Data Stays, Compute Moves: GBase 8a Cloud Data Warehouse Solves the 'Sharing vs. Security' Dilemma

Have you ever faced this dilemma: you want to tap into data value but fear data breaches; you want to share and collaborate on data but worry about crossing compliance red lines. Data is either locked in a safe gathering dust, or it "travels naked" during migration—this is a persistent headache for nearly every data-sensitive industry.

GBase 8a Cloud Data Warehouse (GCDW) offers a perfect solution: data stays, compute moves. Let the raw data rest securely on-premises while compute power comes to it proactively, extracting maximum data value while safeguarding privacy. In this article, we will unpack this innovative technology of "data lying flat, computing doing the work".

Data Stays, Compute Moves: A Counter-Traditional Approach

What is the traditional data processing model?

To analyze data, you must move it from point A to point B, clean, process, compute, and then either move it back or make a copy. This model of "data chasing compute" may work when data volumes are small, but once they reach petabytes, the cost of migration becomes astronomical—and more critically, each move doubles the risk of leakage.

"Data stays, compute moves" flips this entirely: raw data remains securely on-premises, compute tasks are broken into small units and distributed to nodes where the data resides, and only the final results are aggregated back.The data never leaves its home, the job gets done, compliance is maintained, and data silos are bridged.Industries extremely sensitive to privacy, such as finance, healthcare, and government affairs, have embraced this approach as the gold standard.

Three Key Pillars of GBase 8a Cloud Data Warehouse

GBase 8a Cloud Data Warehouse uses a three-step approach to embed the "data stays, compute moves" philosophy into its DNA from the architecture level.

Storage-Compute Separation: Letting Data and Compute Stay in Their Own Spaces

In traditional data warehouses, storage and compute are tightly coupled—if you want to scale compute, you must also scale storage, forcing data to move and driving up costs rapidly. GBase 8a Cloud Data Warehouse adopts a storage-compute separation architecture: storage is independent from compute, both evolve independently without interference.

Raw data is pinned to dedicated storage nodes, rock-solid. Compute nodes act like "shared bicycles"—available on demand, scaled elastically: during traffic peaks, new compute nodes are spun up in seconds; during lulls, compute resources are automatically released. Compute power proactively "comes to the data", so data no longer needs to be shuffled around. This approach can save over 30% in resource costs and minimize data leakage risks to the utmost.

MPP Engine: Computing On-Site

If data stays put, how does compute move? It relies on the MPP (Massively Parallel Processing) distributed computing engine. The system breaks a large query into countless small tasks, dispatches them to each node holding data, and has each node work locally, returning only the aggregated results.

What does this mean? When querying petabytes of data, there's no need to wait for data to transfer slowly; thousands of nodes compute in parallel, response times drop to seconds, with performance 10 to 100 times better than traditional architectures. Moreover, only results are transmitted, not raw data, greatly reducing bandwidth consumption.

Privacy Computing: Making Data Usable but Not Visible

Some scenarios are more complex—for example, several banks want to jointly build risk control models but do not wish to share customer data. GBase 8a Cloud Data Warehouse supports privacy computing technologies such as federated learning and private set intersection, enabling multiple parties to collaborate on computations without exchanging raw data.

Each party's data remains securely in-house; only encrypted model parameters or computation results are exchanged. In the end, the model is built, risk control capabilities are enhanced, and no raw data was ever exposed. This is true "usable but not visible".

Proven in Practice: Finance, Insurance, and Government Have Adopted It

This philosophy is not just theoretical. A major state-owned bank and GBASE established a joint innovation lab, using GBase 8a Cloud Data Warehouse to upgrade core systems. With nearly 1,000 nodes, it supports key scenarios such as risk data marts, regulatory reporting, and financial and accounting platforms. Raw data remains on the bank's premises, compute nodes are dynamically orchestrated, and resource provisioning has been shortened from weeks to hours. The fact that the compliance-heavy financial sector can use it with confidence proves the product's robustness.

PICC Life Insurance used a GBase 8a cluster to build a data warehouse, with nearly 100 nodes handling over 100 TB of data. Branch BI batch processing performance improved by over 70%, and overnight batch jobs now complete in under 5 hours. Data value is fully unleashed, and efficiency soars.

Additionally, government, telecommunications, and power sectors—industries that hold massive amounts of sensitive data—are also using GBase 8a Cloud Data Warehouse to uphold compliance and awaken dormant data assets.

Not Just Secure, but Cost-Effective

Beyond privacy protection and computing efficiency, GBase 8a Cloud Data Warehouse has two hidden bonus skills: cost saving and space saving. Its columnar compression ratio can reach up to 1:30, meaning a dataset requiring 30 TB can be compressed to just 1 TB, slashing storage and hardware investments. At the same time, it supports scaling to 4,096 nodes, covering everything from small setups to ultra-large scale.

With these robust capabilities, GBase 8a Cloud Data Warehouse was selected for the '2025 China Cloud Ecosystem Typical Application Case Collection', becoming a benchmark product for the implementation of "data stays, compute moves".

With the Data Security Law and the Personal Information Protection Law taking effect, 'data stays, compute moves' is no longer a technical option but a compliance necessity. GBase 8a Cloud Data Warehouse, with its combination of storage-compute separation, MPP engine, and privacy computing, enables data to be fully exploited even while 'resting', maintaining privacy boundaries while bridging data silos.

Looking ahead, GBase databases will continue to focus on this path, bringing the vision of "usable but not visible" data into more industry practices — so that organizations no longer have to choose between security and value.