Virtual Network Resource Management In Multi-tenant Data Centers

Multi-tenant data centers(MTDCs)have gained increasing popularity in modern business services due to their effective and secure sharing of resources through the utilization of hypervisor-based mechanisms to isolate computing resources among co-located tenants on the same end host.However,current network virtualization techniques employed in MTDCs are often static,inflexible,and inefficient,leading to higher data center costs and suboptimal network resource utilization.Furthermore,tenants in MTDCs may not have complete control over the underlying infrastructure and virtual network resources,which could expose security risks for co-allocated tenants in the network.For instance,VxLAN is widely adopted for network virtualization in data centers;however,it does not offer tenant selfcontrol over virtual resources in MTDC.Network slicing has emerged as a solution by offering separate controllers for each slice.However,accurately specifying tenants poses challenges.To address this,the introduction of virtual MAC addresses can enhance tenantawareness in network slicing for MTDC environments.Moreover,virtual resource allocation typically involves path selection between source and destination,along with Traffic Shaping and Policing.To enhance virtual resource allocation,we can defining virtual resources as virtual ports within the MTDC network and assigning unique virtual MAC addresses to each tenant.This allocation enables individual virtual ports for tenants,facilitating the application of network policies and enables the implementation of tenant-aware source routing,reducing forwarding decision time in physical substrate nodes.Therefore,it is imperative to develop a new network virtualization and virtual network resource allocation mechanism that can efficiently manage virtual resources while ensuring control for both tenants and MTDC operators.This research proposes three innovative contributions for virtual network resource management in MTDC.The essential key features of each contribution are discussed below:■ Tenant-aware network slicing using novel vMAC:a novel virtual MAC address(vMAC)is introduced to encode tenants’ Virtual Data Center(VDC)identifier enabling the network to be aware of each tenant in MTDC.This approach stands in contrast to the current method of using MAC addresses,IP addresses,or TCP/UDP ports to define network slices,which do not inherently identify tenants within the network.Therefore,the proposed mechanism allows for the creation of a customizable virtual network topology,termed as vTopology,with separate controllers for each tenant,providing tenant isolation,improving path resilience,and preventing misconfiguration inside the tenant’s network from impacting co-tenants in MTDC.Additionally,vMAC address structure allows for expanding the MTDC hosting capability to 220 VDC.Furthermore,this research introduces an Optimal Back-Forwarding Path(OBFP)to overcome data traffic forwarding problems between virtual ports of one single virtual switch located over different physical switches and minimize forwarding decision time.■ Tenant-aware virtual network resource allocation:a novel approach for tenantaware allocation that utilizes the vMAC address to match each tenant in the network and allocate the VDC of each tenant based on the virtual port allocated in substrate nodes in the MTDC network.Although virtual resource allocation mechanisms currently employ software-defined network(SDN)controllers to allocate virtual network resources in data center SDN networks,the SDN controller lacks awareness of each tenant in the network,which creates complications for virtual resource management in MTDCs for MTDC operators and risks tenant resource sharing.Employing the proposed vTopology module in the SDN controller,allows the SDN controller to be aware of each tenant in the MTDC,facilitating virtual network resource allocation with tenant awareness.In addition,this study puts forth source routing and flow matching policies based on vMAC address that facilitate the forwarding of each tenant’s traffic along a predetermined path from source to destination.The proposed method has the potential to enhance the management of virtual network resources,enable traffic classification on a per-tenant basis,and promote the equitable utilization of physical nelwork resources.As a result,it may lead to improved efficiency and acceptance rates in the allocation and distribution of virtual resources.■VDC fair bandwidth allocation:a solution is proposed to address the significant challenge of uneven bandwidth allocation within and between VDCs in an MTDC environment.This is a major issue faced by cloud computing and MTDC administrators.The implementation of bandwidth guarantees in an MTDC faces several challenges,such as ensuring scalability and providing unique bandwidth constraints to virtual nodes within a VDC.Hence,in this research,the FairShare contribution is proposed to ensure equitable bandwidth allocation among virtual nodes of VDC,as well as between VDCs.This approach aims to address the challenge of dropped virtual links within a VDC and optimize the number of active virtual links and the substrate path set for each VDC.The goal is to minimize overall bandwidth wastage by enhancing fairness in load utilization among tenants’ virtual nodes and decreasing waste bandwidth costs.In summary,this research presents three contributions for enhancing virtual resource management in MTDCs.Firstly,the vTopology module introduces a new vMAC to encode tenants’ VDC identifier,allowing the network to be aware of each tenant.Secondly,the Tenant-aware resource allocation(TARA)approach utilizes the vMAC address to match each tenant and allocate VDCs based on virtual ports allocated in substrate nodes.Finally,the FairShare model ensures fair bandwidth allocation among virtual nodes of VDCs and between VDCs,minimizing overall bandwidth wastage and reducing associated costs.