Modeling and Evaluation of a Data Center Sovereignty

Abstract

Technology Sovereignty aims at protecting the interests of the consumers belonging to a sovereign state. The new laws on network sovereignty monitor and govern networks inside a state. They protect the constitutional rights of citizens and ensure data security. However, they do not focus on the infrastructure and hardware associated with these networks. Ensuring a robust end-to-end network fuels the need for reliable hardware manufacturers and an appropriate network architecture that can handle multiple failures efficiently. To understand technological sovereignty in communication networks, the Data Center Network (DCN) is targeted as it is a critical part of the digital society. Data centers are dedicated physical facilities that act as storage houses for large amounts of data. Though several fault-tolerance studies have been performed in DCNs, none have studied the role of hardware manufacturers in DCN Sovereignty. The unavailability of components from a manufacturer can lead to multiple failures. So, it is necessary to build a sovereign DCN without creating a dependency on the manufacturer(s). In this work, to evaluate the sovereignty of a DCN, (i) Multiple failure scenarios depending on manufacturer reliability are evaluated, and (ii) Design guidelines on how to choose the number of hardware manufacturers and how to arrange them in the DCN topology are presented.

Modeling and Evaluation of a Data Center Sovereignty with Software Failures

“…The inter-DCN traffic is considered as a 'special rack' [23] and the R DCN is set to 90% : 10%. All these values follow the model in our previous work [9].…”

“…This effect is evaluated numerically in terms of connectivity, and max-flow. This paper is a continuation of our previous work [9], which explored only HW failures. The paper is structured as follows: Section II briefs the related work.…”

“…Our previous work [9] on DCN sovereignty considered HW failures only. Three different approaches-homogeneous, heterogeneous and robustness surfaces were implemented to determine some basic guidelines for the DCN operators.…”

“…In the A LRS , the sequence blue-purple-yellow is repeated. From [9], the A LRS performs the best. Therefore, all simulations in this current work use A LRS .…”

“…This section discusses the DCN implementation in simulations. The additions to the model in [9] are discussed in detail. Servers, ToR, aggregate, and core switches are arranged in the Top of the Rack architecture throughout this study.…”