Dynamic Performance Management in Multi-tenanted Business Process Servers Using Nonlinear Control

Abstract

Abstract. The methodologies to develop multi-tenanted architectures have been investigated in the recent literature due to the popularity of cloud computing. A number of challenges need to be overcome if multitenanted architectures are to be effective and efficient. Among the challenges is the management of performance properties while effectively sharing the limited resources between the tenants. This work presents an approach to design such a management system for a multi-tenanted business process server. This approach not only enables performance to be maintained at different levels for different tenants depending on their priorities, but also autonomously detects the overloads of aggressive tenants and dynamically changes the control objectives to safeguard the business operations of other tenants. The novelty of the proposed approach is the use of the nonlinear feedback control. The experiment results indicate that the proposed nonlinear control approach achieves the objectives much better compared to the existing fixed and linear control techniques.

SDSN@RT: A middleware environment for single‐instance multitenant cloud applications

“…Tenants can define their applications (processes or message mediation flows) that use their own customized software elements. To achieve the performance differentiation among tenants, the regulation functions such as global admission controllers or schedulers are used to fairly allocate middleware-level computing resources among tenants 17 .…”

“…For example, the fine-grained variants and variation points, and the tangling of normal processing logic with variant differentiation logic can make the application complex 13,14 . Regarding the regulation design of multi-tenant applications, some works support tenant-specific policy enforcement 15 (without sharing or variation of policies among tenants), quality-aware application variant generation 16 , and specific 17,18,19,20 or configurable 21 regulation mechanisms. However, these works have not considered the runtime sharing and tenant-specific variations in the regulation design or to runtime changes to the regulation design.…”

A Systematic Mapping Study on the Multi-tenant Architecture of SaaS Systems

“…On a database level, Moon et al [19] presented a load balancer for multi-tenant databases to increase the performance and sharing of resources among tenants. Patikirikorala et al [20] developed an approach that uses a nonlinear replenished control to keep the performance in distinct usage levels for different tenants, depending on their priorities. It enables the detection of overload, therefore the control of tenants operations can be dynamically changed.…”