AI 용어집
인공지능 완전 사전
RAM Cloud
Storage architecture where all data resides in RAM, offering access times on the order of microseconds and throughput of several million operations per second.
Distributed In-Memory Data Grid
Distributed system that partitions and replicates data across multiple memory nodes, providing horizontal scalability and high availability for transactional applications.
Vectorized Processing
Computing approach that processes datasets in vector blocks to maximize CPU cache utilization and SIMD instructions, reducing loop overhead.
Memory-Mapped Files
Mechanism directly linking a process's virtual address space to disk files, allowing near-instantaneous data access without explicit reads.
Cache Coherence
Protocol ensuring data consistency between multiple cache levels in multi-processor systems during parallel in-memory computing operations.
Persistent Memory
Hybrid storage technology combining DRAM speed with data persistence after power loss, such as Intel Optane DC Persistent Memory.
Spark RDD (Resilient Distributed Datasets)
Fundamental Apache Spark abstraction representing immutable partitioned in-memory collections, with transformation lineage for fault tolerance.
Off-Heap Memory
Memory allocated outside the Java heap to escape the Garbage Collector, allowing manual management and predictable performance for critical data.
Apache Ignite
In-memory distributed computing platform combining data grid, database, and streaming processing with native ACID consistency for high-performance applications.
SAP HANA
Enterprise in-memory database platform combining OLTP and OLAP processing with columnar calculation engine and vector compression for real-time analytics.
Redis
Open-source in-memory data structure supporting optional persistence, master-slave replication, and clustering for applications requiring microsecond latencies.
Data Grid
Middleware architecture distributing data and computations across a cluster of memory nodes, providing elastic scalability and automatic load balancing.
In-Memory Analytics
Analytical approach fully loading datasets into RAM for instant execution of complex queries and interactive calculations without disk size constraints.
In-Memory Data Fabric
Abstraction layer unifying access to distributed heterogeneous in-memory data, providing consistent APIs and automatic placement and replication optimizations.