A5039050101- Cloud Computing and Resource Management
- Distributed systems and fault tolerance
- Advanced Database Systems and Queries
- Advanced Data Storage Technologies
- Software System Performance and Reliability
- Distributed and Parallel Computing Systems
- Blockchain Technology Applications and Security
- Data Management and Algorithms
- Caching and Content Delivery
- Parallel Computing and Optimization Techniques
- Real-Time Systems Scheduling
Huawei Technologies (China)
2022
Institute of Computing Technology
2017-2019
Chinese Academy of Sciences
2017-2019
University of Chinese Academy of Sciences
2019
An ever increasing number of configuration parameters are provided to system users. But many users have used one setting across different workloads, leaving untapped the performance potential systems. A good can greatly improve a deployed under certain workloads. with tens or hundreds parameters, it becomes highly costly task decide which leads best performance. While such requires strong expertise in both and application, commonly lack expertise. To help tap systems, we present BestConfig,...
We observe that the time bottleneck during recovery phase of an IMDB (In-Memory DataBase system) shifts from log replaying to index rebuilding after state-of-art techniques for instant have been applied. In this paper, we investigate checkpoints eliminate bottleneck. However, improper designs may lead inconsistent or incur severe performance degradation. For correctness challenge, combine two techniques, i.e. , deferred deletion entries, and on-demand clean-up dangling entries recovery,...
Fast database engines have become an essential building block in many systems and applications. Yet most of them are designed based on on-premise solutions do not directly work the cloud. Existing cloud-native mostly disk resident databases that follow a storage-centric design exploit potential modern cloud infrastructure, such as manycore processors, large main memory persistent memory. However, in-memory infrequent untapped.
Fast in-memory key value stores are the keys to building large-scale Internet services. The state-of-the-art solutions mainly focus on optimizing performance for read-intensive workloads. Nevertheless, a wide range of applications demonstrate significant amount updates and queries, which scale poorly with current implementations. In this paper, we present BiloKey, highly scalable store multi-core machines, significantly outperforming Redis Memcached variety mixed read write To achieve this,...
To support the variety of Big Data use cases, many related systems expose a large number user-specifiable configuration parameters. Highlighted in our experiments, MySQL deployment with well-tuned parameters achieves peak throughput as 12 times much one default setting. However, finding best setting for tens or hundreds is mission impossible ordinary users. Worse still, applications require multiple co-deployed same cluster. As these can interact to affect overall performance, they must be...
Highly-available datastores are widely deployed for online applications. However, many applications not contented with the simple data access interface currently provided by highly-available datastores. Distributed transaction support is demanded such as large-scale payment used Alipay or Paypal. Current solutions to distributed can spend more than half of whole processing time in commit. An efficient atomic commit protocol highly desirable. This paper presents HACommit protocol, a logless...
Although ACID is the previous golden rule for transaction support, durability now not a basic requirement data storage. Rather, high availability becoming first-class property required by online applications. We show that of almost surely stronger than durability. thus propose ACIA (Atomicity, Consistency, Isolation, Availability) as new standard support. Essentially, shift from to due change assumed conditions management. Four major condition changes exist. With transactions, more diverse...
This document outlines the approach to supporting cross-node transactions over a Redis cluster.
High-efficiency in-memory caching databases are the key to building large-scale Internet services. Well-designed cache can reduce pressure on network servers and response delay of applications. The popular memory system Memcached Redis have been successfully deployed by famous enterprises, such as Amazon, GitHub Sina. state-of-the-art solutions mainly focus optimizing performance providing a suit static policies for various Take key-value data structure store example, it provides LRU, LFU...