Joint Dimensioning of Server and Network Infrastructure for Resilient Optical Grids/Clouds

Abstract

We address the dimensioning of infrastructure, comprising both network and server resources, for large-scale decentralized distributed systems such as grids or clouds. We design the resulting grid/cloud to be resilient against network link or server failures. To this end, we exploit relocation: Under failure conditions, a grid job or cloud virtual machine may be served at an alternate destination (i.e., different from the one under failurefree conditions). We thus consider grid/cloud requests to have a known origin, but assume a degree of freedom as to where they end up being served, which is the case for grid applications of the bag-of-tasks (BoT) type or hosted virtual machines in the cloud case. We present a generic methodology based on integer linear programming (ILP) that: 1) chooses a given number of sites in a given network topology where to install server infrastructure; and 2) determines the amount of both network and server capacity to cater for both the failure-free scenario and failures of links or nodes. For the latter, we consider either failure-independent (FID) or failure-dependent (FD) recovery. Case studies on Europeanscale networks show that relocation allows considerable reduction of the total amount of network and server resources, especially in sparse topologies and for higher numbers of server sites. Adopting a failure-dependent backup routing strategy does lead to lower resource dimensions, but only when we adopt relocation (especially for a high number of server sites): Without exploiting relocation, potential savings of FD versus FID are not meaningful.

Resilience options for provisioning anycast cloud services with virtual optical networks

“…Hence, the classical notion of a (source,destination)-based traffic matrix disappears [14]. While we previously developed scalable methods solve the resilient anycast dimensioning problem [15]- [17], that work did not consider synchronization between distinct working and backup data center locations (as opposed to the current paper). We believe this is the first work to discuss this in depth: previously we only sketched initial ideas in [6].…”

“…We will adopt a column generation (CG) approach, as this tends to be a highly scalable solution methodology (e.g., its application in [15], [17]). That means that we will divide the model into a Restricted Master Problem (RMP) and a Pricing Problem (PP).…”

“…We will protect against single link failures as well as single data center failures. (Extension to generic failures modeled as shared risk groups is fairly trivial, e.g., using a similar approach as [17]. )…”

Anycast (re)routing of multi-period traffic in dimensioning resilient backbone networks for multi-site data centers

“…Furthermore, this anycast principle can also be used for resiliency: if a failure affects either the DC or the (optical) network connection towards it, backup can be provided at an alternate location. Previous work has studied the potential benefits (in terms of server and network resource capacity) of exploiting anycast in network dimensioning problems with static traffic (e.g., [2]). …”

Time-varying resilient virtual network mapping for multi-location cloud data centers

“…In the context of optical grids and cloud computing, the design of resilient networks has been studied within the anycast framework, see, e.g., [4,5]. However, we are not aware of any work on time-varying anycast traffic exploiting protection rerouting or reconfiguration.…”