Context:

Scheduling in cloud is complicated as a result of multi-tenancy. Diverse tenants have different requirements, including service functions, response time, QoS and throughput. Diverse tenants require different scheduling capabilities, resource consumption and competition. Multi-tenancy scheduling approaches have been developed for different service models, such as Software as a Service (SaaS), Platform as a service (PaaS), Infrastructure as a Service (IaaS), and Database as a Service (DBaaS).

Objective:

In this paper, we survey the current landscape of multi-tenancy scheduling, laying out the challenges and complexity of software engineering where multi-tenancy issues are involved. This study emphasises scheduling policies, cloud provisioning and deployment with regards to multi-tenancy issues. We conduct a systematic literature review of research studies related to multi-tenancy scheduling approaches on cloud platforms determine the primary scheduling approaches currently used and the challenges for addressing key multi-tenancy scheduling issues.

Method:

We adopted a systematic literature review method to search and review many major journal and conference papers on four major online electronic databases, which address our four predefined research questions. Defining inclusion and exclusion criteria was the initial step before extracting data from the selected papers and deriving answers addressing our inquiries.

Results:

Finally, 53 papers were selected, of which 62 approaches were identified. Most of these methods are developed without cloud layers’ limitation (43.40%) and on SaaS, most of scheduling approaches are oriented to framework design (43.75%).

Conclusion:

The results have demonstrated most of multi-tenancy scheduling solutions can work at any delivery layer. With the difference of tenants’ requirements and functionalities, the choice of cloud service delivery models is changed. Based on our study, designing a multi-tenancy scheduling framework should consider the following 3 factors: computing, QoS and storage resource. One of the potential research foci of multi-tenancy scheduling approaches is on GPU scheduling.

中文翻译：

---

对多租户云平台上的调度方法的系统回顾

内容：

由于多租户的关系，云中的调度非常复杂。不同的租户有不同的要求，包括服务功能，响应时间，QoS和吞吐量。不同的租户需要不同的调度能力，资源消耗和竞争。已经针对不同的服务模型开发了多租户调度方法，例如软件即服务（SaaS），平台即服务（PaaS），基础架构即服务（IaaS）和数据库即服务（DBaaS）。

目的：

在本文中，我们调查了多租户调度的当前状况，提出了涉及多租户问题的软件工程的挑战和复杂性。这项研究强调了有关多租户问题的调度策略，云供应和部署。我们对与云平台上的多租户调度方法有关的研究进行系统的文献综述，以确定当前使用的主要调度方法以及解决关键的多租户调度问题的挑战。

方法：

我们采用了系统的文献综述方法，以在四个主要的在线电子数据库上搜索和审查许多主要期刊和会议论文，这些文献解决了我们四个预先定义的研究问题。定义收录和排除标准是从所选论文中提取数据并得出解决我们问题的答案之前的第一步。

结果：

最后，选择了53篇论文，其中确定了62种方法。这些方法大多数是在不受云层限制的情况下开发的（43.40％），在SaaS上，大多数调度方法都是针对框架设计的（43.75％）。

结论：

结果表明，大多数多租户调度解决方案都可以在任何交付层工作。随着租户需求和功能的不同，云服务交付模式的选择也发生了变化。根据我们的研究，设计多租户调度框架应考虑以下三个因素：计算，QoS和存储资源。多租户调度方法的潜在研究重点之一是GPU调度。