Network Virtualization – The Last frontier in Data Center Virtualization
Woj 270003B7NV Visits (5019)
Server, storage, and network… the three horsemen that make up the core physical components of the modern day data center. The nuance in this statement is that ‘the modern day data center” is not a physical one. Virtualization has become an integral technology used across the three physical aspects of the data center to allow clients to more efficiently leverage their physical investments. Server (or sometimes called compute or runtime) virtualization is probably the most widely deployed implementation of virtualization, followed by storage and then network. But, network virtualization has many parallels with server and storage virtualization and the seemingly recent interest in network virtualization is driven by widespread deployment of server virtualization. The basic function of network virtualization is that it enables the creation of private networks for individual applications or customers that give them isolation from other applications and customers that use that network. From that perspective, network virtualization is already very widely used, and has been for many years. Telecommunications providers have been providing virtual private networks for their users using a shared network infrastructure for decades, although the technologies used to provide them have evolved over the years. For instance x.25, Frame-relay, ATM, and more recently MPLS are some of the technologies that carriers have deployed to be able to virtualize their shared networks for improved efficiencies and better quality of service. Today, MPLS based virtual private networks are widely used but we see services based on the older technologies linger on.
Now, if we take a closer look at corporate IT infrastructure and data centers (i.e. non-Telco), virtual private networks are also widely deployed in the form of virtual LANs with Ethernet Switches. They are used to provide finer levels of isolation and protection for different tenants such as departments or applications - for example, Finance applications, HR applications, and so on. So, what’s the big deal and why all the new excitement?
traditional model of supporting virtual networks has limitations which are
being stressed and exposed with the increasing sophistication of server
virtualization and the needs of Cloud deployments. Within each server that is virtualized,
there is virtual network switch that is part of the hypervisor. Virtual server mobility is placing new
demands on networks as the server’s logical network address has to be retained
even as its physical location changes.
An added dimension here is that virtual switches are typically managed
The new excitement in network virtualization is coming from two areas, both aimed at making the network more flexible and agile:
(1) growing capabilities of Virtual Network Switches that are embedded in
Server hypervisors, and in particular the ability to form Virtual Network
Overlays. A Virtual Network Overlay
enables Virtual System administrators to deploy Virtual Servers and
applications with much greater flexibility and without requiring frequent
changes to physical network switches, thereby reducing the need for coordinated
network changes that was previously required.
In the coming years, we will see organizations deploying these new network virtualization technologies. The adoption rates, however, will vary based on the needs of the organizations and their risk profile for new technologies. Although these technology approaches each have their pros and cons, a common thread here for Software Defined Networks is the need for a Network API to provide network users with the means to create and change network configurations to adapt to user/workload requirements. IBM is working with industry partners and organizations such as OpenStack to develop a network service abstraction and API for Cloud and other applications. We are also extending our management capability to support and exploit network virtualization in various forms.