Achieve predictable performance
summary
A single FPGA per server is integrated into Intel big‑data frameworks and APIs so unmodified applications use it transparently, offering look-aside and inline acceleration for streaming, network and local storage paths. Implementations show a high‑performance compression for Hadoop shuffle, partner Bitstream reporting about 3x performance improvement for Spark ingress, Outlogic reporting 3x for NoSQL key‑lookups, and Swarm64 delivering up to 1M queries or updates per second for real‑time Postgres (≈5x vs non‑accelerated).
AI-written summary · about
chapters (5)
related talks
17mHardware Accelerated Big Data Processing
26mReimagine Real-time
48mAnalytics Architectures on Microsoft Fabric
23mEnabling SQL to FPGA Using HLS for Filtering/Stream Data Transformations
39mBig Data Systems Performance: The Little Shop of Horrors
27mFPGA-Accelerated Compactions for LSM based Key-Value Store
28mAchieving 10Gbps Line-rate Key-value Stores with FPGAs
32mFuture-Proofing your Architecture - AI at Any Scale with HPE & Aerospike