6,742 Hits in 4.1 sec


Guoliang Li, Xuanhe Zhou, Shifu Li, Bo Gao
2019 Proceedings of the VLDB Endowment  
To address these problems, we propose a query-aware database tuning system QTune with a deep reinforcement learning (DRL) model, which can efficiently and effectively tune the database configurations.  ...  We propose a Double-State Deep Deterministic Policy Gradient (DS-DDPG) model to enable query-aware database configuration tuning, which utilizes the actor-critic networks to tune the database configurations  ...  Learning Discrete Configuration Patterns Using Deep Learning. We choose the deep learning method to map queries to discrete configuration patterns.  ... 
doi:10.14778/3352063.3352129 fatcat:6l7joisrvbacnaids5ucur4li4


Vishal Sharma, Curtis Dyreson
2022 Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing  
., continuous or while the DBMS is processing workloads, index selection using a novel online deep reinforcement learning technique using our proposed priority experience sweeping.  ...  There is a growing need for a database with an ability to learn and adapt to evolving workloads. This paper proposes Indexer++, an autonomous, workload-aware, online index tuner.  ...  We would like to extend our appreciation to all of the manuscript reviewers who provided us with valuable feedback that helped us to significantly improve the overall quality of our work.  ... 
doi:10.1145/3477314.3507691 fatcat:5r2zxxioybc7jausbethdkigfa

A Survey of Large-Scale Deep Learning Serving System Optimization: Challenges and Opportunities [article]

Fuxun Yu, Di Wang, Longfei Shangguan, Minjia Zhang, Xulong Tang, Chenchen Liu, Xiang Chen
2022 arXiv   pre-print
Deep Learning (DL) models have achieved superior performance in many application domains, including vision, language, medical, commercial ads, entertainment, etc.  ...  This survey aims to summarize and categorize the emerging challenges and optimization opportunities for large-scale deep learning serving systems.  ...  learning recommendation model DLRM [31] .  ... 
arXiv:2111.14247v2 fatcat:yoeol5xrj5guhh2tiulcvopmue

A Survey of Multi-Tenant Deep Learning Inference on GPU [article]

Fuxun Yu, Di Wang, Longfei Shangguan, Minjia Zhang, Chenchen Liu, Xiang Chen
2022 arXiv   pre-print
Deep Learning (DL) models have achieved superior performance.  ...  With such strong computing scaling of GPUs, multi-tenant deep learning inference by co-locating multiple DL models onto the same GPU becomes widely deployed to improve resource utilization, enhance serving  ...  One example is the Microsoft Deep Learning Inference Service (DLIS) system [35] .  ... 
arXiv:2203.09040v3 fatcat:utvpoyvvajfhfghgpf45nxnbne

RIBBON: Cost-Effective and QoS-Aware Deep Learning Model Inference using a Diverse Pool of Cloud Computing Instances [article]

Baolin Li, Rohan Basu Roy, Tirthak Patel, Vijay Gadepally, Karen Gettings, Devesh Tiwari
2022 arXiv   pre-print
RIBBON saves up to 16% of the inference service cost for different learning models including emerging deep learning recommender system models and drug-discovery enabling models.  ...  Deep learning model inference is a key service in many businesses and scientific discovery processes.  ...  Secondly, some workloads such as the recommendation models come with large embedding tables requiring tens of GBs of memory, general deep learning workloads can also have a large model size and large data  ... 
arXiv:2207.11434v1 fatcat:t5p6sqigizfftkyfpcgimedupq

No DBA? No regret! Multi-armed bandits for index tuning of analytical and HTAP workloads with provable guarantees [article]

R. Malinga Perera, Bastian Oetomo, Benjamin I. P. Rubinstein, Renata Borovica-Gajic
2021 arXiv   pre-print
Furthermore, our bandit framework outperforms deep reinforcement learning (RL) in terms of convergence speed and performance volatility (providing up to 58% speed-up).  ...  We propose a self-driving approach to online index selection that eschews the DBA and query optimiser, and instead learns the benefits of viable structures through strategic exploration and direct performance  ...  three broad types of workloads, allowing us to compare different aspects of the recommendation process: 1.  ... 
arXiv:2108.10130v1 fatcat:36irphbtdfaoxjx6ef6u5gwznq

Opportunistic View Materialization with Deep Reinforcement Learning [article]

Xi Liang, Aaron J. Elmore, Sanjay Krishnan
2019 arXiv   pre-print
Carefully selected materialized views can greatly improve the performance of OLAP workloads. We study using deep reinforcement learning to learn adaptive view materialization and eviction policies.  ...  Results suggest that: (1) DQM can outperform heuristic when their assumptions are not satisfied by the workload or there are temporal effects like period maintenance, (2) even with the cost of learning  ...  The database community has extensively studied view recommendation systems that take in a historical query workload, a database schema, and possibly a cost model to recommend the best views to create  ... 
arXiv:1903.01363v1 fatcat:hgpngvm6ubdpvgrtzd6ykpep7q


Tamer Eldeeb, Zhengneng Chen, Asaf Cidon, Junfeng Yang
2022 Proceedings of the Fifth International Workshop on Exploiting Artificial Intelligence Techniques for Data Management  
Neuroshard represents past queries as a neural hypergraph, and uses Deep Reinforcement Learning with Multi-Task learning to generate a learned partitioner that is able to optimize for multiple objectives  ...  We present Neuroshard, the first system that learns shard assignments directly from the workload, and optimizes for multiple sharding objectives simultaneously.  ...  ACKNOWLEDGMENTS We thank Kexin Pei and Chengzhi Mao for many useful discussions on Deep Learning and introducing us to GradNorm. We also thank Haonan Wang for assisting with experimental evaluation.  ... 
doi:10.1145/3533702.3534908 fatcat:2l5a5v3s4nfmjjyypqredhge7u

A Demonstration of Willump: A Statistically-Aware End-to-end Optimizer for Machine Learning Inference

Peter Kraft, Daniel Kang, Deepak Narayanan, Shoumik Palkar, Peter Bailis, Matei Zaharia
2020 Proceedings of the VLDB Endowment  
Systems for ML inference are widely deployed today, but they typically optimize ML inference workloads using techniques designed for conventional data serving workloads and miss critical opportunities  ...  Willump automatically cascades feature computation for classification queries: Willump classifies most data inputs using only high-value, low-cost features selected by a cost model, improving query performance  ...  In Figure 4 , we show the code for the Music benchmark, which performs music recommendation by querying precomputed features from a remote data store.  ... 
dblp:journals/pvldb/KraftKNPBZ20 fatcat:toomjppxyfeg3hbmaacfumdwxe

PreQR: Pre-training Representation for SQL Understanding

Xiu Tang, Sai Wu, Mingli Song, Shanshan Ying, Feifei Li, Gang Chen
2022 Proceedings of the 2022 International Conference on Management of Data  
A new SQL encoder is then established by adopting the attention mechanism to support on-the-fly query-aware schema linking.  ...  However, most existing learning-based methods adopt the one-hot encoding for SQL query representation, unable to catch complicated semantic context, e.g. structure of query, database schema definition  ...  Therefore, PreQR encoding can be applied to support many learning tasks on database system, such as query log analysis, recommendation and outlier detection.  ... 
doi:10.1145/3514221.3517878 fatcat:55pvzguqujc5jkn5gtpcjm3ymi

PICASSO: Unleashing the Potential of GPU-centric Training for Wide-and-deep Recommender Systems [article]

Yuanxing Zhang, Langshi Chen, Siran Yang, Man Yuan, Huimin Yi, Jie Zhang, Jiamang Wang, Jianbo Dong, Yunlong Xu, Yue Song, Yong Li, Di Zhang (+3 others)
2022 arXiv   pre-print
Using the same hardware budget in production, PICASSO on average shortens the walltime of daily training tasks by 7 hours, significantly reducing the delay of continuous delivery.  ...  To remove this roadblock to the development of recommender systems, we propose a novel framework named PICASSO to accelerate the training of recommendation models on commodity hardware.  ...  We thank Lixue Xia, Wencong Xiao, Shiru Ren, Zhen Zheng, and Zheng Cao for useful pointers regarding the writing of this paper.  ... 
arXiv:2204.04903v2 fatcat:kr4qnafxwvgahfjzfxtxjiwx7a

ILX: Intelligent "Location+X" Data Systems (Vision Paper) [article]

Walid G. Aref, Ahmed M. Aly, Anas Daghistani, Yeasir Rayhan, Jianguo Wang, Libin Zhou
2022 arXiv   pre-print
Queries and operations that are performed on location data warrant the use of database systems. Despite that, location data is being supported in data systems as an afterthought.  ...  Typically, relational or NoSQL data systems that are mostly designed with non-location data in mind get extended with spatial or spatiotemporal indexes, some query operators, and higher level syntactic  ...  ILX's query language should have embedded in it personalized recommendation operators, e.g., based on location-aware Collaborative Filtering, to rank the query engine's responses.  ... 
arXiv:2206.09520v2 fatcat:6f7aqxy225hbbci4zxf2dtz4ie

BI-REC: Guided Data Analysis for Conversational Business Intelligence [article]

Venkata Vamsikrishna Meduri, Abdul Quamar, Chuan Lei, Vasilis Efthymiou, Fatma Ozcan
2021 arXiv   pre-print
We define the space of data analysis in terms of BI patterns, augmented with rich semantic information extracted from the OLAP cube definition, and use graph embeddings learned using GraphSAGE to create  ...  We propose a two-step approach to explore the search space for useful BI pattern recommendations.  ...  They extract patterns from prior workloads and use Machine Learning (ML) techniques to recommend historical queries similar to the current query/session.  ... 
arXiv:2105.00467v1 fatcat:uuwo5axmyfa2jabyitwpgtpguq


Ka-Ho Chow, Umesh Deshpande, Sangeetha Seshadri, Ling Liu
2022 Proceedings of the Seventeenth European Conference on Computer Systems  
This paper presents DeepRest, a deep learningdriven resource estimation system.  ...  DeepRest identifies system anomalies by verifying whether the resource utilization is justifiable by how the application is being used.  ...  Workload Generation. We generate workloads based on real-world behaviors using Locust [10] .  ... 
doi:10.1145/3492321.3519564 fatcat:4w4yvupslreppdgivbltelhgey

Deploying a Steered Query Optimizer in Production at Microsoft

Wangda Zhang, Matteo Interlandi, Paul Mineiro, Shi Qiao, Nasim Ghazanfari, Karlen Lie, Marc Friedman, Rafah Hosn, Hiren Patel, Alekh Jindal
2022 Proceedings of the 2022 International Conference on Management of Data  
Modern analytical workloads are highly heterogeneous and massively complex, making generic query optimizers untenable for many customers and scenarios.  ...  In this paper, we continue a recent line of work in steering a query optimizer towards better plans for a given workload, and make major strides in pushing previous research ideas to production deployment  ...  Recent work on instance optimization have proposed to use machine learning to learn from a given user workload different components of a query optimizer that indirectly lead to better query plan choices  ... 
doi:10.1145/3514221.3526052 fatcat:bvpga6in7jblzo7bms3ta4daje
« Previous Showing results 1 — 15 out of 6,742 results