XDB
XDV

XDB

XDB supports unparalleled speed by supporting massive parallel processing based on SQL and GPU-accelerated operations. It uses the CPU and GPU together to deliver optimal performance and allows you to analyze all areas of complex and massive data. XDB is an ideal SQL engine for large data requiring high speed, and it will unlock deeper data value and business insight that could not be easily obtained before.

Features

ㆍSupport distributed massive parallel processing based on GPU-accelerated computation
ㆍAnalyze all areas of complex and massive data at high speed through the combination of CPU and GPU
ㆍThere is no limit to the processing capacity, and it is possible to analyze even petabyte-level large-capacity data
ㆍReduce data collection and query execution time
ㆍSignificantly faster performance than CPU-based solutions
ㆍSupport standard SQL and various programming languages, APIs, and data sources

Functions

ㆍMPP (Massively Parallel Processing)
ㅤ: Distributed parallel processing architecture using multiple GPU cards
ㅤㅤProcess the SQL query in milliseconds and return the result
ㆍDistributed Architecture
ㅤ: Distribute data across multiple GPUs, each processing data independently
ㅤㅤOutput results are merged into CPU, providing faster data loading speed
ㅤㅤAccelerate big data processing by implementing Multi Node and Multi GPU
ㆍMaster Node ? Data Node
ㅤ: Efficient use of resources with distributed architecture load balancing consisting of a master node and multiple data nodes
ㆍOpen Architecture
ㅤ: Easy integration with various devices and platforms
ㅤㅤGoogle gRPC, JDBC, ODBC / Major programming languages and APIs / Supports various data sources / BI visualization tools
ㆍStandard SQL
ㅤ: Compatible with ANSI-92 SQL / Excellent query performance
ㅤㅤUsers can develop products using their existing SQL knowledge
ㅤㅤSupport complex aggregate operations and federated queries

Key of Data Processing

Columnar

- Based on open Columnar Memory Format
- Save Raw data in column unit
- Improved performance by reducing memory usage and data migration when performing queries

Chunking

- Separate and store large columnar data into small chunks
- Process for efficient use of GPU resources
- Suitable for performing ad-hoc queries

Metadata

- Functions equivalent to indexes in existing databases
- Storing data-related information such as min, max, datatype
- Minimize unnecessary I/O, CPU, and memory usage through data skipping

Compression

- When data is stored, it automatically compresses according to the data tendency
- Efficient use of disk space with high-performance compression technology
- Improve database performance and reduce usage time

Zero Copy

- Minimize the copy process that occurs when data is transferred over the network
- Saving CPU and memory resources
- Fast data processing and efficient memory management

GPU Caching

- Utilizes CPU memory and GPU memory together for data storage
- Minimize access latency by keeping specific data in GPU high-bandwidth memory
- Ideal for real-time big data processing