GBase 8a MPP Cluster V9 effortlessly handles multi-tenancy demands.
As enterprise IT evolves, data volumes are growing at an unprecedented pace. The trend toward data centralization has gained widespread acceptance in sectors like finance, telecommunications, and energy, and many large enterprises have begun building their own data centers. These enterprises require their big data platforms or data warehouses to support multi-tenancy, enabling provincial branches to access resources on demand. The GBase 8a MPP Cluster V9 cluster perfectly fulfills these requirements through its virtual cluster and resource management technologies.
First, let’s look at the GBase 8a MPP Cluster V9 virtual cluster technology. A virtual cluster is a resource isolation mechanism that allows logical sub-clusters to be independently planned and deployed, and their cluster size and computing resources to be scaled independently. All logical sub-clusters share a single access point, a unified metadata view, unified resource management, unified execution scheduling, and unified authentication and authorization. Most importantly, virtual clusters enable seamless data exchange between sub-clusters, making it extremely easy for headquarters to distribute shared data to provincial branches.
Figure: Virtual Cluster Multi-Tenant Application Diagram
Looking at this through a practical use case: the multiple logical sub-clusters correspond to multiple provincial branches. Each branch can independently deploy, start, stop, and scale its sub-cluster. For branch users, using a logical sub-cluster feels exactly like using a standalone cluster. From headquarters’ perspective, the physical and data resources of all branches are centrally managed, utilized, accessed, and maintained, while still enabling on-demand allocation, physical data isolation, and fault isolation.
Beyond the scenarios described above, the GBase 8a MPP Cluster V9 also fully supports multi-data center deployments across cities. For example, if a bank has data centers in Beijing, Shanghai, and Xi’an, the management nodes responsible for scheduling services can be deployed across all three. In the simplest setup, deploy three nodes in Beijing, three in Shanghai, and three in Xi’an. These nine management nodes form a complete scheduling cluster that collectively serves requests. Similarly, logical sub-clusters can be deployed locally in each location, allowing the cluster to process requests close to each branch and deliver wide-area load balancing. Combined with the GBase 8a MPP Cluster V9 mirror cluster feature, high availability for massive cross-data-center data warehouses becomes effortless.
Next, let’s examine the resource management technology. The virtual cluster technology described above partitions and manages resources for provincial branches at the enterprise architecture level. To fully implement multi-tenancy, resource management needs to further refine the physical resources inside logical sub-clusters. Within the allocated physical machines, each branch can divide resources such as CPU, memory, storage, network, and priority into resource pools according to user demand. These pools are then bound to users, enabling physical resource isolation and dynamic control. In this way, provincial branches meet the headquarters’ unified management requirements while retaining the flexibility to allocate and use their assigned resources, smoothing the transition triggered by data centralization.
Figure: Resource Partitioning Diagram
GBase 8a MPP Cluster V9 also implements the Open Service Broker standard protocol. Through cloud and PaaS platforms, GBase 8a can effectively manage the relationship between resources and resource consumers in a multi-tenant model, enabling open resource sharing while meeting the dynamic business demands of resource users.