Guangfa Securities Co., Ltd. was established in September 1991 and is one of the first comprehensive securities firms in China. Since 1994, the company has consistently ranked among the top ten securities firms in the country. With a correct development strategy, standardized management, and a sound risk control mechanism, it has developed into one of the most influential securities companies in the market.

With the explosive increase in online business volume, in order to prevent financial risks such as illegal operations and money laundering from the source, financial regulatory agencies have increasingly strict regulatory requirements for recording (audio and video), original vouchers, account opening materials, and other aspects of the business process. This is a great challenge for the information systems of various securities firms.

In addition, in daily securities business, Guangfa Securities' business system gradually adopts AI technology to improve business processing efficiency and reduce labor costs. Unstructured data such as customer identification documents, remote account opening videos, scanned contracts, customer service center voice, and enterprise related information grow by 40-50% annually.

More than 80% of the data generated by securities companies every day is unstructured data. With continuous data growth, securities services are developing towards intelligence, agility, and value, posing challenges to backend storage.

The driving force of change: upgrading securities informatization

Unstructured data, mainly consisting of images, pictures, and scanned documents, has surged rapidly, with data volume jumping from TB level to PB level. Every year, a large number of applications are added, increasing complexity and posing new challenges to the access performance of massive files.

The internal business of the institution is complex and there are data barriers, requiring the establishment of a unified data platform to improve cross business sharing efficiency. Technological evolution drives the containerization deployment of applications, which requires storage absorption container technology to provide persistent storage services. Utilizing the value of big data and leveraging AI to achieve business innovation, such as AIOps, to further reduce costs and improve efficiency is also a trend. If the storage itself can provide AI capabilities to support intelligent operation and maintenance, it is naturally a icing on the cake.

G Securities Company (hereinafter referred to as "G Securities") has chosen distributed storage, opening up a new exploration of intensive digital archive management.

Urgent change: Long term pain points drive new solutions to be implemented

From an external perspective, before 2014, the securities industry used a decentralized management approach for customer files, which posed multiple risks and put pressure on frontline employees, hindering business operations. From an internal demand perspective, business data mainly consists of small files and is growing rapidly; In addition, the "dual recording" standard requires files to be stored for 20 years and accessed at any time, resulting in severely inadequate storage performance.

G Securities is facing the following pain points: NAS storage for 70 million files, poor second level latency performance; The bottleneck of NAS is prominent, making expansion difficult; Multi site management is inconvenient, and efficient collaboration between primary and backup centers is difficult; Insufficient bandwidth during peak business hours and inability to allocate resources reasonably; Slow reading of large files, lagging video playback, and poor user experience; The efficiency of manual operation and maintenance is low, and building an intelligent operation and maintenance platform is essential. All of this has prompted G Securities to seek new storage solutions in the market.

Numerous proposals competing: Why did G Securities choose SandStone?

Among the alternative solutions, one is the traditional storage solution, which uses high-end NAS to meet the demand, but frequent expansion and performance bottlenecks in the future are still headache inducing; Another type is distributed storage solutions, which have natural advantages in capacity, performance, scalability, and other aspects.

Regarding the aforementioned pain points, the selection of solutions should focus on PB level capacity and flexible expansion; Performance without bottlenecks, high access performance; Simple operation and significant benefits; The testing docking is smooth and the business is launched quickly; Timely after-sales service, providing original factory support.

Among competitors, there are well-known traditional storage companies and emerging storage startups. How did Sugiyama ultimately win recognition from G Securities?

Product and solution improvement

SandStone MOS massive object storage supports millisecond level access to 500 million files. Unique AI processing engine helps optimize quality inspection algorithms. The business that has been integrated into MOS does not require the development of metadata migration function. Both metadata retrieval and master data reading are completed internally within MOS, significantly improving efficiency.

SandStone's three-layer data protection solution also gives customers peace of mind: by using replica and erasure mechanisms, data reliability is improved; Replace external backup software and hardware with built-in backup function; Ensuring business continuity through cross data center disaster recovery.

Excellent test results

Customers are mainly concerned about whether MOS can smoothly integrate with existing businesses, as well as product reliability and stability. The S3 interface of MOS passed the interface test smoothly, and the customer completed the single system docking in just one hour. Over the past month, Sugiyama has successfully passed all the abnormal tests.

Thorough implementation of services

The customer only made minor modifications to the original business and seamlessly migrated the original data to MOS. To avoid affecting securities trading, the project implementation is carried out from 18:00 to 06:00 the next day, providing attentive service.

4. Massive, intelligent, and agile, SandStone provides a new choice for archive management

Starting from the macro architecture of the archive center, as shown in Figure 1, MOS stores the data in the source data area and interacts with the computing layer. By deploying dual active deployment in the same city, achieve RPO minute level disaster recovery.

Specifically, the implementation of the solution is shown in Figure 2.

(1) MOS replaces NAS to ensure millisecond level access; Using small file merging to improve space utilization;

(2) Based on distributed architecture, flexible expansion, linear growth in capacity and performance;

(3) Cross data center dual active mode, multiple data centers can read and write simultaneously, ensuring high continuity and high availability;

(4) Reasonably allocate bandwidth and priority through intelligent QoS to avoid resource preemption;

(5) Business outlets access storage through HTTPS to improve file reading efficiency;

(6) MOS alarm information is connected to G Securities alarm management platform through SNMP to achieve refined supervision.

In June 2017, MOS was officially launched, deploying 12 servers, 4 switches, and over 100 hard drives to build a 1PB storage cluster.

Business data is gradually migrating, MOS supports millisecond level access to 500 million files

After more than half a year of stable operation, G Securities has gradually integrated its one-stop counters and investment research platforms into MOS. Due to data growth and insufficient original space, the customer conducted its first expansion in 2019. By creating a new EC fault domain, the space utilization rate has been increased from 50% to 67%, with significant cost-effectiveness.

As of March 2020, the raw capacity of the storage cluster exceeded 800TB, with a total of over 560 million objects and more than 70000 files written per day. The latency was less than 100 milliseconds, and the performance was unmatched by NAS.

MOS interfaces with the logging system to support intelligent operation and maintenance with intelligent storage

To build an intelligent operations and maintenance (AIOps) platform, migrate the logging system to MOS. Based on distributed architecture to ensure high availability and reduce the impact of hardware changes; Empowering AIOps with MOS characteristics, such as resource optimization through intelligent QoS; Utilize AI to optimize operation and maintenance strategies.

MOS integrated intelligent data processing engine, fully unleashing the value of data

MOS's unique intelligent data processing engine helps optimize quality inspection algorithms, unleash data value, and achieve one-stop management of data storage, processing, disaster recovery, and value discovery.

MOS has been running steadily without any malfunctions, earning praise from customers.

5. Significant improvement in archive management capabilities, helping customers achieve greater value

MOS helps G Securities build a centralized digital archive management platform, bringing significant improvements. Customers can handle business at any branch across the country, and all branches can share and use digital files, making it more convenient to access and borrow customer information online. The archive center can conduct content checks on double recorded videos, such as whether the face is clear and whether there is a company logo. If the system review fails, manual review will be conducted, which is highly efficient.

The economic benefits of the new plan are also very obvious. The price of MOS per TB is about 50% lower than that of high-end NAS, eliminating the hardware facilities required for document format conversion clusters and watermark clusters, saving nearly 100000 yuan in costs.

The future direction of architecture evolution is to combine MOS's intelligent data processing engine to unleash the value of data. For example, using AI to optimize quality inspection algorithms; Simplify data structures and improve data quality through format conversion tools; Empowering operations with AI to reduce costs and improve efficiency. Second, embrace Internet finance, accelerate container deployment, and improve microservice capabilities.

The successful practice of G Securities' archive management platform has verified the high quality of SandStone's products and solutions, and laid a solid foundation for future industry practices.