What is Hologres?
Hologres is an all-in-one real-time data warehouse engine that is compatible with PostgreSQL. It supports online analytical processing (OLAP) and ad hoc analysis of PB-scale data. Hologres supports online data serving at high concurrency and low latency. It is deeply integrated with MaxCompute, Flink and DataWorks, provides a full-stack data warehouse solution that integrates online and offline processing.

What is Proxima?
Proxima is a high-performance software library from Alibaba DAMO Academy for approximate nearest neighbor search for vectors. Compared to open-source alternatives like Faiss, Proxima excels in stability and performance. It provides industry-leading methods and modules, supporting various application scenarios such as images, videos, and faces.

What is Hologres External Database?
The Hologres External Database feature allows you to load metadata from external data sources into Hologres without migrating the actual data. This makes it easy to manage both internal and external data within a single system, enabling unified metadata management in a lakehouse architecture. It also enables seamless data integration by providing real-time analysis and querying of external data.

What is Hologres Serverless Computing?
Hologres supports Serverless Computing. You can designate large SQL jobs (e.g., SQL jobs with high CPU or memory overhead) to be executed in a fully managed Serverless Computing resource pool. This eliminates the need to reserve fixed-size computing resources, and each job can use independent Serverless Computing resources, ensuring resource isolation and avoiding resource contention and interference.

What are the Suitable Scenarios for Hologres Serverless Computing?
The following scenarios are suitable for the Serverless Computing feature:1.Scenarios where large SQL jobs frequently encounter OOM (Out of Memory) errors, and there is an expectation to improve job success rates and instance stability.2.Scenarios where the current instance has a lot of idle resources during low traffic periods, and there is an expectation to improve resource utilization and reduce costs.3.Scenarios where the current instance experiences resource constraints during peak traffic periods, and even doubling the resources through time-sharing elasticity is not enough to alleviate the resource pressure, with the expectation of further increasing the amount of resources and solving peak traffic problems.

Can Hologres Serverless Computing Accelerate Fixed Plan?
Serverless Computing resources are shared by all users in the current availability zone, and users are scheduled fairly. Query requests may be affected by the resource water level of the current Serverless Computing Pool, which may increase the queuing time; dedicated computing resources are reserved for the current instance. The instance or computing group will not be recycled when it is in the running state, and will not be affected by other user instances. Therefore, in scenarios where query latency is sensitive, it is more suitable to use exclusive computing resources. Serverless Computing cannot support online service scenarios, including Flink real-time writing, DataWorks data integration real-time writing, Holo Client real-time writing, and online point queries through Fixed Plan. If you have various scenario requirements optimized through Fixed Plan, this part of the scenario needs to be run through exclusive computing resources.