Sara Ayoubi

MINTED: M ulticast VI rtual N e T work E mbedding in Cloud Data Centers With D elay Constraints

Network virtualization is regarded as the pillar of cloud computing, enabling the multi-tenancy concept where multiple Virtual Networks (VNs) can cohabit the same substrate network. With network virtualization, the problem of allocating resources to the various tenants, commonly known as the Virtual Network Embedding problem, emerges as a challenge. Its NP-Hard nature has drawn a lot of attention from the research community, many of which however overlooked the type of communication that a given VN may exhibit, assuming that they all exhibit a one-to-one (unicast) communication only. In this paper, we motivate the importance of characterizing the mode of communication in VN requests, and we focus our attention on the problem of embedding VNs with a one-to-many (multicast) communication mode. Throughout this paper, we highlight the unique properties of multicast VNs and its distinct Quality of Service (QoS) requirements, most notably the end-delay and delay-variation constraints for delay-sensitive multicast services. Further, we showcase the limitations of handling a multicast VN as unicast. To this extent, we formally define the VNE problem for Multicast VNs (MVNs) and prove its NP-Hard nature. We propose two novel approach to solve the Multicast VNE (MVNE) problem with end-delay and delay variation constraints: A 3-Step MVNE technique, and a Tabu-Search algorithm. We motivate the intuition behind our proposed embedding techniques, and provide a competitive analysis of our suggested approaches over multiple metrics and against other embedding heuristics.

Protection plan design for cloud tenants with bandwidth guarantees

In cloud data centers, where hosted applications share the underlying network resources, network-bandwidth guarantees have been shown to improve predictability of application performance and cost. However, recent empirical studies have also shown that often data center devices and links are not all that reliable and that failures may cause service outages, rendering significant revenue loss for the affected tenants, as well as the cloud operator. Accordingly, cloud operators are pressed to offer both reliable and predictable performance for the hosted applications. While much work has been done on solving both problems separately, this paper seeks to develop a joint framework by which cloud operators can offer both performance and availability guarantees for the hosted tenants. In particular, this paper considers a simple model to abstract the bandwidth guarantees requirement for the tenant and presents a protection plan design which consists of backup virtual machines placement and bandwidth provisioning to optimize the internal data center traffic. We show through solid motivational examples that finding the optimal protection plan design is highly perplexing, and encompasses several constituent challenges. Owing to its complexity, we decompose it into two subproblems, and solve them separately. First, we invoke a placement subproblem of the minimum number of backup VMs and then we attempt to find the most efficient correspondence between backup and primary VMs (i.e., protection plan) which minimizes the bandwidth redundancy. Our numerical evaluation shows that our two-step method is both scalable and accurate; further, it performs much better than a baseline method where placement of backup VMs is done at random.

Multicast Tree Repair and Maintenance in the Cloud

Network virtualization enables the multi-tenancy concept where multiple tenantss services can cohabit the same substrate network and share its resources. With multi-tenancy, the problem of allocating resources to the various tenants emerges as a challenging problem. This former is commonly known as the virtual network embedding problem (VNE), which has attracted numerous effort from the research industry due to its NP-Hard nature. Yet, most of the existing work overlook the various modes of communication a virtual network (VN) can exhibit, assuming it is always a one-to-one communication between virtual machines (VMs). The recent technological advancements (such as Software Defined Networks (SDNs)) have paved the way for efficient multicast in data center networks, thereby leveraging the support of services and applications which multicast data in large volumes. While much work has been devoted for studying the problem of multicast virtual network (MVN) embedding in the cloud, little attention has been paid to investigating the impact of failure on this service class. In this paper, we study the impact of facility node failure on embedded MVNs, and introduce a novel post-failure restoration scheme to repair failed MVNs while maintaining their requested Quality of Service (QoS). Our numerical results prove that our suggested method achieves encouraging restoration ratio in considerably fast execution time.

Towards Scalable Traffic Management in Cloud Data Centers

Cloud Computing is becoming a mainstream paradigm, as organizations, large and small, begin to harness its benefits. This novel technology brings new challenges, mostly in the protocols that govern its underlying infrastructure. Traffic engineering in cloud data centers is one of these challenges that has attracted attention from the research community, particularly since the legacy protocols employed in data centers offer limited and unscalable traffic management. Many advocated for the use of VLANs as a way to provide scalable traffic management, however, finding the optimal traffic split between VLANs is the well known NP-Complete VLAN assignment problem. The size of the search space of the VLAN assignment problem is huge, even for small size networks. This paper introduce a novel decomposition approach to solve the VLAN mapping problem in cloud data centers through column generation. Column generation is an effective technique that is proven to reach optimality by exploring only a small subset of the search space. We introduce both an exact and a semi-heuristic decomposition with the objective to achieve load balancing by minimizing the maximum link load in the network. Our numerical results have shown that our approach explores less than 1% of the available search space, with an optimality gap of at most 4%. We have also compared and assessed the performance of our decomposition model and state of the art protocols in traffic engineering. This comparative analysis proves that our model attains encouraging gain over its peers.

Pro-Red: Prognostic redesign of survivable virtual networks for cloud data centers

This paper deals with the problem of proactive survivability of Virtual Networks (VNs) residing in a cloud data center. In all of the previous work, the protection schemes consists of augmenting the VNs with a pre-determined number of backup nodes. Further, to reduce the amount of provisioned resources, various backup resource sharing schemes are introduced, which are mainly employed when mapping the augmented VN onto the substrate network. This renders the existing redesign techniques agnostic to the backup resource sharing in the substrate network, and highly dependent on the efficiency of the adopted mapping approach. In this paper, we swerve from this dogmatic approach, and introduce Pro-Red, a novel prognostic redesign technique that is capable of foretelling (encouraging) the backup resource sharing in the substrate network, prior to the embedding phase. Our numerical results prove that this redesign technique achieves lower-cost mapping solutions and greatly enhances the attainable backup sharing, boosting the overall networks admissibility.

Surviving link failures in multicast VN embedded applications

Virtual network embedding (VNE) is defined as the allocation of network resources to multiple virtual networks (VNs) and is recognized to be a challenging task to perform efficiently. Virtual network survivability is a new term that describes the measures taken to provide a failure-proof VN against physical link and/or node failure. Indeed, a single link or node failure in a substrate network can bring down multiple hosted VNs, i.e., the ones that utilize that failed link or node. As such, virtual network survivability becomes an essential part of VNE. While much work has been dedicated to studying the impact of a variety of failure cases in a VN, little attention has been directed towards studying the link failure impact on multicast virtual network (MVN) applications, which principally restrict end-to-end delay and delay variation measures. In fact, most of the introduced survivability schemes adopt protection techniques by reserving backup resources prior to embedding, which inevitably leads to under-utilization of the network resources. In this paper, we first investigate the impact of physical link failure on MVNs. Then, we introduce a novel recovery approach to restore MVNs while considering their end-delay and delay variation requirements. Simulation experiments prove that our recovery technique achieves good restoration ratio in considerably fast execution time and low link mapping cost with little impact on the admittance ratio.

A Reliable Embedding Framework for Elastic Virtualized Services in the Cloud

This paper proposes a novel framework for managing the resource provisioning of reliable virtual networks (VN) in the cloud. This includes handling the placement of VN requests while providing availability guarantees, as well as reconfiguring/adapting their placement as their request changes over time. This is particularly interesting for services with periodic resource demands. Given the heterogeneous failure rates of physical network components, the placement and reconfiguration must ensure that the selected hosts for each VN meets its availability requirements. The existing work on availability-aware VN placement has overlooked the case of “availability over-provisioning,” as well as the fact that VN requests are subject to change over time. To this extent, we propose a novel framework that consists of two main modules; JENA: a tabu-based availability-aware resource allocation (embedding) module for VNs that achieves “just-enough” availability guarantees, and ARES: a reliable reconfiguration module to adapt the embedding of hosted services as they scale. Further, we introduce the concept of “protection-domains” and “protection-policies” to equip our proposed modules with the ability to augment services with redundant/backup nodes to enhance their reliability. Our numerical results show that our framework enhances networks admissibility (with 33% lower blocking compared to existing work), and in return increases the cloud providers long term revenue, compared to peer and benchmark algorithms.

Towards Promoting Backup-Sharing in Survivable Virtual Network Design

In a virtualized infrastructure where multiple virtual networks (or tenants) are running atop the same physical network (e.g., a data center network), a single facility node (e.g., a server) failure can bring down multiple virtual machines, disconnecting their corresponding services and leading to millions of dollars in penalty cost. To overcome losses, tenants or virtual networks can be augmented with a dedicated set of backup nodes and links provisioned with enough backup resources to assume any single facility node failure. This approach is commonly referred to as Survivable Virtual Network (SVN) design. The achievable reliability guarantee of the resultant SVN could come at the expense of lowering the substrate network utilization efficiency, and subsequently its admissibility, since the provisioned backup resources are reserved and remain idle until failures occur. Backup-sharing can replace the dedicated survivability scheme to circumvent the inconvenience of idle resources and reduce the footprints of backup resources. Indeed the problem of SVN design with backup-sharing has recurred multiple times in the literature. In most of the existing work, designing an SVN is bounded to a fixed number of backup nodes; further backup-sharing is only explored and optimized during the embedding phase. This renders the existing redesign techniques agnostic to the backup resource sharing in the substrate network, and highly dependent on the efficiency of the adopted mapping approach. In this paper, we diverge from this dogmatic approach, and introduce ProRed, a novel prognostic redesign technique that promotes the backup resource sharing at the virtual network level, prior to the embedding phase. Our numerical results prove that this redesign technique achieves lower-cost mapping solutions and greatly enhances the achievable backup sharing, boosting the overall networks admissibility.

Offering Resilient and Bandwidth Guaranteed Services in Multi-tenant Cloud Networks: Harnessing the Sharing Opportunities

The sharing of computing and networking resources in the cloud is challenged by several obstacles, such as providing bandwidth guarantees for a predictable performance of the hosted applications, as well as maintaining the availability of their services following outages. Therefore, the wide scale adoption of this emerging computing paradigm remains highly dependent on overcoming these challenges. In fact, a lack of bandwidth guarantees extends the completion time for jobs, thus increasing expenses for clients paying for their time of use. In addition, outages in data centers may result in severe revenue losses for both, the cloud operators and their clients alike. To overcome these challenges, cloud operators should be empowered with a strategic design plan that is able to guarantee resilient and predictable performance for hosted applications. Such a plan consists of provisioning additional backup resources (e.g.virtual machines, bandwidth) while ensuring efficient network bandwidth utilization. In this work, we study the design of various facets of such a plan. Namely, we exploit several bandwidth sharing opportunities in multi-tenant cloud networks while offering resilient and bandwidth guaranteed services. In contrast to previous works which target cloud clients satisfaction, we focus on optimizing network bandwidth utilization in order to increase the cloud operators revenues while maintaining such bandwidth allocation transparent to the clients. Through several motivational examples, and numerical studies, we highlight the sharing opportunities and show that they are able to increase cloud operators revenues by an average of 21.4% while providing up to 50% of bandwidth gain in the network.

Energy-Aware Placement and Scheduling of Network Traffic Flows with Deadlines on Virtual Network Functions

Hardware MiddleBoxes represent a vital part in todays networks. Despite their important roles, they are accompanied by several problems, namely, their lack of flexibility, high capital and operation expenditures, and power consumption. Network Function Virtualization is one promising solution to address these problems. This trend replaces the MiddleBoxes by software-based entities. Indeed, these Virtual Network Functions promise to alleviate the numerous disadvantages brought by their hardware counterparts. One of these most serious issuesis the steadily increasing power consumption. Studies suggest that the Virtual Network Functions will reduce the electricity costs needed to turn on and operate the hardware functions. In order to further optimize the power consumption of the network, an efficient framework, capable of placing and scheduling traffic on these VNFs, is needed. Such a framework allows to optimally place and schedule the flows to be serviced, and placing theunused servers in energy saving modes. In this article, we assume VNFs are already placed on physical machines, each hosting a subset of the functions. We consider traffic flows with deadlines. We aim at assigning and scheduling flows to VNFs in the most energy efficient manner. We formulate this problem mathematically and, owing to its complexity, present an efficient algorithmic method for solving the problem. We compareour heuristic with two other approaches, one of which aims to minimize the makespan, and the other to minimize number of servers used. We show that our heuristic combines the advantages of both approaches and generates better results by consuming up to 31.3% and 46.1% energy less than other two approaches respectively.