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

Abstract

Some service providers offer services for which physical devices, terminals and servers are installed at the customer's site, e.g. in a hospital setting where physical terminals are installed in patient rooms and used for communication. These devices work together to provide the service, forming service workflows. Due to a recent trend where parts of the management system for such services is offloaded to cloud environments, such services can no longer be isolated in a private subnet, that is specifically dimensioned for its purpose. Part of the network flows must then pass over a larger section of the customer network, and the network load can increase as new services are added, or when services are upgraded. Because of this, it is important to determine whether there is sufficient network and server capacity in the customer network to add new service workflows before they are deployed, or before an existing service is upgraded, ensuring a sufficient level of quality can be guaranteed.In this article we focus on how the impact of service workflows can be determined, ensuring service workflows do not negatively impact each other's execution. In particular, we determine an impact analysis strategy to evaluate the degree to which a given set of service workflows can be guaranteed in a given network topology. As not all flows are continuously active, the approach is designed to support the sharing of network and server resources using a hierarchically specified resource sharing model. We then evaluate the quality of the resulting solutions using two use cases, and the execution speed of the designed algorithms. For the two evaluation use cases, we find that the developed hierarchical algorithm requires ±42% and ±52% less resources than an approach without resource sharing, without any workflow failures occurring during the simulations.

Hierarchical network-aware placement of service oriented applications in Clouds

“…In this paper we, by contrast, do take the underlying network into account, and the algorithms are designed hierarchically, ensuring they scale better. We use the principles described in [16], where we discussed the network impact of adding and removing services within a cloud environment, to formally define the communication between services. Our previous work focused solely on determining the network impact of service changes, and not on allocating the services.…”

“…In a subsequent optimization, an effort is made to increase the amount of bandwidth that is guaranteed. This two-stage approach to allocating bandwidth is similar to the one we described in [16].…”

“…The additional network constraints are similar to those introduced in [16]. Two additional constraints are needed to correctly represent network flows: edge capacity limitations, expressed in Equation (10), limiting the amount of bandwidth of network edges, and flow conservation, which ensures no flow is lost within the network, which is expressed in Equation (11) and Equation (12).…”

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

“…This is especially important if some features are allocated in different networks and part of the communication must pass over the Internet. Filters such as this can be created by determining the impact on the underlying network of instantiating these services, which we previously discussed in [32].…”

Algorithms for efficient data management of component-based applications in cloud environments

“…In our previous work [14], algorithms for determining the impact of the underlying network on the performance of component-based applications were studied. The focus was specifically on determining the impact of a given service mapping, whereas this work focuses on determining the service mapping itself.…”