A Year of Growth in CAPI Solutions
Transcription
A Year of Growth in CAPI Solutions
A Year of Growth in CAPI Solutions Bruce Wile, CAPI Chief Engineer IBM Revolutionizing the Datacenter Join the Conversation #OpenPOWERSummit One year later…..Feb 2015 vs. Feb 2016 Feb 2015 • 2 Accelerators: IBM Data Engine for NoSQL Algo-Logic L3 Order Book • Key Message to CAPI Team: “Go make more accelerators” • 1 Developer Kit Card (Nallatech) Feb 2016 • Dozens of accelerators • 3 Developer Kit Companies (Nallatech, Alpha-Data, Semptian) Each with new cards in the works 2 Join the Conversation #OpenPOWERSummit CAPI Acceleration Types Accelerator Building Block ~A C SAN 48-16G 1 G G8264 ~A C Accelerator External IO on FPGA (available) 840 1 12 1-2 Accelerator Library 146 GB 15 K 146 GB 15 K SAS SAS Function API Power 8S22LC Power 8S22LC Power 8S22LC Power 8S22LC Application Application Application Application Power 8S22LC Power 8S22LC Power 8S22LC Application + Accelerator 3 146 GB 15 K 146 GB 15 K SAS SAS 146 GB 15 K 146 GB 15 K SAS SAS 146 GB 15 K 146 GB 15 K SAS SAS 146 GB 15 K 146 GB 15 K SAS SAS 146 GB 15 K 146 GB 15 K SAS SAS 146 GB 15 K 146 GB 15 K SAS SAS 146 GB 15 K 146 GB 15 K SAS SAS 42U 41U 40U 39U 38U 37U 36U 35U 34U 33U 32U 31U 30U 29U 28U 27U 26U 25U 24U 23U 22U 21U 20U 19U 18U 17U 16U 15U 14U 13U 12U 11U 10U 9U 8U 7U 6U 5U 4U 3U 2U 1U Accelerator Types: Transparent: API is preexisting/common such that the application does not need to change to take advantage of acceleration. Integration Required: Application developer must write code to utilize the accelerator Full Solution: Application and accelerator are part of a complete solution sold to end customers. Full Solution Join the Conversation #OpenPOWERSummit Catalog of CAPI Accelerators – By Type Transparent Integration Required Full Solution Erasure Code IBM Data Engine for No-SQL (“CAPI-Flash”) Genomics Processing GZIP Compression Fast-Fourier Transfer PairHMM Accelerator Monte Carlo Risk Analysis LA Library CV Library Bitwise Encryption DNN Library Key Value Store (KVS) Dynamic Time Warp Pattern Match L3 Order Book Mood Detection Object Detection Object Recognition With many more coming….. 4/1/2016 4 Object Tracking Join the Conversation #OpenPOWERSummit Graph Analytics Djikstra Novara-Fuzzy Text Search Petascale Indexing CAPI Development Ecosystem CAPI Developer Kits available from…. CAPI Frameworks, Examples ….. PSL + DMA P8 AFU DMA Mover CAPI Framework AFU AccDNN-Caffe to CAPI P8 + AFU PSL PSL Sim Engine Memcopy Example 4/1/2016 5 Join the Conversation #OpenPOWERSummit Database Analytics: Four Paradigms for Comparison A Data B DB2 IO Card Data MEMORY IO Card DB2 MEMORY FPGA Analytics Dev Driver Analytics DD Mem Processor Chip Processor Chip SW Only PCI-E FPGA Acceleration + Accelerated Analytics on FPGA gains performance for many algorithms - Device driver overhead and programming difficulty D DB2 Analytics POWER8 Chip CAPI FPGA Acceleration + Accelerated Analytic on FPGA gains performance for many algorithms + Shared Memory model for ease of programming and fast access + Performance gains through pipelining of analyzed data (vs. large block release) + Frees processor threads for other work KEY: MEMORY CAPI MEMORY FPGA CAPI Data C DB2 IO Card FPGA Data Analytics POWER8 Chip CAPI FPGA Acceleration with Database Integration (Data in Motion) + All advantages of CAPI FPGA Acceleration + Single data flow into processor IO yields much higher performance HW Data Path Join Conversation #OpenPOWERSummit 6 SW the Threads Algorithm Workload Acceleration of Informix Description: This is a project to prove the Value of Power 8 for typical IoT Workload Acceleration, exemplified by Dynamic Time Warping: a subsequence similarity measurement algorithm accelerated with FPGA Example: I want to know the times motif like “A” appeared in the history. A Client Value: Faster subsequence similarity search over the historical IoT data than CPU based computing, balanced in-database analysis(without moving data to a separate analytics platform) even under consistent data ingestion. 4/1/2016 7 Join the Conversation #OpenPOWERSummit Emotion Detection for Retail Analytics Vision Accelerator Opportunity – Gain valuable Platform shopper insights by inferring shopper demographic and behavior patterns from many strategically placed cameras. Expression Recognition Face Detect Gender Classify Age Classify Vision Accelerator Interconnect CAPI Bridge Gender: Male Age: 20 – 30 Gaze: 5 seconds Impression: positive Join the Conversation #OpenPOWERSummit Product Detection