This tutorial is aimed at presenting BigDataBench 5.0--a scalable big data and data-driven artificial inteligence (inshort, Al) benchmark suite, We are glad to introduce the following interesting topics:

(1)The challenges and motivation for characterizing modern Big Data and Al workloads.
(2)Why scalable benchmarking methodology matters? What is a data motif? How many data motifs cancharacterize and compose modern comprehensive Big Data and Al workloads? Micro benchmarks of Modern BigData and Ai workloads.
(3)The essentials of modern big data and Al workloads, e.g., lmage classification, lmage generation, Text-to-TextTranslation, lmage-to-text, lmage-to-image, Speech-to-text, Word embedding, Face embedding, Object detectionRecommendation, Graph Model, clustering, classification, PageRank, Feature extraction, Search engine indexing.Their data sets,algorithms, and involved data motifs.

We will give a tutorial on BigDataBench at ASPLOS 2018 in Williamsburg, VA, USA.

March 24, 2018 (Saturday),09:00 - 12:00 (Half Day)

ROOM:TBD

As a multi-discipline, i.e., system, architecture, data management and machine learning, research and engineering effort from both industry and academia, BigDataBench (IISWC’13, HPCA’14, PACT’16, TPDS’17) is an open-source big data and AI benchmark suite. It is roughly estimated that there are 400+ published paper citing or using BigDataBench since 2014.

In this tutorial, first, we will present the challenges in defining modern datacenter workloads that are the foundation of system and architecture research；Second, we report our summary work on identifying eight data dwarfs from comprehensive big data and AI workloads, each of which captures the common requirements of each class of unit of computation while being reasonably divorced from individual implementations. Third, we present BigDataBench --- a data dwarf-based Big Data and AI benchmark suite. Our benchmark suite includes micro benchmarks, each of which is a single data dwarf, components benchmarks, which consists of the data dwarf combinations, and end-to-end application benchmarks, which are the combinations of component benchmarks.

The current version — BigDataBench 4.0 is significant upgrade, and provides 13 representative real-world data sets and 47 benchmarks, including online services, offline analytics, graph analytics, AI, data warehouse, NoSQL, and streaming workloads. Finally, we propose using the combinations of the eight data dwarfs with different weights to mimic the benchmarks in BigDataBench. Our proxy benchmarks shorten the execution time by 100s times on the real systems while they are qualified for both earlier architecture design and later system evaluation across different architectures.