GBase 8a MPP Cluster V9 vs. Mainstream MPP Analytical Databases

GBase 8a MPP Cluster V9, featuring its second-generation federated architecture, has been deployed at over a hundred customers, powerfully proving its strengths in architectural advancement, cluster scale, high availability, scalability, and security.
 

GBase 8a MPP Cluster vs. Greenplum

Architecture

Cluster Architecture: GBase 8a MPP adopts a federated architecture where the primary cluster can scale out, eliminating single points of failure and bottlenecks. GP's master node cannot scale out, making it prone to single points of failure and bottlenecks.

Node-Level Architecture: GBase 8a MPP is based on a multi-threaded architecture that fully leverages the multi-core power of a single server. GP is based on a multi-process architecture, making it difficult to achieve maximum parallel processing capability on multi-core CPUs in a single node.

Core Features

Data Storage: Both employ columnar and hybrid row-column storage and have comparable capabilities.

DML Operations: The ability to update compressed data is comparable.

Backup: GBase 8a MPP Cluster supports both full and incremental backups. GP only supports full backups, with no incremental backup capability.

Resource Isolation: GBase 8a MPP Cluster supports resource management and priority control, enabling multi-tenancy, which is comparable to GP.

Concurrency: GBase 8a MPP Cluster achieves read-write concurrency by separating work scopes. GP uses MVCC for concurrency. Each approach has its own advantages.

Data Distribution: GBase 8a MPP Cluster can store data distributedly or use replicated tables, providing an edge in star-schema scenarios. GP stores data distributedly, requiring data movement for small-table joins.

Reliability

High Availability: GBase 8a MPP has unparalleled high availability advantages, while GP suffers from fatal flaws including potential data loss and inability to guarantee data integrity and consistency.

Stability Under High Concurrency: GBase 8a MPP's primary cluster supports automatic load balancing and online node expansion, improving stability and providing an overwhelming advantage in high-concurrency scenarios. GP's single-master architecture suffers from a fatal concurrency bottleneck.

Scalability

GBase 8a MPP has strong scalability advantages. GP does not support online expansion; requiring the cluster to become read-only during expansion is a critical drawback that affects business continuity.

Compatibility

In terms of hardware platforms, GBase 8a MPP supports a broader range of chips and hardware. On the software side, it supports a wider range of operating systems and middleware.

Security

GBase 8a MPP supports transparent column-level encryption, offering better security.

Usability

GBase 8a MPP provides more user-friendly, convenient, and comprehensive operation and management features. GP relies on third-party management tools without vendor support and maintenance.

Extended Features

Scale-In: GBase 8a MPP supports scale-in, which is more aligned with real-world operation scenarios. GP's lack of scale-in capability is a critical shortcoming in production.

Cross-Cluster Operations: GBase 8a MPP offers unparalleled cross-cluster capabilities. GP does not support operations between clusters.

Cluster Virtualization: GBase 8a MPP can build larger-scale clusters through virtual cluster technology, a pioneering advancement in the MPP field.

Multi-Engine Database: GBase 8a MPP Cluster employs a SQL engine and a hybrid architecture with greater development prospects. GP supports SQL and MapReduce engines.

Integrated Appliance: Both have respective strengths in integrated software-hardware appliance capabilities.

Technical Support

GBase 8a MPP provides vendor 7x24 technical support, with a 4-hour on-site response (2 hours in emergencies) for customers in China. This represents an unparalleled advantage in localized service.

GBase 8a MPP Cluster vs. Chinese Competitors

GBase 8a MPP Cluster has clear advantages in performance, availability, usability, scalability, reliability, stability, reference cases, and market share.

Overall Architecture

GBase 8a MPP adopts a federated architecture where the coordinator and compute nodes each form high-availability clusters. This resolves single-point bottlenecks and avoids the complexity of synchronization among management nodes, delivering excellent scalability.

A certain Chinese competitor forms high availability using an active-standby node approach, which may become a bottleneck in large clusters under high concurrency, limiting scalability.

Node-Level Architecture

GBase 8a MPP Cluster is based on a lightweight multi-threaded architecture with low resource consumption that fully leverages multi-core and multi-threading advantages. A certain Chinese competitor uses a multi-process architecture that consumes more resources, incurs heavy context switching overhead, and disrupts data caches, limiting overall performance and scalability.

Data Storage

GBase 8a MPP Cluster is built on native columnar storage design, delivering better support and higher performance for massive data analytics scenarios.

Row and Column Storage: GBase 8a MPP Cluster uses a true columnar engine, providing high performance in analytics. The Chinese competitor imposes many restrictions on its columnar engine, limiting analytics performance.

Data Distribution: GBase 8a MPP Cluster supports HASH distribution tables (including multi-column HASH), random distribution tables, replicated tables, and nocopies tables. Multi-column HASH distribution solves data skew that can occur with single-column HASH. The Chinese competitor supports HASH distribution and replicated tables.

Data Partitioning: GBase 8a MPP Cluster supports RANGE, LIST, HASH, LINEAR HASH, KEY, and LINEAR KEY partitioning. The Chinese competitor only supports RANGE partitioning.

Data Compression: GBase 8a MPP Cluster uses advanced compression algorithms, reducing storage space by 1-20 times and improving I/O performance accordingly. The Chinese competitor supports compression in columnar mode, allowing compression of inactive historical data to reduce space and lower procurement and maintenance costs.

Indexing

GBase 8a MPP Cluster employs coarse-grained intelligent indexes designed for massive data analytics, offering higher performance with zero maintenance and low bloat. The Chinese competitor optimizes query indexes originally for OLTP databases, not specifically tailored for OLAP statistical queries.

Data Loading

Both can achieve efficient parallel data loading; GBase 8a MPP Cluster supports a more comprehensive set of data formats and network protocols.

Backup and Recovery

Both provide backup and recovery tools or features, supporting various backup media and full/incremental backups.

Interfaces and SQL

Both offer multiple programming and management interfaces, with GBase 8a MPP Cluster supporting a richer set. Both have good support for standard SQL and analytical functions; the Chinese competitor provides more comprehensive support for extensions after SQL92.

High Availability

GBase 8a MPP Cluster achieves multi-active high availability for all components, offering greater availability.

High Availability Mechanism: GBase 8a MPP Cluster's coordinator and data compute nodes each form multi-active HA clusters, with automatic failover upon node failure without service interruption. The Chinese competitor implements instance-level comprehensive HA.

Multi-Replica Mechanism: GBase 8a MPP Cluster can use flexible multi-shard mechanisms to reduce the barrel effect, controlling the replica count and distribution of each shard for high availability. The Chinese competitor employs a security ring for shard backup, effectively using only a primary and a standby; the secondary replica only serves as cold backup when the primary fails.