Large-scale retrieval systems must operate under strict latency constraints while maintaining high recall. Sparse retrieval offers efficiency and interpretability, whereas dense retrieval provides stronger semantic matching. Although hybrid approaches combine both signals, their interaction is often limited, especially under intersection-based retrieval. We introduce HybridSparse, an end-to-end hybrid retrieval framework that strengthens sparse--dense interaction across modeling, training, and serving. It adopts a unified encoder with a shared backbone and jointly optimizes lexical and semantic representations through co-training. To further improve alignment, we incorporate hybrid score regularization and consistency distillation, enabling more stable and effective hybrid scoring. Experiments on public benchmarks demonstrate consistent improvements over strong sparse, dense, and hybrid baselines. In large-scale production deployment for Bing advertisement retrieval, HybridSparse delivers a +1.30% RPM gain, highlighting its practical impact.

We present VQTalker, a Vector Quantization-based framework for multilingual talking head generation that addresses the challenges of lip synchronization and natural motion across diverse languages. Our approach is grounded in the phonetic principle that human speech comprises a finite set of distinct sound units (phonemes) and corresponding visual articulations (visemes), which often share commonalities across languages. We introduce a facial motion tokenizer based on Group Residual Finite Scalar Quantization (GRFSQ), which creates a discretized representation of facial features. This method enables comprehensive capture of facial movements while improving generalization to multiple languages, even with limited training data. Building on this quantized representation, we implement a coarse-to-fine motion generation process that progressively refines facial animations. Extensive experiments demonstrate that VQTalker achieves state-of-the-art performance in both video-driven and speech-driven scenarios, particularly in multilingual settings. Notably, our method achieves high-quality results at a resolution of 512 × 512 pixels while maintaining a lower bitrate of approximately 11 kbps. Our work opens new possibilities for cross-lingual talking face generation.

We propose a novel and general framework to disentangle video data into its dynamic motion and static content components. Our proposed method is a self-supervised pipeline with less assumptions and inductive biases than previous works: it utilizes a transformer-based architecture to jointly generate flexible implicit features for frame-wise motion and clip-wise content, and incorporates a low-bitrate vector quantization as an information bottleneck to promote disentanglement and form a meaningful discrete motion space. The bitrate-controlled latent motion and content are used as conditional inputs to a denoising diffusion model to facilitate self-supervised representation learning. We validate our disentangled representation learning framework on real-world talking head videos with motion transfer and auto-regressive motion generation tasks. Furthermore, we also show that our method can generalize to other types of video data, such as pixel sprites of 2D cartoon characters. Our work presents a new perspective on self-supervised learning of disentangled video representations, contributing to the broader field of video analysis and generation.

Self-consistency-based approaches, which involve repeatedly sampling multiple outputs and selecting the most consistent one as the final response, prove to be remarkably effective in improving the factual accuracy of large language models. Nonetheless, existing methods usually have strict constraints on the task format, largely limiting their applicability. In this paper, we present Integrative Decoding (ID), to unlock the potential of self-consistency in open-ended generation tasks. ID operates by constructing a set of inputs, each prepended with a previously sampled response, and then processes them concurrently, with the next token being selected by aggregating of all their corresponding predictions at each decoding step. In essence, this simple approach implicitly incorporates self-consistency in the decoding objective. Extensive evaluation shows that ID consistently enhances factuality over a wide range of language models, with substantial improvements on the TruthfulQA (+11.2%), Biographies (+15.4%) and LongFact (+8.5%) benchmarks. The performance gains amplify progressively as the number of sampled responses increases, indicating the potential of ID to scale up with repeated sampling.

Existing methods for code generation use code snippets as seed data, restricting the complexity and diversity of the synthesized data. In this paper, we introduce a novel feature tree-based synthesis framework, which revolves around hierarchical code features derived from high-level abstractions of code. The feature tree is constructed from raw data and refined iteratively to increase the quantity and diversity of the extracted features, which captures and recognizes more complex patterns and relationships within the code. By adjusting the depth and breadth of the sampled subtrees, our framework provides precise control over the complexity of the generated code, enabling functionalities that range from function-level operations to multi-file scenarios. We fine-tuned widely-used base models to obtain EpiCoder series, achieving state-of-the-art performance on multiple benchmarks at both the function and file levels. In particular, empirical evidence indicates that our approach shows significant potential in the synthesizing of repository-level code data. Our code and data are publicly available.

Transformer-based Large Language Models (LLMs) have become increasingly important. However, scaling LLMs to longer contexts incurs slow inference speed and high GPU memory consumption for caching key-value (KV) vectors. This paper presents RetrievalAttention, a training-free approach to both accelerate the decoding phase and reduce GPU memory consumption by pre-building KV vector indexes for fixed contexts and maintaining them in CPU memory for efficient retrieval. Unlike conventional KV cache methods, RetrievalAttention integrate approximate nearest neighbor search (ANNS) indexes into attention computation. We observe that off-the-shelf ANNS techniques often fail due to the out-of-distribution (OOD) nature of query and key vectors in attention mechanisms. RetrievalAttention overcomes this with an attention-aware vector index. Our evaluation shows RetrievalAttention achieves near full attention accuracy while accessing only 1-3\% of the data, significantly reducing inference costs. Remarkably, RetrievalAttention enables LLMs with 8B parameters to handle 128K tokens on a single NVIDIA RTX4090 (24GB), achieving a decoding speed of 0.107 seconds per token.

Large language models (LLMs) have empowered intelligent agents to execute intricate tasks within domain-specific software such as browsers and games. However, when applied to general-purpose software systems like operating systems, LLM agents face three primary challenges. Firstly, the action space is vast and dynamic, posing difficulties for LLM agents to maintain an up-to-date understanding and deliver accurate responses. Secondly, real-world tasks often require inter-application cooperation, demanding farsighted planning from LLM agents. Thirdly, agents need to identify optimal solutions aligning with user constraints, such as security concerns and preferences. These challenges motivate AndroidArena, an environment and benchmark designed to evaluate LLM agents on a modern operating system. To address high-cost of manpower, we design a scalable and semi-automated method to construct the benchmark. In the task evaluation, AndroidArena incorporates accurate and adaptive metrics to address the issue of non-unique solutions. Our findings reveal that even state-of-the-art LLM agents struggle in cross-APP scenarios and adhering to specific constraints. Additionally, we identify a lack of four key capabilities, i.e. understanding, reasoning, exploration, and reflection, as primary reasons for the failure of LLM agents. Furthermore, we provide empirical analysis on the failure of reflection, and improve the success rate by 27% with our proposed exploration strategy. This work is the first to present valuable insights in understanding fine-grained weakness of LLM agents, and offers a path forward for future research in this area. Environment, benchmark, prompt, and evaluation code for AndroidArena are released at https://github.com/AndroidArenaAgent/AndroidArena.

The paper introduces AniTalker, an innovative framework designed to generate lifelike talking faces from a single portrait. Unlike existing models that primarily focus on verbal cues such as lip synchronization and fail to capture the complex dynamics of facial expressions and nonverbal cues, AniTalker employs a universal motion representation. This innovative representation effectively captures a wide range of facial dynamics, including subtle expressions and head movements. AniTalker enhances motion depiction through two self-supervised learning strategies: the first involves reconstructing target video frames from source frames within the same identity to learn subtle motion representations, and the second develops an identity encoder using metric learning while actively minimizing mutual information between the identity and motion encoders. This approach ensures that the motion representation is dynamic and devoid of identity-specific details, significantly reducing the need for labeled data. Additionally, the integration of a diffusion model with a variance adapter allows for the generation of diverse and controllable facial animations. This method not only demonstrates AniTalker's capability to create detailed and realistic facial movements but also underscores its potential in crafting dynamic avatars for real-world applications. Synthetic results can be viewed at https://github.com/X-LANCE/AniTalker.

Caching is crucial for accelerating data access, employed as a ubiquitous design in modern systems at many parts of computer systems. With increasing core count, and shrinking latency gap between cache and modern storage devices, hit-path scalability becomes increasingly critical. However, existing production in-memory caches often use list-based management with promotion on each cache hit, which requires extensive locking and poses a significant overhead for scaling beyond a few cores. Moreover, existing techniques for improving scalability either (1) only focus on the indexing structure and do not improve cache management scalability, or (2) sacrifice efficiency or miss-path scalability.

Audio-driven talking face has attracted broad interest from academia and industry recently. However, data acquisition and labeling in audio-driven talking face are labor-intensive and costly. The lack of data resource results in poor synthesis effect. To alleviate this issue, we propose to use TTS (Text-To-Speech) for data augmentation to improve few-shot ability of the talking face system. The misalignment problem brought by the TTS audio is solved with the introduction of soft-DTW, which is first adopted in the talking face task. Moreover, features extracted by HuBERT are explored to utilize underlying information of audio, and found to be superior over other features. The proposed method achieves 17%, 14%, 38% dominance on MSE score, DTW score and user study preference repectively over the baseline model, which shows the effectiveness of improving few-shot learning for talking face system with TTS augmentation.

While recent research has made significant progress in speech-driven talking face generation, the quality of the generated video still lags behind that of real recordings. One reason for this is the use of handcrafted intermediate representations like facial landmarks and 3DMM coefficients, which are designed based on human knowledge and are insufficient to precisely describe facial movements. Additionally, these methods require an external pretrained model for extracting these representations, whose performance sets an upper bound on talking face generation. To address these limitations, we propose a novel method called DAE-Talker that leverages data-driven latent representations obtained from a diffusion autoencoder (DAE). DAE contains an image encoder that encodes an image into a latent vector and a DDIM-based image decoder that reconstructs the image from it. We train our DAE on talking face video frames and then extract their latent representations as the training target for a Conformer-based speech2latent model. During inference, DAE-Talker first predicts the latents from speech and then generates the video frames with the image decoder in DAE from the predicted latents. This allows DAE-Talker to synthesize full video frames and produce natural head movements that align with the content of speech, rather than relying on a predetermined head pose from a template video. We also introduce pose modelling in speech2latent for pose controllability. Additionally, we propose a novel method for generating continuous video frames with the DDIM-based image decoder trained on individual frames, eliminating the need for modelling the joint distribution of consecutive frames directly. Our experiments show that DAE-Talker outperforms existing popular methods in lip-sync, video fidelity, and pose naturalness. We also conduct ablation studies to analyze the effectiveness of the proposed techniques and demonstrate the pose controllability of DAE-Talker.

On-device Deep Neural Network (DNN) inference consumes significant computing resources and development efforts. To alleviate that, we propose LUT-NN, the first system to empower inference by table lookup, to reduce inference cost. LUT-NN learns the typical features for each operator, named centroid, and precompute the results for these centroids to save in lookup tables. During inference, the results of the closest centroids with the inputs can be read directly from the table, as the approximated outputs without computations.

Embedding-based retrieval methods construct vector indices to search for document representations that are most similar to the query representations. They are widely used in document retrieval due to low latency and decent recall performance. Recent research indicates that deep retrieval solutions offer better model quality, but are hindered by unacceptable serving latency and the inability to support document updates. In this paper, we aim to enhance the vector index with end-to-end deep generative models, leveraging the differentiable advantages of deep retrieval models while maintaining desirable serving efficiency. We propose Model-enhanced Vector Index (MEVI), a differentiable model-enhanced index empowered by a twin-tower representation model. MEVI leverages a Residual Quantization (RQ) codebook to bridge the sequence-to-sequence deep retrieval and embedding-based models. To substantially reduce the inference time, instead of decoding the unique document ids in long sequential steps, we first generate some semantic virtual cluster ids of candidate documents in a small number of steps, and then leverage the well-adapted embedding vectors to further perform a fine-grained search for the relevant documents in the candidate virtual clusters. We empirically show that our model achieves better performance on the commonly used academic benchmarks MSMARCO Passage and Natural Questions, with comparable serving latency to dense retrieval solutions.

Approximate Nearest Neighbor Search (ANNS) on high dimensional vector data is now widely used in various applications, including information retrieval, question answering, and recommendation. As the amount of vector data grows continuously, it becomes important to support updates to vector index, the enabling technique that allows for efficient and accurate ANNS on vectors.

Current state-of-the-art document retrieval solutions mainly follow an index-retrieve paradigm, where the index is hard to be directly optimized for the final retrieval target. In this paper, we aim to show that an end-to-end deep neural network unifying training and indexing stages can significantly improve the recall performance of traditional methods. To this end, we propose Neural Corpus Indexer (NCI), a sequence-to-sequence network that generates relevant document identifiers directly for a designated query. To optimize the recall performance of NCI, we invent a prefix-aware weight-adaptive decoder architecture, and leverage tailored techniques including query generation, semantic document identifiers, and consistency-based regularization. Empirical studies demonstrated the superiority of NCI on two commonly used academic benchmarks, achieving +21.4% and +16.8% relative enhancement for Recall@1 on NQ320k dataset and R-Precision on TriviaQA dataset, respectively, compared to the best baseline method.

Vector quantization (VQ) based ANN indexes, such as Inverted File System (IVF) and Product Quantization (PQ), have been widely applied to embedding based document retrieval thanks to the competitive time and memory efficiency. Originally, VQ is learned to minimize the reconstruction loss, i.e., the distortions between the original dense embeddings and the reconstructed embeddings after quantization. Unfortunately, such an objective is inconsistent with the goal of selecting ground-truth documents for the input query, which may cause severe loss of retrieval quality. Recent works identify such a defect, and propose to minimize the retrieval loss through contrastive learning. However, these methods intensively rely on queries with ground-truth documents, whose performance is limited by the insufficiency of labeled data. In this paper, we propose Distill-VQ, which unifies the learning of IVF and PQ within a knowledge distillation framework. In Distill-VQ, the dense embeddings are leveraged as "teachers'', which predict the query's relevance to the sampled documents. The VQ modules are treated as the "students'', which are learned to reproduce the predicted relevance, such that the reconstructed embeddings may fully preserve the retrieval result of the dense embeddings. By doing so, Distill-VQ is able to derive substantial training signals from the massive unlabeled data, which significantly contributes to the retrieval quality. We perform comprehensive explorations for the optimal conduct of knowledge distillation, which may provide useful insights for the learning of VQ based ANN index. We also experimentally show that the labeled data is no longer a necessity for high-quality vector quantization, which indicates Distill-VQ's strong applicability in practice. The evaluations are performed on MS MARCO and Natural Questions benchmarks, where Distill-VQ notably outperforms the SOTA VQ methods in Recall and MRR. Our code is avaliable at https://github.com/staoxiao/LibVQ.

Deep matching models aim to facilitate search engines retrieving more relevant documents by mapping queries and documents into semantic vectors in the first-stage retrieval. When leveraging BERT as the deep matching model, the attention score across two words are solely built upon local contextualized word embeddings. It lacks prior global knowledge to distinguish the importance of different words, which has been proved to play a critical role in information retrieval tasks. In addition to this, BERT only performs attention across sub-words tokens which weakens whole word attention representation. We propose a novel Global Weighted Self-Attention (GLOW) network for web document search. GLOW fuses global corpus statistics into the deep matching model. By adding prior weights into attention generation from global information, like BM25, GLOW successfully learns weighted attention scores jointly with query matrix Q and key matrix K. We also present an efficient whole word weight sharing solution to bring prior whole word knowledge into sub-words level attention. It aids Transformer to learn whole word level attention. To make our models applicable to complicated web search scenarios, we introduce combined fields representation to accommodate documents with multiple fields even with variable number of instances. We demonstrate GLOW is more efficient to capture the topical and semantic representation both in queries and documents. Intrinsic evaluation and experiments conducted on public data sets reveal GLOW to be a general framework for document retrieve task. It significantly outperforms BERT and other competitive baselines by a large margin while retaining the same model complexity with BERT. The source code is available at https://github.com/GLOW-deep/GLOW.

The in-memory algorithms for approximate nearest neighbor search (ANNS) have achieved great success for fast high-recall search, but are extremely expensive when handling very large scale database. Thus, there is an increasing request for the hybrid ANNS solutions with small memory and inexpensive solid-state drive (SSD). In this paper, we present a simple but efficient memory-disk hybrid indexing and search system, named SPANN, that follows the inverted index methodology. It stores the centroid points of the posting lists in the memory and the large posting lists in the disk. We guarantee both disk-access efficiency (low latency) and high recall by effectively reducing the disk-access number and retrieving high-quality posting lists. In the index-building stage, we adopt a hierarchical balanced clustering algorithm to balance the length of posting lists and augment the posting list by adding the points in the closure of the corresponding clusters. In the search stage, we use a query-aware scheme to dynamically prune the access of unnecessary posting lists. Experiment results demonstrate that SPANN is 2X faster than the state-of-the-art ANNS solution DiskANN to reach the same recall quality 90% with same memory cost in three billion-scale datasets. It can reach 90% recall@1 and recall@10 in just around one millisecond with only about 10% of original memory cost. Code is available at: https://github.com/microsoft/SPTAG.

To achieve good result quality and short query response time, search engines use specific match plans on Inverted Index to help retrieve a small set of relevant documents from billions of web pages. A match plan is composed of a sequence of match rules, which contain discrete match rule types and continuous stopping quotas. Currently, match plans are manually designed by experts according to their several years’ experience, which encounters difficulty in dealing with heterogeneous queries and varying data distribution. In this work, we formulate the match plan generation as a Partially Observable Markov Decision Process (POMDP) with a parameterized action space, and propose a novel reinforcement learning algorithm Parameterized Action Soft Actor-Critic (PASAC) to effectively enhance the exploration in both spaces. In our scene, we also discover a skew prioritizing issue of the original Prioritized Experience Replay (PER) and introduce Stratified Prioritized Experience Replay (SPER) to address it. We are the first group to generalize this task for all queries as a learning problem with zero prior knowledge and successfully apply deep reinforcement learning in the real web search environment. Our approach greatly outperforms the well-designed production match plans by over 70% reduction of index block accesses with the quality of documents almost unchanged, and 9% reduction of query response time even with model inference cost. Our method also beats the baselines on some open-source benchmarks1.

To estimate time complexity for a given algorithm is important for algorithm designers. Usually, time complexity means the "analytical" time complexity which needs to be proved by strict math derivation. We propose to estimate the "numerical" time complexity (NTC), which measures the minimum number of operations an algorithm has to spend, as well as capture the intrinsic laws of time complexity. The unique challenges include: (1) How to make a machine learning model has the same ability as a real-world CPU (2) How to measure the minimum number of required arithmetic operations for a given problem. To tackle these challenges, we first propose a memory-based recurrent calculation network to mimic the functions of CPU and then we propose a self-adaptive selection gate for deciding when the mimic calculation process should stop. In addition, we use a symbolic learning method to find the time complexity formula. We train and test our model on four basic algorithms: long integer addition, 1-dim max-pooling, outer product, and sorting. Experiment results demonstrate that our model can precisely predict the numerical time complexity as well as the time complexity formula for each algorithm. We also conduct many visualizations to prove the effectiveness and correctness of our model.

Nowadays, many MapReduce applications consist of groups of jobs with dependencies among each other, such as iterative machine learning applications and large database queries. Unfortunately, the MapReduce framework is not optimized for these multi-job applications. It does not explore the execution overlapping opportunities among jobs and can only schedule jobs independently. These issues significantly inflate the application execution time. This paper presents Pipeline Improvement Support with Critical chain Estimation Scheduling (PISCES), a critical chain optimization (a critical chain refers to a series of jobs which will make the application run longer if any one of them is delayed), to provide better support for multi-job applications. PISCES extends the existing MapReduce framework to allow scheduling for multiple jobs with dependencies by dynamically building up a job dependency DAG for current running jobs according to their input and output directories. Then using the dependency DAG, it provides an innovative mechanism to facilitate the data pipelining between the output phase (map phase in the Map-Only job or reduce phase in the Map-Reduce job) of an upstream job and the map phase of a downstream job. This offers a new execution overlapping between dependent jobs in MapReduce which effectively reduces the application runtime. Moreover, PISCES proposes a novel critical chain job scheduling model based on the accurate critical chain estimation. Experiments show that PISCES can increase the degree of system parallelism by up to 68 percent and improve the execution speed of applications by up to 52 percent.

The lack of labeled data is one of the main challenges when building a task-oriented dialogue system.Existing dialogue datasets usually rely on human labeling, which is expensive, limited in size, and in low coverage.In this paper, we instead propose our framework auto-dialabel to automatically cluster the dialogue intents and slots.In this framework, we collect a set of context features, leverage an autoencoder for feature assembly, and adapt a dynamic hierarchical clustering method for intent and slot labeling.Experimental results show that our framework can promote human labeling cost to a great extent, achieve good intent clustering accuracy (84.1%), and provide reasonable and instructive slot labeling results.

Many applications perform real-time analysis on data streams. We argue that existing solutions are poorly matched to the need, and introduce our new Freeze-Frame File System. Freeze-Frame FS is able to accept streams of updates while satisfying "temporal reads" on demand. The system is fast and accurate: we keep all update history in a memory-mapped log, cache recently retrieved data for repeat reads, and use a hybrid of a real-time and a logical clock to respond to read requests in a manner that is both temporally precise and causally consistent. When RDMA hardware is available, the write and read throughput of a single client reaches 2.6GB/s for writes and 5GB/s for reads, close to the limit (about 6GB/s) on the RDMA hardware used in our experiments. Even without RDMA, Freeze Frame FS substantially outperforms existing options for our target settings.

MapReduce is an effective tool for parallel data processing. One significant issue in practical MapReduce applications is data skew: the imbalance in the amount of data assigned to each task. This causes some tasks to take much longer to finish than others and can significantly impact performance. This paper presents LIBRA, a lightweight strategy to address the data skew problem among the reducers of MapReduce applications. Unlike previous work, LIBRA does not require any pre-run sampling of the input data or prevent the overlap between the map and the reduce stages. It uses an innovative sampling method which can achieve a highly accurate approximation to the distribution of the intermediate data by sampling only a small fraction of the intermediate data during the normal map processing. It allows the reduce tasks to start copying as soon as the chosen sample map tasks (only a small fraction of map tasks which are issued first) complete. It supports the split of large keys when application semantics permit and the total order of the output data. It considers the heterogeneity of the computing resources when balancing the load among the reduce tasks appropriately. LIBRA is applicable to a wide range of applications and is transparent to the users. We implement LIBRA in Hadoop and our experiments show that LIBRA has negligible overhead and can speed up the execution of some popular applications by up to a factor of 4.

MapReduce is a widely used parallel computing framework for large scale data processing. The two major performance metrics in MapReduce are job execution time and cluster throughput. They can be seriously impacted by straggler machines-machines on which tasks take an unusually long time to finish. Speculative execution is a common approach for dealing with the straggler problem by simply backing up those slow running tasks on alternative machines. Multiple speculative execution strategies have been proposed, but they have some pitfalls: (i) Use average progress rate to identify slow tasks while in reality the progress rate can be unstable and misleading, (ii) Cannot appropriately handle the situation when there exists data skew among the tasks, (iii) Do not consider whether backup tasks can finish earlier when choosing backup worker nodes. In this paper, we first present a detailed analysis of scenarios where existing strategies cannot work well. Then we develop a new strategy, maximum cost performance (MCP), which improves the effectiveness of speculative execution significantly. To accurately and promptly identify stragglers, we provide the following methods in MCP: (i) Use both the progress rate and the process bandwidth within a phase to select slow tasks, (ii) Use exponentially weighted moving average (EWMA) to predict process speed and calculate a task's remaining time, (iii) Determine which task to backup based on the load of a cluster using a cost-benefit model. To choose proper worker nodes for backup tasks, we take both data locality and data skew into consideration. We evaluate MCP in a cluster of 101 virtual machines running a variety of applications on 30 physical servers. Experiment results show that MCP can run jobs up to 39 percent faster and improve the cluster throughput by up to 44 percent compared to Hadoop-0.21.

Data center applications present significant opportunities for multiplexing server resources. Virtualization technology makes it easy to move running application across physical machines. In this paper, we present an approach that uses virtualization technology to allocate data center resources dynamically based on application demands and support green computing by optimizing the number of servers actively used. We abstract this as a variant of the relaxed on-line bin packing problem and develop a practical, efficient algorithm that works well in a real system. We adjust the resources available to each VM both within and across physical servers. Extensive simulation and experiment results demonstrate that our system achieves good performance compared to the existing work.

Many Internet applications can benefit from an automatic scaling property where their resource usage can be scaled up and down automatically by the cloud service provider. We present a system that provides automatic scaling for Internet applications in the cloud environment. We encapsulate each application instance inside a virtual machine (VM) and use virtualization technology to provide fault isolation. We model it as the Class Constrained Bin Packing (CCBP) problem where each server is a bin and each class represents an application. The class constraint reflects the practical limit on the number of applications a server can run simultaneously. We develop an efficient semi-online color set algorithm that achieves good demand satisfaction ratio and saves energy by reducing the number of servers used when the load is low. Experiment results demonstrate that our system can improve the throughput by 180% over an open source implementation of Amazon EC2 and restore the normal QoS five times as fast during flash crowds. Large scale simulations demonstrate that our algorithm is extremely scalable: the decision time remains under 4 s for a system with 10 000 servers and 10 000 applications. This is an order of magnitude improvement over traditional application placement algorithms in enterprise environments.

Cloud computing allows business customers to scale up and down their resource usage based on needs. Many of the touted gains in the cloud model come from resource multiplexing through virtualization technology. In this paper, we present a system that uses virtualization technology to allocate data center resources dynamically based on application demands and support green computing by optimizing the number of servers in use. We introduce the concept of "skewness” to measure the unevenness in the multidimensional resource utilization of a server. By minimizing skewness, we can combine different types of workloads nicely and improve the overall utilization of server resources. We develop a set of heuristics that prevent overload in the system effectively while saving energy used. Trace driven simulation and experiment results demonstrate that our algorithm achieves good performance.