditional mid-range storage arrays. Bank channels are classified into three categories: 

  1. Branch channels that include ATMs, tellers, and self-service banking systems 
  2. Electronic channels including online and mobile banking 
  3. Innovative channels composed of third-party platforms and banks’ proprietary applications 

The access performance demands of channel systems change as the number of active users grows. In channels accessed by tens of millions of users, a storage system needs to provide 30,000 IOPS, and hundreds of thousands of IOPS are required for hundreds of millions of users. The database and report services have higher requirements on IOPS and latency. The varied storage performance demands and the coexistence of virtual and physical platforms pose huge challenges to traditional storage systems.

To address these challenges, Huawei used FusionStorage distributed cloud storage to build a new storage resource pool that was widely compatible with various hypervisor devices. In addition to the existing VMware vSphere on the customer’s live network, FusionStorage supports Huawei’s FusionSphere cloud environment, preventing vendor lock-in. FusionStorage differs from traditional siloed architecture, in that the storage services for development and testing, channel access, and database applications provided by FusionStorage use storage pools. It virtualizes hundreds of servers into a scale-out storage pool that expands linearly as the server quantity increases. Each node in the storage system boasts both computing and storage capabilities. FusionStorage adopts a two-layer read cache mechanism (Layer-1: memory cache; Layer-2: SSD cache), which helps reduce data access times. A single node carries 30 percent more applications than peer products with the same hardware configuration. Under large-scale concurrent service models, FusionStorage provides better performance. For example, it improves the report analysis efficiency by 30 percent for CMB. The current cloud resource pool constructed by Huawei and CMB is able to withstand concurrent access by 200 million users.

Transformation

Huawei’s FusionStorage distributed cloud storage has enabled CMB to rapidly eliminate the performance bottlenecks of traditional storage systems. By combining the back-end cloud infrastructure with a big data application platform, FusionStorage facilitated fast service deployment and expansion, along with accelerated product development. FusionStorage was deployed in development and testing systems, followed by report, channel access, database, and big data platforms to simplify O&M for the bank. This helped expedite storage resource provisioning and elevate storage utilization at 40 percent lower TCO. This also enabled CMB to expand its entire IT infrastructure to the cloud, to bolster its financial IT strategy of building a networked, digital, intelligent bank.

Coping With Elastic Performance Demands of Frequently Accessed Services

Upgrading storage systems is the first step towards business transformation. The traditional siloed IT storage architecture and data processing platform of China Merchants Bank can no longer meet the ever-increasing demands of e-banking services, imposing huge challenges on storage systems.

For example, a few years ago, when the number of mobile banking users exceeded 30 million, the number of WeChat red envelope users exceeded 200 million. The WeChat mobile App is ubiquitous in China, and the WeChat red envelope is a feature based on the Chinese tradition of giving families or friends money as a gift. On the Double 11 shopping day in 2017, the number of mobile banking transactions exceeded 10 million, while the number of WeChat red envelope transactions reached 1 billion. This is 1,000 times more transactions than mobile banking.

Storage systems must be flexibly configured to cope with ultra-large-scale, concurrent, and unpredictable service access, responding to thousands of billions of requests. In a traditional architecture, a RAID group is created first, after which LUNs are allocated to provide services for external systems. When the overall system performance is relatively sufficient, but a RAID group cannot meet the performance requirements, then multiple RAID groups and LUNs need to be adjusted. Services have different performance requirements at different time points. During resource planning, the resources of a storage system are usually configured to meet peak performance demand, but in most cases this level of performance is required for only a short period of time, leaving a large number of resources idle most of the time. Frequent adjustments in system configuration bring O&M risks. CMB has found that traditional storage systems can no longer adapt to the impact of Internet financing services.

For services that see explosive growth in performance demand in a certain period of time, FusionStorage uses refined resource allocation and QoS control to flexibly allocate resources specific to application loads. It adopts a special resource supply mode, to allowing the application to obtain more resources in specific scenarios. The IOPS and bandwidth resources of a storage pool can be properly allocated to applications with different priorities, and small-capacity EVS disks can be temporarily allocated to provide high performance for short periods of time. This enables meeting sudden, short-lived spikes in demand. After hardware capacity expansion, both the capacity and performance of the upper-layer application resource pool automatically increase. Also, FusionStorage simplifies resource management and prevents frequent adjustments of RAID and LUN levels. Resource provisioning and reclamation can be completed within minutes, facilitating agile development. It has been proven in practice that FusionStorage improves the resource provisioning efficiency by more than 10 times compared with traditional storage.

Sustainable Distributed Cloud Storage

Cloud computing has seen increasingly broad adoption in CMB’s data centers. Based on the rapid VM provisioning function, FusionStorage allows users to create, provision, and mount VMs in batches. By improving the provisioning efficiency of storage resources, FusionStorage shortens the TTM of new applications. FusionStorage supports flexible resource expansion. It supports scale-out of disks, nodes, cabinets, or resource pools. Also, it enables volume- or node-level scale-down to support flexible infrastructure changes driven by application demands, and application-transparent replacement of hardware life cycles that do not require data migration.

The core difference between FusionStorage and traditional storage lies in expansion capability. After FusionStorage is deployed on the cloud data center, all hardware resources in the center can be allocated to any service. No additional performance planning or adjustment is required when the overall system performance and capacity meet the requirements. This minimizes the risk of system breakdown posed by frequent configurations. According to the system evaluation carried out by CMB, when using traditional storage to build cloud resource pools, they must buy higher-level, larger-capacity, next-generation products when the hardware warranty expires. Also, the data migration service fee may reach 1 million RMB (US$145,727). In contrast, with FusionStorage solutions CMB only needs to replace the hardware that has expired with data automatically synchronized between legacy and new devices. FusionStorage supports service-unaware device replacement in batches and automated data synchronization. This reduces data migration costs and prevent service interruption.

Moreover, FusionStorage simplifies storage management by using a management platform to manage distributed clusters in a unified manner. This also simplifies O&M, improves efficiency, and ensures optimal resource configuration, to reduce the TCO by 40 percent within the warranty period. Huawei’s cloud storage enables CMB to achieve fast service deployment and expansion at a short TTM.

China Merchants Bank has spoken highly of Huawei’s FusionStorage, “Huawei’s FusionStorage distributed cloud storage delivers superior performance, achieving linear performance growth as the number of servers increases. Currently, FusionStorage provides higher reliability and performance than traditional storage. In the future, we intend to use FusionStorage at a larger scale and gradually expand the application of FusionStorage in non-critical services. We will use distributed storage to first move storage systems, and later the whole IT infrastructure, to the cloud.”

Benefits
  • Simultaneous access by hundreds of millions of users
  • Scale-out of 1,000s of nodes with linear performance growth
  • 10x higher efficiency in storage resource provisioning
  • Fast service deployment and short TTM
  • 40 percent lower TCO
  • Flexible, on-demand storage use with simple storage management
  • TOP