arXiv subjects

Mingjie Tang

Publications and source records attributed to Mingjie Tang.

34 records · Page 2Linked to original sources

arXiv2024

GPTuner: A Manual-Reading Database Tuning System via GPT-Guided Bayesian Optimization

Modern database management systems (DBMS) expose hundreds of configurable knobs to control system behaviours. Determining the appropriate values for these knobs to improve DBMS performance is a long-standing problem in the database community. As there is an increasing number of knobs to tune and each knob could be in continuous or categorical values, manual tuning becomes impractical. Recently, automatic tuning systems using machine learning methods have shown great potentials. However, existing approaches still incur significant tuning costs or only yield sub-optimal performance. This is because they either ignore the extensive domain knowledge available (e.g., DBMS manuals and forum discussions) and only rely on the runtime feedback of benchmark evaluations to guide the optimization, or they utilize the domain knowledge in a limited way. Hence, we propose GPTuner, a manual-reading database tuning system. Firstly, we develop a Large Language Model (LLM)-based pipeline to collect and refine heterogeneous knowledge, and propose a prompt ensemble algorithm to unify a structured view of the refined knowledge. Secondly, using the structured knowledge, we (1) design a workload-aware and training-free knob selection strategy, (2) develop a search space optimization technique considering the value range of each knob, and (3) propose a Coarse-to-Fine Bayesian Optimization Framework to explore the optimized space. Finally, we evaluate GPTuner under different benchmarks (TPC-C and TPC-H), metrics (throughput and latency) as well as DBMS (PostgreSQL and MySQL). Compared to the state-of-the-art approaches, GPTuner identifies better configurations in 16x less time on average. Moreover, GPTuner achieves up to 30% performance improvement (higher throughput or lower latency) over the best-performing alternative.

cs.DB↗

arXiv2024

Attack as Defense: Run-time Backdoor Implantation for Image Content Protection

As generative models achieve great success, tampering and modifying the sensitive image contents (i.e., human faces, artist signatures, commercial logos, etc.) have induced a significant threat with social impact. The backdoor attack is a method that implants vulnerabilities in a target model, which can be activated through a trigger. In this work, we innovatively prevent the abuse of image content modification by implanting the backdoor into image-editing models. Once the protected sensitive content on an image is modified by an editing model, the backdoor will be triggered, making the editing fail. Unlike traditional backdoor attacks that use data poisoning, to enable protection on individual images and eliminate the need for model training, we developed the first framework for run-time backdoor implantation, which is both time- and resource- efficient. We generate imperceptible perturbations on the images to inject the backdoor and define the protected area as the only backdoor trigger. Editing other unprotected insensitive areas will not trigger the backdoor, which minimizes the negative impact on legal image modifications. Evaluations with state-of-the-art image editing models show that our protective method can increase the CLIP-FID of generated images from 12.72 to 39.91, or reduce the SSIM from 0.503 to 0.167 when subjected to malicious editing. At the same time, our method exhibits minimal impact on benign editing, which demonstrates the efficacy of our proposed framework. The proposed run-time backdoor can also achieve effective protection on the latest diffusion models. Code are available.

cs.CR↗

arXiv2024

mLoRA: Fine-Tuning LoRA Adapters via Highly-Efficient Pipeline Parallelism in Multiple GPUs

Transformer-based, pre-trained large language models (LLMs) have demonstrated outstanding performance across diverse domains, particularly in the emerging {\em pretrain-then-finetune} paradigm. Low-Rank Adaptation (LoRA), a parameter-efficient fine-tuning method, is commonly used to adapt a base LLM to multiple downstream tasks. Further, LLM platforms enable developers to fine-tune multiple models and develop various domain-specific applications simultaneously. However, existing model parallelism schemes suffer from high communication overhead and inefficient GPU utilization when training multiple LoRA tasks across GPUs and machines. In this paper, we present mLoRA, a parallelism-efficient fine-tuning system designed for training multiple LoRA across GPUs and machines. mLoRA introduces a novel LoRA-aware pipeline parallelism scheme that efficiently pipelines independent LoRA adapters and their distinct fine-tuning stages across GPUs and machines, along with a new LoRA-efficient operator to enhance GPU utilization during pipelined LoRA training. Our extensive evaluation shows that mLoRA can significantly reduce average fine-tuning task completion time, e.g., by 30\%, compared to state-of-the-art methods like FSDP. More importantly, mLoRA enables simultaneous fine-tuning of larger models, e.g., two Llama-2-13B models on four NVIDIA RTX A6000 48GB GPUs, which is not feasible for FSDP due to high memory requirements. Hence, mLoRA not only increases fine-tuning efficiency but also makes it more accessible on cost-effective GPUs. mLoRA has been deployed in AntGroup's production environment.

cs.LG↗

arXiv2024

MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts

Fine-tuning Large Language Models (LLMs) is a common practice to adapt pre-trained models for specific applications. While methods like LoRA have effectively addressed GPU memory constraints during fine-tuning, their performance often falls short, especially in multi-task scenarios. In contrast, Mixture-of-Expert (MoE) models, such as Mixtral 8x7B, demonstrate remarkable performance in multi-task learning scenarios while maintaining a reduced parameter count. However, the resource requirements of these MoEs remain challenging, particularly for consumer-grade GPUs with less than 24GB memory. To tackle these challenges, we propose MixLoRA, an approach to construct a resource-efficient sparse MoE model based on LoRA. MixLoRA inserts multiple LoRA-based experts within the feed-forward network block of a frozen pre-trained dense model and employs a commonly used top-k router. Unlike other LoRA-based MoE methods, MixLoRA enhances model performance by utilizing independent attention-layer LoRA adapters. Additionally, an auxiliary load balance loss is employed to address the imbalance problem of the router. Our evaluations show that MixLoRA improves about 9% accuracy compared to state-of-the-art PEFT methods in multi-task learning scenarios. We also propose a new high-throughput framework to alleviate the computation and memory bottlenecks during the training and inference of MOE models. This framework reduces GPU memory consumption by 40% and token computation latency by 30% during both training and inference.

cs.CL↗

arXiv2024

DLRover-RM: Resource Optimization for Deep Recommendation Models Training in the Cloud

Deep learning recommendation models (DLRM) rely on large embedding tables to manage categorical sparse features. Expanding such embedding tables can significantly enhance model performance, but at the cost of increased GPU/CPU/memory usage. Meanwhile, tech companies have built extensive cloud-based services to accelerate training DLRM models at scale. In this paper, we conduct a deep investigation of the DLRM training platforms at AntGroup and reveal two critical challenges: low resource utilization due to suboptimal configurations by users and the tendency to encounter abnormalities due to an unstable cloud environment. To overcome them, we introduce DLRover-RM, an elastic training framework for DLRMs designed to increase resource utilization and handle the instability of a cloud environment. DLRover-RM develops a resource-performance model by considering the unique characteristics of DLRMs and a three-stage heuristic strategy to automatically allocate and dynamically adjust resources for DLRM training jobs for higher resource utilization. Further, DLRover-RM develops multiple mechanisms to ensure efficient and reliable execution of DLRM training jobs. Our extensive evaluation shows that DLRover-RM reduces job completion times by 31%, increases the job completion rate by 6%, enhances CPU usage by 15%, and improves memory utilization by 20%, compared to state-of-the-art resource scheduling frameworks. DLRover-RM has been widely deployed at AntGroup and processes thousands of DLRM training jobs on a daily basis. DLRover-RM is open-sourced and has been adopted by 10+ companies.

cs.DC↗

arXiv2024

BadPart: Unified Black-box Adversarial Patch Attacks against Pixel-wise Regression Tasks

Pixel-wise regression tasks (e.g., monocular depth estimation (MDE) and optical flow estimation (OFE)) have been widely involved in our daily life in applications like autonomous driving, augmented reality and video composition. Although certain applications are security-critical or bear societal significance, the adversarial robustness of such models are not sufficiently studied, especially in the black-box scenario. In this work, we introduce the first unified black-box adversarial patch attack framework against pixel-wise regression tasks, aiming to identify the vulnerabilities of these models under query-based black-box attacks. We propose a novel square-based adversarial patch optimization framework and employ probabilistic square sampling and score-based gradient estimation techniques to generate the patch effectively and efficiently, overcoming the scalability problem of previous black-box patch attacks. Our attack prototype, named BadPart, is evaluated on both MDE and OFE tasks, utilizing a total of 7 models. BadPart surpasses 3 baseline methods in terms of both attack performance and efficiency. We also apply BadPart on the Google online service for portrait depth estimation, causing 43.5% relative distance error with 50K queries. State-of-the-art (SOTA) countermeasures cannot defend our attack effectively.

cs.CV↗

arXiv2024

Couler: Unified Machine Learning Workflow Optimization in Cloud

Machine Learning (ML) has become ubiquitous, fueling data-driven applications across various organizations. Contrary to the traditional perception of ML in research, ML workflows can be complex, resource-intensive, and time-consuming. Expanding an ML workflow to encompass a wider range of data infrastructure and data types may lead to larger workloads and increased deployment costs. Currently, numerous workflow engines are available (with over ten being widely recognized). This variety poses a challenge for end-users in terms of mastering different engine APIs. While efforts have primarily focused on optimizing ML Operations (MLOps) for a specific workflow engine, current methods largely overlook workflow optimization across different engines. In this work, we design and implement Couler, a system designed for unified ML workflow optimization in the cloud. Our main insight lies in the ability to generate an ML workflow using natural language (NL) descriptions. We integrate Large Language Models (LLMs) into workflow generation, and provide a unified programming interface for various workflow engines. This approach alleviates the need to understand various workflow engines' APIs. Moreover, Couler enhances workflow computation efficiency by introducing automated caching at multiple stages, enabling large workflow auto-parallelization and automatic hyperparameters tuning. These enhancements minimize redundant computational costs and improve fault tolerance during deep learning workflow training. Couler is extensively deployed in real-world production scenarios at Ant Group, handling approximately 22k workflows daily, and has successfully improved the CPU/Memory utilization by more than 15% and the workflow completion rate by around 17%.

cs.DB↗

arXiv2022

PandaDB: Understanding Unstructured Data in Graph Database

Unstructured data(e.g., images, videos, PDF files, etc.) contain semantic information, for example, the facial feature of a person and the plate number of a vehicle. There could be semantic relationships between data items which are not explicitly represented. For example, a person's face may appear in two irrelevant photos. Also, much information is represented as structured data(e.g., the person's name and age). End-users prefer to query the semantic information from unstructured data together with structured data based on the potential relationships among them. However, due to the lack of a unified database system for structured and unstructured data, developers have to comprise multiple systems and runtime together to answer these queries. In this work, we build an open-source graph database named PandaDB to consistently manage and query structured and unstructured data. We first introduce a graph data model to manage structured and unstructured data, then propose a new query language to understand the semantics of the unstructured data in the graph. Next, we develop a new cost model and related query optimization techniques to speed up the unstructured data processing pipeline. Finally, we optimize the unstructured data storage and provide the index to speed up the query processing over unstructured data. PandaDB is widely used in industrial applications like FinTech, Knowledge Graph, and Recommendation System. The results show PandaDB can support a large scale of unstructured data query processing in a graph.

cs.DB↗

arXiv2021

Scalable Relational Query Processing on Big Matrix Data

The use of large-scale machine learning methods is becoming ubiquitous in many applications ranging from business intelligence to self-driving cars. These methods require a complex computation pipeline consisting of various types of operations, e.g., relational operations for pre-processing or post-processing the dataset, and matrix operations for core model computations. Many existing systems focus on efficiently processing matrix-only operations, and assume that the inputs to the relational operators are already pre-computed and are materialized as intermediate matrices. However, the input to a relational operator may be complex in machine learning pipelines, and may involve various combinations of matrix operators. Hence, it is critical to realize scalable and efficient relational query processors that directly operate on big matrix data. This paper presents new efficient and scalable relational query processing techniques on big matrix data for in-memory distributed clusters. The proposed techniques leverage algebraic transformation rules to rewrite query execution plans into ones with lower computation costs. A distributed query plan optimizer exploits the sparsity-inducing property of merge functions as well as Bloom join strategies for efficiently evaluating various flavors of the join operation. Furthermore, optimized partitioning schemes for the input matrices are developed to facilitate the performance of join operations based on a cost model that minimizes the communication overhead.The proposed relational query processing techniques are prototyped in Apache Spark. Experiments on both real and synthetic data demonstrate that the proposed techniques achieve up to two orders of magnitude performance improvement over state-of-the-art systems on a wide range of applications.

cs.DB↗

arXiv2020

STULL: Unbiased Online Sampling for Visual Exploration of Large Spatiotemporal Data

Online sampling-supported visual analytics is increasingly important, as it allows users to explore large datasets with acceptable approximate answers at interactive rates. However, existing online spatiotemporal sampling techniques are often biased, as most researchers have primarily focused on reducing computational latency. Biased sampling approaches select data with unequal probabilities and produce results that do not match the exact data distribution, leading end users to incorrect interpretations. In this paper, we propose a novel approach to perform unbiased online sampling of large spatiotemporal data. The proposed approach ensures the same probability of selection to every point that qualifies the specifications of a user's multidimensional query. To achieve unbiased sampling for accurate representative interactive visualizations, we design a novel data index and an associated sample retrieval plan. Our proposed sampling approach is suitable for a wide variety of visual analytics tasks, e.g., tasks that run aggregate queries of spatiotemporal data. Extensive experiments confirm the superiority of our approach over a state-of-the-art spatial online sampling technique, demonstrating that within the same computational time, data samples generated in our approach are at least 50% more accurate in representing the actual spatial distribution of the data and enable approximate visualizations to present closer visual appearances to the exact ones.

cs.DB↗

arXiv2020

SQLFlow: A Bridge between SQL and Machine Learning

Industrial AI systems are mostly end-to-end machine learning (ML) workflows. A typical recommendation or business intelligence system includes many online micro-services and offline jobs. We describe SQLFlow for developing such workflows efficiently in SQL. SQL enables developers to write short programs focusing on the purpose (what) and ignoring the procedure (how). Previous database systems extended their SQL dialect to support ML. SQLFlow (https://sqlflow.org/sqlflow ) takes another strategy to work as a bridge over various database systems, including MySQL, Apache Hive, and Alibaba MaxCompute, and ML engines like TensorFlow, XGBoost, and scikit-learn. We extended SQL syntax carefully to make the extension working with various SQL dialects. We implement the extension by inventing a collaborative parsing algorithm. SQLFlow is efficient and expressive to a wide variety of ML techniques -- supervised and unsupervised learning; deep networks and tree models; visual model explanation in addition to training and prediction; data processing and feature extraction in addition to ML. SQLFlow compiles a SQL program into a Kubernetes-native workflow for fault-tolerable execution and on-cloud deployment. Current industrial users include Ant Financial, DiDi, and Alibaba Group.

cs.DB↗

arXiv2019

A Natural-language-based Visual Query Approach of Uncertain Human Trajectories

Visual querying is essential for interactively exploring massive trajectory data. However, the data uncertainty imposes profound challenges to fulfill advanced analytics requirements. On the one hand, many underlying data does not contain accurate geographic coordinates, e.g., positions of a mobile phone only refer to the regions (i.e., mobile cell stations) in which it resides, instead of accurate GPS coordinates. On the other hand, domain experts and general users prefer a natural way, such as using a natural language sentence, to access and analyze massive movement data. In this paper, we propose a visual analytics approach that can extract spatial-temporal constraints from a textual sentence and support an effective query method over uncertain mobile trajectory data. It is built up on encoding massive, spatially uncertain trajectories by the semantic information of the POIs and regions covered by them, and then storing the trajectory documents in text database with an effective indexing scheme. The visual interface facilitates query condition specification, situation-aware visualization, and semantic exploration of large trajectory data. Usage scenarios on real-world human mobility datasets demonstrate the effectiveness of our approach.

cs.HC↗

arXiv2019

LocationSpark: In-memory Distributed Spatial Query Processing and Optimization

Due to the ubiquity of spatial data applications and the large amounts of spatial data that these applications generate and process, there is a pressing need for scalable spatial query processing. In this paper, we present new techniques for spatial query processing and optimization in an in-memory and distributed setup to address scalability. More specifically, we introduce new techniques for handling query skew, which is common in practice, and optimize communication costs accordingly. We propose a distributed query scheduler that use a new cost model to optimize the cost of spatial query processing. The scheduler generates query execution plans that minimize the effect of query skew. The query scheduler employs new spatial indexing techniques based on bitmap filters to forward queries to the appropriate local nodes. Each local computation node is responsible for optimizing and selecting its best local query execution plan based on the indexes and the nature of the spatial queries in that node. All the proposed spatial query processing and optimization techniques are prototyped inside Spark, a distributed memory-based computation system. The experimental study is based on real datasets and demonstrates that distributed spatial query processing can be enhanced by up to an order of magnitude over existing in-memory and distributed spatial systems.

cs.DB↗

arXiv2017

Adaptive Processing of Spatial-Keyword Data Over a Distributed Streaming Cluster

The widespread use of GPS-enabled smartphones along with the popularity of micro-blogging and social networking applications, e.g., Twitter and Facebook, has resulted in the generation of huge streams of geo-tagged textual data. Many applications require real-time processing of these streams. For example, location-based e-coupon and ad-targeting systems enable advertisers to register millions of ads to millions of users. The number of users is typically very high and they are continuously moving, and the ads change frequently as well. Hence sending the right ad to the matching users is very challenging. Existing streaming systems are either centralized or are not spatial-keyword aware, and cannot efficiently support the processing of rapidly arriving spatial-keyword data streams. This paper presents Tornado, a distributed spatial-keyword stream processing system. Tornado features routing units to fairly distribute the workload, and furthermore, co-locate the data objects and the corresponding queries at the same processing units. The routing units use the Augmented-Grid, a novel structure that is equipped with an efficient search algorithm for distributing the data objects and queries. Tornado uses evaluators to process the data objects against the queries. The routing units minimize the redundant communication by not sending data updates for processing when these updates do not match any query. By applying dynamically evaluated cost formulae that continuously represent the processing overhead at each evaluator, Tornado is adaptive to changes in the workload. Extensive experimental evaluation using spatio-textual range queries over real Twitter data indicates that Tornado outperforms the non-spatio-textually aware approaches by up to two orders of magnitude in terms of the overall system throughput.

cs.DC↗

arXiv2014

Similarity Group-by Operators for Multi-dimensional Relational Data

The SQL group-by operator plays an important role in summarizing and aggregating large datasets in a data analytic stack.While the standard group-by operator, which is based on equality, is useful in several applications, allowing similarity aware grouping provides a more realistic view on real-world data that could lead to better insights. The Similarity SQL-based Group-By operator (SGB, for short) extends the semantics of the standard SQL Group-by by grouping data with similar but not necessarily equal values. While existing similarity-based grouping operators efficiently materialize this approximate semantics, they primarily focus on one-dimensional attributes and treat multidimensional attributes independently. However, correlated attributes, such as in spatial data, are processed independently, and hence, groups in the multidimensional space are not detected properly. To address this problem, we introduce two new SGB operators for multidimensional data. The first operator is the clique (or distance-to-all) SGB, where all the tuples in a group are within some distance from each other. The second operator is the distance-to-any SGB, where a tuple belongs to a group if the tuple is within some distance from any other tuple in the group. We implement and test the new SGB operators and their algorithms inside PostgreSQL. The overhead introduced by these operators proves to be minimal and the execution times are comparable to those of the standard Group-by. The experimental study, based on TPC-H and a social check-in data, demonstrates that the proposed algorithms can achieve up to three orders of magnitude enhancement in performance over baseline methods developed to solve the same problem.

cs.DB↗

arXiv2014

On Order-independent Semantics of the Similarity Group-By Relational Database Operator

Similarity group-by (SGB, for short) has been proposed as a relational database operator to match the needs of emerging database applications. Many SGB operators that extend SQL have been proposed in the literature, e.g., similarity operators in the one-dimensional space. These operators have various semantics. Depending on how these operators are implemented, some of the implementations may lead to different groupings of the data. Hence, if SQL code is ported from one database system to another, it is not guaranteed that the code will produce the same results. In this paper, we investigate the various semantics for the relational similarity group-by operators in the multi-dimensional space. We define the class of order-independent SGB operators that produce the same results regardless of the order in which the input data is presented to them. Using the notion of interval graphs borrowed from graph theory, we prove that, for certain SGB operators, there exist order-independent implementations. For each of these operators, we provide a sample algorithm that is order-independent. Also, we prove that for other SGB operators, there does not exist an order-independent implementation for them, and hence these SGB operators are ill-defined and should not be adopted in extensions to SQL to realize similarity group-by. In this paper, we introduce an SGB operator, namely SGB-All, for grouping multi-dimensional data using similarity. SGB-All forms groups such that a data item, say O, belongs to a group, say G, if and only if O is within a user-defined threshold from all other data items in G. In other words, each group in SGB-All forms a clique of nearby data items in the multi-dimensional space. We prove that SGB-All are order-independent, i.e., there is at least one algorithm for each option that is independent of the presentation order of the input data.

cs.DB↗