Hologres

The big data warehouse system has evolved from the "complicated" Lambda architecture to the "simplified" real-time offline integrated digital warehouse. Its core is to connect MaxCompute+Hologres offline and real-time digital warehouses based on the streaming computing engine, and realize the hierarchical processing of digital warehouses through metadata and data interchange.

• Metadata auto discovery

  MaxCompute and Hologres have realized two-way metadata automatic discovery and refresh as well as perfect data type support.

• Data sharing and interworking

  Storage direct reading is more than 10 times faster than accessing ordinary surfaces, and supports two-way synchronization of data in millions of rows/second, simplifying data publishing and back flushing scenarios.

• Unified service export

  Hologres directly accelerates the query of MaxCompute data without data movement, reducing data redundancy and realizing BI acceleration.

With the increasing demand of enterprises for data timeliness, the problem of real-time scenarios for real-time processing, real-time storage, and real-time analysis has become increasingly prominent. Based on the concept of Streaming Warehouse, Hologres realizes the efficient flow of real-time data between data warehouse tiers and solves the problem of real-time data warehouse tiers.

• one-stop

  The whole link can be expressed in SQL, and Hologres data at each layer can be reused and checked, facilitating the construction of data layering and multiplexing system of real-time data warehouse.

• High performance

  Flink's powerful real-time computing is perfectly combined with Hologres' extreme real-time writing and updating capabilities, multi-dimensional OLAP and high concurrency spot checking capabilities.

• Enterprise level operation and maintenance

  The operation and maintenance is simpler, the observability is better, and the security capability is stronger. It provides a variety of high availability capabilities to facilitate the construction of an enterprise class streaming warehouse.

Through seamless integration with DLF and OSS, Hologres can directly speed up the reading and writing of data in various formats and types stored on OSS, reduce development, operation and maintenance costs, break data silos, and achieve business insight without moving data (the surface is only used for field mapping, not really storing data) in the form of external tables.

• High performance

  Use vector engine to accelerate OSS/DLF/MaxCompute.

• Openness

  It is convenient to import and export data, and data warehouse data flows freely.

• cost performance

  The exclusive instance lake warehouse resource reuse does not require additional cost calculation, and the shared cluster serverless mode pays by use.

The big model can be widely used in all walks of life, but the ability to answer professional questions in vertical industries is not perfect. Hologres supports vector retrieval capability with high concurrency and low latency, which can be perfectly combined with large models and PAI to complete the construction of enterprise specific question and answer knowledge base.

• Extreme performance

  It supports efficient index construction and vector retrieval with high concurrency and low latency.

• Real time capability

  Support vector data can be written and updated in real-time with high performance, and data can be checked after being written.

• Easy to use

  The use of vector computation can be completed through standard syntax.

For the high availability scenario of online production environment, it provides a master slave multi instance deployment mode of shared storage and computing group instances. In this mode, fault isolation and load isolation are supported, which effectively supports the high availability scenario.

• On demand expansion and reduction

  Warehouse can be pulled up on time or on demand (Scale Out); Warehouse dynamic thermal expansion (Scale Up); Computing and storage are highly scalable and dual elastic.

• Cost reduction and efficiency increase

  Users can use resources as needed, and the cost can be minimized; Based on physical replication, physical files are fully reused, reducing cost and increasing efficiency.

• Computing group resource isolation

  Each computing group is naturally isolated from each other in physical resources to avoid interaction between computing groups and reduce service jitter.