A scalable application placement controller for enterprise data centers

Abstract

Given a set of machines and a set of Web applications with dynamically changing demands, an online application placement controller decides how many instances to run for each application and where to put them, while observing all kinds of resource constraints. This NP hard problem has real usage in commercial middleware products. Existing approximation algorithms for this problem can scale to at most a few hundred machines, and may produce placement solutions that are far from optimal when system resources are tight. In this paper, we propose a new algorithm that can produce within 30 seconds high-quality solutions for hard placement problems with thousands of machines and thousands of applications. This scalability is crucial for dynamic resource provisioning in large-scale enterprise data centers. Our algorithm allows multiple applications to share a single machine, and strives to maximize the total satisfied application demand, to minimize the number of application starts and stops, and to balance the load across machines. Compared with existing state-of-the-art algorithms, for systems with 100 machines or less, our algorithm is up to 134 times faster, reduces application starts and stops by up to 97%, and produces placement solutions that satisfy up to 25% more application demands. Our algorithm has been implemented and adopted in a leading commercial middleware product for managing the performance of Web applications.

Shared resource network-aware impact determination algorithms for service workflow deployment with partial cloud offloading

“…Application placement is used to determine the location of applications within networks [10,11,12] or clouds [13,14,15], taking into account the demand for each application. Application placement is used to coordinate applications.…”

“…Our approach however incorporates CPU limitations and can be applied to varying network layouts. In [11] and [17], an application placement algorithm based on a conversion to a network problem is discussed, but the physical network is not taken into account. Our work by contrast specifically focuses on the underlying network.…”

“…Because of this, two additional constraints, Equation (11) and Equation (12) are added to restrict respectively outgoing and incoming flows. In this formulation, a 1 and a 2 represent the source and the sink service of the commodity c. ∀c ∈ C : ∀a ∈ A : ∀s ∈ S : f((a, s), c) = 0 (unless a = a 1 ) (11) ∀c ∈ C : ∀a ∈ A : ∀s ∈ S : f((s, a), c) = 0 (unless a = a 2 )…”

Allocating resources for customizable multi-tenant applications in clouds using dynamic feature placement

“…In particular, we focus on extending the generic application placement problem [16] and cloud application placement problem [17,18,19] to incorporate both multi-tenancy and software variability. The approach we use for multitenancy uses component-based applications composed using a Service-Oriented Architecture (SOA), making the relations between components another important consideration.…”

“…Application placement algorithms typically focus on server CPU and memory resources [19,26,27,28] or bandwidth limitations [29]. In this article we make use of a generalized approach where arbitrary resources of different types can be used.…”