Deep learning (DL) has become essential across various computer science domains. Accurately predicting iteration time for DL models in diverse cloud data center environments is critical for making high-quality scheduling decisions. Existing approaches neglect the sequential features inherent in the runtime execution, leading to issues such as overlooking DL framework overhead and struggling to handle diverse sizes of DL models, resulting in either low accuracy or slow convergence of the prediction model. This paper introduces ETS, a novel iteration time prediction method utilizing execution trace sliding windows. Our observation reveals that DL models exhibit a highly sequential runtime execution nature. Building upon this insight, we leverage sliding windows to extract a novel type of sequential features from the runtime execution trace. These features comprehensively capture DL framework overhead and address the diversity challenge in DL model sizes. By combining a best-practice method to train a prediction model, we achieve high accuracy and rapid convergence simultaneously. Experimental validation on over 14,000 DL model configurations demonstrates ETS's effectiveness in predicting the iteration time of DL models, achieving a mere 5.9% prediction error with a training time at the 10-minute level, and improving scheduling outcomes by reducing job completion time by 17%.

Graph Database Systems (GDBs) support efficiently storing and retrieving graph data, and have become a critical component in many important applications. Many widely-used GDBs utilize the Gremlin query language to create, modify, and retrieve data in graph databases, in which developers can assemble a sequence of Gremlin APIs to perform a complex query. However, incorrect implementations and optimizations of GDBs can introduce logic bugs, which can cause Gremlin queries to return incorrect query results, e.g., omitting vertices in a graph database. In this paper, we propose Query Di sassembling (QuDi), an effective testing technique to automatically detect logic bugs in Gremlin-based GDBs. Given a Gremlin query Q, QuDi disassembles Q into a sequence of atomic graph traversals TList, which shares the equivalent execution semantics with Q. If the execution results of Q and TList are different, a logic bug is revealed in the target GDB. We evaluate QuDi on six popular GDBs, and have found 25 logic bugs in these GDBs, 10 of which have been confirmed as previously-unknown bugs by GDB developers.

Public clouds provide a bewildering choice of configurations for Deep Learning (DL) models, and the choice of configuration will significantly impact the performance and budget. However, it is an obvious challenge to recommend a near-optimal configuration for a particular DL model from a wide range of candidates. The huge search overhead of finding such a configuration is the notorious cold start problem in state-of-the-art efforts, and this problem becomes more severe when they are faced with unseen DL models.

Big data processing is now the primary mission in remote sensing processing, fortunately, cloud computing provides a feasible approach to perform it efficiently. But the work of resource provisioning, scheduling, and scaling is still inevitable in most cloud computing solutions, it poses a considerable challenge to data analyst. The emerging serverless architecture presents a new paradigm to provide a cloud service, the user only needs to upload function codes and leaves all the other server management jobs to the service provider. It reveals a new possibility of remote sensing processing. This paper presents FaasRS, a framework to process remote sensing images upon serverless platform. FaasRS is built on AWS Lambda, it exposes only simple APIs to operate images, and builds DAG for user’s algorithm. FaasRS splits task by splitting the image into small tiles based on geospatial region, and uses each Lambda worker to perform the computation for one tile. To reduce the redundant operations, we also make optimizations based on the algorithm DAG. FaasRS shows favorable performance and scalability in our evaluation. In the comparison with Spark and Ray, FaasRS shows a significant performance improvement in different type of RS processing jobs.

Cloud providers are presented with a bewildering choice of VM types for a range of contemporary data processing frameworks today. However, existing performance modeling and machine learning efforts cannot pick optimal VM types for multiple frameworks simultaneously, since they are difficult to balance model accuracy and model training cost.

Out of memory (OOM) errors occur frequently in data-intensive applications that run atop distributed data-parallel frameworks, such as MapReduce and Spark. In these applications, the memory space is shared by the framework and user code. Since the framework hides the details of distributed execution, it is challenging for users to pinpoint the root causes and fix these OOM errors. This paper presents a comprehensive characteristic study on 123 real-world OOM errors in Hadoop and Spark applications. Our major findings include: (1) 12% errors are caused by the large data buffered/cached in the framework, which indicates that it is hard for users to configure the right memory quota to balance the memory usage of the framework and user code. (2) 37% errors are caused by the unexpected large runtime data, such as large data partition, hotspot key, and large key/value record. (3) Most errors (64%) are caused by memory-consuming user code, which carelessly processes unexpected large data or generates large in-memory computing results. Among them, 13% errors are also caused by the unexpected large runtime data. (4) There are three common fix patterns (used in 34% errors), namely changing the memory/dataflow-related configurations, dividing runtime data, and optimizing user code logic. Our findings inspire us to propose potential solutions to avoid the OOM errors: (1) providing dynamic memory management mechanisms to balance the memory usage of the framework and user code at runtime; (2) providing users with memory+disk data structures, since accumulating large computing results in in-memory data structures is a common cause (15% errors).

Virtualized resource renting is a key issue in IaaS (Infrastructure-as-a-Service) cloud systems. Suitable resource allocation improves resource utilization and increases profit for application providers. The application provider will obtain better revenues according to the Service Level Agreement (SLA), if they rent more virtual resources. However, they will invest much more capital for renting these virtual resources. How many resources a provider rents has become a key thing for cloud applications. This paper addresses the reconciliation objectives by proposing a profit-driven resource scheduling method for virtualized cloud systems. Compared with traditional methods, our method aims at maximizing the revenues by introducing SLA and the cost of renting cloud resource, instead of increasing resource utilization or decreasing early finishing time. We model the performance of applications with queueing theory; calculate the revenues according to the SLA and renting cost; adjust the amount of virtual resources to adapt to dynamic workloads in period. We have implemented a framework for scheduling virtual resources, and applied it in our IaaS cloud platform OnceCloud. The experimental results demonstrate that our method has advantages over existing ones in revenues.

Most recent distributed file systems have adopted architecture with an independent metadata server cluster. However, potential multiple hotspots and flash crowds access patterns often cause a metadata service that violates performance Service Level Objectives. To maximize the throughput of the metadata service, an adaptive request load balancing framework is critical. We present a distributed cache framework above the distributed metadata management schemes to manage hotspots rather than managing all metadata to achieve request load balancing. This benefits the metadata hierarchical locality and the system scalability. Compared with data, metadata has its own distinct characteristics, such as small size and large quantity. The cost of useless metadata prefetching is much less than data prefetching. In light of this, we devise a time period-based prefetching strategy and a perfecting-based adaptive replacement cache algorithm to improve the performance of the distributed caching layer to adapt constantly changing workloads. Finally, we evaluate our approach with a hadoop distributed file system cluster.

Rule execution is the core step of rule-based semantic web reasoning. However, most existing approaches are centralized, which cannot scale out to reason big semantic web datasets. In this paper, we described a kind of semantic web rule execution mechanism using MapReduce programming model, which not only can handle RDFS and OWL ter Horst semantic rules, but also can be used in SWRL reasoning. Theoretical analysis is present on the scalability of this rule execution mechanism. Result shows that it can scale well as Mapreduce framework.

PaaS is one of the most popular paradigms of cloud computing and the performance guarantee of PaaS-oriented applications has been critically concerned. Performance models, such as a Layer Queue Network (LQN) model, are efficient at performance guaranteeing in a highly dynamic computing environment for their capabilities of capacity planning. However, it is not a trivial work to build and employ such models for a PaaS platform because of the lack of designs and the transaction-intensive feature of the PaaS-oriented applications. In this paper, a PaaS-oriented performance modeling approach is proposed. The LQN model of the PaaS-oriented application can be dynamically built through tracing their interactions with the PaaS platform. And the CPU consumptions, which are important to the accuracy of the LQN model, can be refined through a Kalman-filter method. A modeling tool is implemented and experimental results have shown the effectiveness of our approach.

E-commerce systems are typically QoS-sensitive, so QoS-oriented tunings of e-commerce servers are very important for such systems. However, existing e-commerce benchmarks are insufficient for supporting QoS-oriented tunings, because some critical QoS features of e-commerce systems cannot be precisely evaluated by them. One example of these features is the integrality of service, which is usually expressed as a session, provided to customers. This paper presents a QoS-oriented e-commerce benchmark, which is named Bench4Q and is an extension of TPC-W supporting QoS-oriented tuning of e-commerce servers. The main features of Bench4Q include: (1) supporting session-based metrics analysis and (2) simulating QoS-sensitive load for QoS-oriented capacity analysis. We illustrate the promising benefits of these features for QoS-oriented tuning of an e-commerce server by a series of Bench4Q benchmarking on a typical e-commerce server.