The founding team of Huawei's distributed storage starts a business, with revenue more than tripling last year. "Panlian Information" secures tens of millions of yuan in financing | Exclusive report by Yingke

Text by | Hu Yiting

Edited by | Yuan Silai

Yingke learned that Shenzhen Fanlian Information Technology Co., Ltd. (hereinafter referred to as "Fanlian Information"), a domestic provider of distributed storage solutions, recently completed a Pre-A+ round of financing worth tens of millions of yuan. The round was led by Xinfang Capital, followed by institutions such as Liwan Industrial Investment, and Hongbo Investment served as the exclusive financial advisor. The funds from this round will be mainly invested in the iterative development of distributed storage products, the R & D of products such as storage and computing all-in-one machines, as well as the layout and upgrade of the sales system.

Since its establishment in 2021, Fanlian Information has entered the trillion - level market of intelligent computing centers. Relying on its independently controllable technology system, the company has launched a technology ecosystem centered around the high - performance distributed file system UbiXFS. Currently, its distributed storage product matrix includes the UbiPower distributed all - flash storage system, which features extreme performance, and the UbiScale distributed hybrid - flash storage system, which balances cost - effectiveness. It has served benchmark customers such as China Mobile, Shenzhen Supercomputing Center, and Zhongbei Communication, and is also targeting overseas markets.

In terms of key technologies, UbiPower innovatively adopts a "memory - centric" architecture design, through the in - depth integration of a full user - space software stack and RDMA network technology. Jing Ning, the CEO of Fanlian Information, said that compared with the current best - in - class products, this can increase the concurrent task carrying capacity by 50% and reduce the average latency by 34%. Based on better performance, it builds an AI solution of "storage + intelligent computing".

With its microsecond - level response characteristics, in the scenario of large - model training, when an AI model with hundreds of billions of parameters undergoes high - frequency weight updates, UbiPower can ensure a stable throughput of millions of IO requests per second. It can rapidly schedule randomly selected training samples to the GPU cluster for training. At the same time, it can achieve high - speed persistent storage of models with hundreds of billions of parameters, significantly reducing read and write latency and maximizing the resource utilization rate of GPUs.

In terms of product combination applications, UbiPower and UbiScale can form a hierarchical storage solution to balance high performance and low cost. The former builds a PB - level high - performance storage layer, while the latter expands an EB - level mass storage pool. The solution presents a unified namespace to the computing side for easy use. Internally, based on an intelligent data scheduling algorithm in the high - performance storage layer, data can flow freely between the high - performance storage layer and the mass storage layer, achieving an organic combination of sub - millisecond response for hot data and EB - level storage for cold data.

The high - performance storage market is mainly targeted at intelligent computing scenarios. It is a niche market in China that currently lacks self - developed underlying technologies, has a relatively fast market growth rate, and strong customer demand for substitution. According to a relevant report released by IDC, global data will increase from 33 ZB (10 trillion trillion bytes) in 2018 to 175 ZB in 2025. The proportion of Chinese data in global data will increase from 23.4% to 27.8%, making China the world's largest data sphere.

To penetrate this market, recently Fanlian Information also launched the UbiCube 5000 series of AI servers and the UbiCube 10000 storage and computing all - in - one machine. The former can support mainstream GPU cards such as NVIDIA and Atlas; the latter features an innovative native storage architecture based on DPU, which can greatly simplify the system architecture of intelligent computing center cluster equipment and offers advantages such as ultra - low latency and ultra - high density, facilitating management and operation and maintenance. It is designed for AI industry applications and private deployment scenarios.

In terms of the team, Jing Ning, the CEO of Fanlian Information, was the first general manager of the distributed storage division at Huawei, and most of the founding members are from Huawei's storage team.

Investors' views:

Huang Chong, the founding partner of Xinfang Capital, said: "Fanlian Information has quickly carved out a niche for itself in the wave of intelligent computing center construction in China with its self - developed ultra - low latency distributed file system. Moreover, at the beginning of the year, closely following DeepSeek, it launched the DeepSeek large - model all - in - one machine in a short period, fully demonstrating its precise ability to grasp market opportunities and efficient execution ability. This has made Xinfang Capital deeply aware of Fanlian Information's profound strength and professional level in technological innovation and business implementation, and has further strengthened the confidence of both parties in cooperation for joint development and exploring the future."

A relevant person in charge of Guangzhou Baietan Fund Management Co., Ltd. said: "Against the backdrop of the digital wave sweeping across various industries, Fanlian Information provides enterprises with efficient, innovative, and practical storage solutions, demonstrating a high level of professionalism and innovation ability, and has great development potential and investment value. Led by a group of people from Huawei's storage team, represented by Jing Ning, Fanlian Information is bound to become a new - era 'unicorn' enterprise in the field of AI storage."