The "New Normal" or An Evolution
summary
SQREAM positions a GPU-accelerated columnar SQL database running on POWER9 with NVIDIA NVLink that ingests up to 3 TB/hour, auto-compresses 100 TB down to 10–20 TB, and shifts heavy analytics from CPUs to GPUs. Benchmarks show 2x faster loading on 6 billion TPC-H rows and 150–370% faster queries (up to 3.7×) versus x86, and a telco case cuts day‑long Hadoop queries to a couple of hours for complex call‑noise modeling.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 0:28Common Data Challenges
- 1:50The New Normal
- 2:35A personal story
- 3:22Has the situation really changed
- 4:30Data has exploded
- 5:26Challenges
- 6:34Limitations
- 7:34Statistics
- 8:00Two Questions
- 8:31How are you handling these challenges
- 9:59What is SQREAM
- 11:08Customers and partners
- 11:31The 4 Pillars of Massive Data Analytics
- 14:07Traffic warden
- 14:55SEAM on Power9
- 15:25Real Use Case
- 16:22Questions
related talks
33mBuilding smarter ML and AI models and making them more accurate
26mColumnar Storage: Redefining Data Management for the Modern Era
44mQueries in Big Data: the Power of GPU‑based NoSQL Databases
51mPl/cuda the fusion of HPC grade power
26mReimagine Real-time
46mKeynote: When Hardware and Databases Collide
39mUnlock the potential of your data: Optimize your data store architecture...
35mAccelerating PostgreSQL Full Table Scan OLAP queries