Changjie Guo’s research while affiliated with Peking University and other places

Analytical Solutions increasingly play a key role in the modern enterprise business. Currently, such solutions are usually very costly for customer to consume, as the deployment cost is high due to high performance hardware requirement and complex software configuration. Moreover, such on-premises solutions are not suitable for the occasional analytics consumers. On the other hand, Analytical solutions are also hard for solution providers to deliver cost efficient service, since the cost is high in initial customer engagement as well as conducting incremental service. To deliver analytical solutions in a cost-effective way, we propose the idea of "analytical cloud", which is designed to provide on-demand decision support, analytical capabilities, and computational resources in a manageable cloud environment. In this paper, we summarize the potential business models supported by this new delivery model and identify the technical issues required to be addressed in such business models.

Bioinformatics is a developing interdisciplinary science which combines information technologies into biological researches. The techniques from this emerging field have shown great potential in many business areas including drug design, agriculture, and so on. Meanwhile, this new computational field has also been one of the largest consumers of computational power, as the analyses in bioinformatics are often extremely computationally or data intensive. Although there are already several projects which have done tentative exploration on deploying bioinformatics applications to cloud environments, the deployment is ad-hoc and restricted to a single private cloud environment. Moreover, the complexity of various demands of bench biologists and bioinformaticians also brings new challenges to bioinformatics cloud development. In this paper, we first identify the key participants and their interactions in a public bioinformatics cloud environment, where bioinformatic analyses are consumed as services on top of a cloud infrastructure. After that, we propose a research framework to discuss the domain-specific technical challenges in delivering such a solution. Finally, we summarize the existing related research efforts based on our framework and introduce our ongoing Web Lab project.

Multi-tenant is a key characteristic for cost effective Software as a Service (SaaS) applications which drive down total cost of ownership for both service consumers and providers. This paper describes our research in designing & building a cost-effective, secure, customizable, scalable and non-intrusive multi-tenant database which greatly accelerates the migration and development of SaaS applications. We analyze the requirements and gaps in traditional database when supporting SaaS scenario, and then propose a novel nonintrusive multi-tenant database framework to address these challenges. Some key considerations and different implementation approaches in designing and implementation such a framework are discussed and compared. This paper also identifies some potential database performance optimization approaches in the multi-tenant scenario.

In Database-as-a-Service (DBaaS), a large number of tenants share DBaaS resources (CPU, I/O and Memory). While the DBaaS provider runs DBaaS to "share" resources across the entire tenant population to maximize resource utilization and minimize cost, the tenants subscribe to DBaaS at a low price point while still having resources conceptually "isolated" according to service level agreements (SLAs). To optimize this dichotomy of goals, we propose a dynamic resource allocation framework that periodically re-allocates resources to tenants to maximize resource utilization while tolerating a low risk of SLA violations. We model the resource allocation problem as a modified unbounded knapsack problem. The model introduces an additional fairness constraint to assign residual resources to active tenants, while avoiding that few tenants consume all residual resources. Performed experiments demonstrate the effectiveness and efficiency of the proposed allocation algorithm for a synthetic workload with burstiness and predicted tenant behavior.

As one of the key characteristics of software as a Service (SaaS), multi-tenancy aims to support massive customers by sharing application instances and databases. To achieve the high economies of scale, one of the most issues needing to be solved in the real industry is that, given a fixed number of nodes, how to optimally place on-boarding tenants to maximize the total supported number of tenants without violating their SLA requirements. This paper focuses on this problem, which is called On-line Tenant Placement Problem (OTPP). In order to calculate the resource consumption of on-boarding tenants, a novel resource consumption estimation model for multi-tenant pattern is proposed in this paper. Based on this model, we explore the complexity of OTPP. A robust heuristic is proposed for the OTPP. The simulation experimental results show the high effectiveness and the good efficiency of our algorithm.