Introduction

The Internet is rigid, with lots of inflexible connections and makeshift network overlays. A new solution with open protocols, expandable architecture, flexible, easy inter-connection, and scalability that solves almost all existing network problems is obviously needed. Network virtualization allows multiple service providers to make use of the same system while being isolated from each other, and deploys various services on based on demand by utilizing and sharing network resources provided by separate multiple infrastructure providers. A virtual network seems to be a network, but is virtual in the sense that it is not necessarily a physical network of cables or wireless devices knotted together.

VMware Training – Resources (Intense)

A networking environment is virtualization-compliant if it is able to support the co-existence of multiple virtual networks on the same physical infrastructure. In a virtualized network environment, various service providers can deploy and manage innovative services on virtual networks for the end users by efficiently sharing and utilizing network resources without the normal limitations found in the existing Internet environment, such as deployment glitches, geographical constraints, or cumbersome system configurations.

The traditional Internet/telephony service is broken and divided into two separate and independent arms, namely:

  1. The infrastructure providers who assemble and manage physical ICT infrastructure. They have the ability to host multiple service providers on the same hardware.
  2. The service providers, who create virtual networks by creating and deploying network services on the platforms provided by the infrastructure providers. Multiple service providers can be hosted on the same physical infrastructure without affecting their operations.

The above arrangement makes economic sense, as it leads to immense reductions in cost for both parties and also promotes competition. Smaller network operators who do not have the financial muscle to roll out their own infrastructure can deploy their services on leased large-scale network infrastructure and network services can be deployed on demand. The functionalities that each service provider can deploy over the virtualized infrastructure include simple telephony service, connectivity solutions, cloud service, and other advanced network innovations.

Types of Virtualization

Infrastructure Virtualization

This combines computer and network hardware resources together with software solutions to create a virtual network. Examples include virtual local area network (VLAN) and virtual private network (VPN). It also includes network storage virtualization, where users can store and access data on storage devices and networks such as a storage area network or RAID through applications deployed on the platform.

System Virtualization

This is meant to solve the problem of integration among system devices, such as desktops, laptops, smart phones, tablets, etc., by ensuring that all devices can inter-connect and are compatible with each other.

Software Virtualization

This solves the platform problem among executable software programs from different software producers so that programs that are written in one language can be interpreted and executed on various platforms. It also ensure that software applications run virtually and are independent of servers so that they can be accessed anywhere without need the software program to be installed a present physical server as in SaaS (software as a service).

Network Virtualization Historical Perspective

This perspective will be examined under the following networks

  • Virtual local area networks (VLAN)
  • Virtual private networks (VPN)
  • Overlay networks.

1. Virtual Local Area Network

A virtual local area network (VLAN) is a group of servers or hosts under a common interest that are logically connected together to with a single broadcast domain irrespective of their physical connectivity characteristics. All frames in a VLAN have a VLAN ID in the MAC header. Switches on VLAN use both the destination MAC address and the VLAN ID to forward frames. Multiple switches from multiple VLANs can be connected together using trunking, which allows data to be carried over a single link between multiple switches. Since VLANs are logically configured with software, they are flexible in terms of network administration, management, security, and re-configuration and are more cost-effective.

2. Virtual Private Network

A virtual private network (VPN) is a dedicated communications network of one or more organizations that are scattered over multiple locations and are connected through tunnels over public communication networks (e.g., the Internet). Each VPN site contains customer edge (CE) devices such as hosts or routers, which are connected to one or more provider edge (PE) routers.

There are different types of VPNs, including:

  1. Layer 3 VPN (L3VPN)

    makes use of layer 3 tunneling protocols to transfer data between the distributed VPN sites. They include the carrier protocol (e.g., IP) used by the service provider network to carry the VPN packets; the encapsulating protocol used to wrap the original data in order to secure it; and the passenger protocol, which is the original data that is transmitted. Sender CE devices envelop the passenger packets and route them into the carrier network or Internet. When the data packets reach the receiver CE devices, they are extracted and injected into receiver networks.

  2. Layer 2 VPN (L2VPNs)

    provides a two-way end-to-end connection by transporting Layer 2 frames between various sites. It can also provide an IP-only LAN-like service (IPLS), in which only IP packets (IPv4 or IPv6) are carried.

  3. Layer 1 VPN

    enables multiple virtual client networks to operate over a common Layer 1 core infrastructure. The main high point of L1VPN is its multi-service backbone, where many companies can offer their own services with payloads of any layer. This allows each company to have its own independent IP address, separate Layer 1 resource view, independent policies, and complete isolation from other service providers.

3. Overlay Networks

An overlay network is a network built on top of one or more existing physical networks. Overlays do not disrupt or cause any changes to the underlying network. As a result, overlay networks are used to deploy new innovations and upgrades on the network. Overlays on an existing network are usually implemented in the application layer. However, it is also possible to implement it at lower layers of the network. The Internet started out as an overlay on top of existing telecommunication networks.

Parties Involved in Network Virtualization

Infrastructure Provider (InP):

The infrastructure provider (InP) deploys, powers, and manages the physical network resources. Different service providers can roll out their services on the infrastructure through programmable logical interfaces.

Service Providers (SP):

They lease infrastructure resources from one or multiple InPs and they create, deploy, and manage customized virtual network services to the end users. A service provider can also provide network services to other service providers by partitioning its resources to create subsidiary or child virtual networks. This is known as recursion or nesting.

Brokers:

They are mediators who broker deals between the major players.

End Users:

End users in the network virtualization model are similar to existing Internet service subscribers and they also have the luxury to choose one or multiples service providers.

Network Virtualization Topography

Design Goals

Over the decades, researches on network virtualization have progressed from focusing on a particular objective to covering multiple ones. While the past projects are focused individually on objectives such as security, flexibility, or programmability, recent virtual network projects focuses on most of them in a single objective. In order to produce network virtualization, each of the following design objectives should be met.

1. Flexibility and Heterogeneity

All network virtualization projects must have certain amount of flexibility and heterogeneity, which is determined by the underlying networking technology and the layer at which virtualization is administered. Dependence on specific technologies reduces the amount of flexibility in the network. The lower layer virtualization is implemented, the more flexible and heterogeneous the network will be. Service providers should have the freedom to arbitrarily choose network topology, routing, and forwarding functionalities they think will best suit their business needs and there should be no need for cumbersome co-ordination with other players in the network environment.

2. Manageability

Manageability in network virtualization has to be addressed at both micro and macro levels by clear separation of policy from mechanism and absolute accountability of the SPs and InPs. By separating SPs from InPs, network virtualization will simplify network management tasks and introduce accountability at every strata of networking.

3. Isolation

Network virtualization must ensure isolation between co-existing SPs to improve security, privacy, and fault-tolerance. Bugs and mis-configurations emanating from one network should remain within the network and not affect the others.

4. Programmability

The network must be easily programmed to ensure flexibility, heterogeneity, and manageability. This will enable SPs to deploy customized protocols and diverse services. The focus should be more on enabling secure programming that is easy and effective without being vulnerable to threats.

5. Scalability

The ability to expand network service according to demand is one of the fundamental principles of network virtualization. Scalability comes as an indispensable requirement of this principle. InPs must be able to scale up or increase network capacity during peak demand periods without affecting the quality of service and performance.

6. Experimental and Deployment

Experimentation ensures that services can be deployed directly to the end users even though the service is in the testing phase. Experimental projects such as PlanetLab, GENI, VINI, and FEDERICA were conducted with real traffic and active network conditions.

PlanetLab GENI

FEDERICA

PlanetLab is a global research network that supports the development of new network services. Since the beginning of 2003, more than 1,000 researchers at top academic institutions and industrial research labs have used PlanetLab to develop new technologies for distributed storage, network mapping, peer-to-peer systems, distributed hash tables, and query processing.PlanetLab currently consists of 1353 nodes at 717 sites.
https://www.planet-lab.org/
GENI is a new, nationwide suite of infrastructure supporting “at scale” research in networking, distributed systems, security, and novel applications. It is supported by the National Science Foundation, and available without charge for research and classroom use.
https://www.geni.net/
FEDERICA is a two-and-a-half-year European project to implement an experimental network infrastructure for trialing new networking technologies.
This infrastructure is intended to be agnostic as to the type of protocols, services and applications that may be trialed, whilst allowing disruptive experiments to be undertaken. The aim is to develop mechanisms that will allow such experiments to be run over existing production networks without adverse effect.

http://www.fp7-federica.eu/

7. Heterogeneity

Heterogeneity of the networking technologies, such as optical, wireless, and sensor; cross-platform domains; as well as user devices is paramount. These should be able to interconnect without the need for any technology-specific solutions.

8. Legacy Support

Network virtualization projects must ensure backward compatibility and integration, since we can’t just do away with previous technologies. It takes a long time for some devices to be phased out. Also, new technologies are usually very costly for medium and small businesses to acquire. Previous technologies must be able to connect to the network without need for special configuration or protocol.

Advantages of Network Virtualization

Lower technology costs as services providers do not need to acquire and assemble expensive equipment before they can roll out.

  • Heterogeneity and legacy support will ensure higher utilization rates.
  • It is less complex and easy for users to utilize.
  • Higher quality of service.
  • Supports dynamic migration.
  • Isolation ensures privacy and security.
  • Scalable and easy to expand.

Key Research Challenges

Several technical challenges are currently militating against rapid network virtualization. Such related problems include interfacing, signaling, bootstrapping, implementation of routers, protocols, and embedding of virtual networks on shared physical infrastructure, as well as effective management of resource. Others include how to handle network failure mobility and security. They are summarized below:

1. Interfacing

Infrastructure providers must provide well-defined interfaces to allow service providers to deploy and express their requirements. Other interfaces between SPs and end users and between InPs and SPs must also be programmed and standardized.

2. Signaling and Bootstrapping

Before a virtual network can go live, a SP must have network connectivity to the InPs in order to be able to deploy their services. Hence network connectivity is a prerequisite. Bootstrapping capabilities should also be in place to allow service providers to synthesize the network resources allocated to them. These would require the emergence of another network player that will provide connectivity to handle these issues, or an out-of-band mechanism.

3. Resource Allocation and Network Topography

In order to ensure efficient allocation and management of resources, InPs must be able to determine the network status with regard to capacity of nodes, links, and the topology of the networks they manage. Also, two different InPs must also be able to exchange links to make an end-to-end virtual network. This will enable VNs to interact with themselves.

4. Usage Policing and Admission Control

Infrastructure providers must ensure that resources are not over-allocated to a particular service provider by and implementing admission control and performing accurate accounting through the use of algorithms to ensure optimal performance of the network.

5. Virtual Nodes and Links

ICT equipment vendors have been promoting virtual routers and switches as a means of simplifying core network architecture, thereby decreasing capital expenditure (CAPEX), and for VPN purposes. There is now more focus on increasing the number of virtual nodes a router can hold. Though the ability to create tunnels over multiple physical links is already being utilized by VPNs, similar idea can be extended to virtual networks too to reduce encapsulation and multiplexing costs.

6. Naming and Addressing

As a result of the heterogeneous nature of the virtual networking environment, there is a new challenge of how to issue names and IDs and addresses to multiple users who might want to move from one SP to the other without forfeiting their IDs. It is the same scenario when a SP deploys on different InPs, as it wouldn’t make sense for the same SP to bear different IDs. Therefore incorporating support for such heterogeneity in multiple dimensions is a fundamental challenge in the context of network virtualization.

7. Mobility and Dynamism

The virtual network environment is highly motile. SPs may desire to extend their services to other geographical areas while users often move from one place to another. To pinpoint the exact location of a resource or user at a particular moment and forward data packets accordingly is a cumbersome virtualization challenge that needs an effective solution.

8. Failure Handling

Bugs and failures that crop up in the in the underlying physical network components may affect the performance of the whole network. Prompt detection, isolation and remediation are very important in any networking environment.

9. Security and Privacy

Tunnels and encryption used to isolate virtual networks may become vulnerable to threats if they are not properly secured. The InPs and SPs have to ensure that there is very little downtime if a successful attack occurs in order to mitigate the effects of the attack.

10. Interoperability Issues

Overlaying virtual networks can spread across many organizations, political domains, and networking technologies from different vendors. Each networking technology has its own set of unique characteristics and therefore requires specific solutions to ensure they can all operate in harmony on the virtual network. Enabling virtualization in face of these challenges requires innovative solutions for interoperability to ensure smooth interactions between many contrasting underlying infrastructures, while providing a generic and transparent management interface for SPs to easily deploy their services remains a herculean task.

11. Network Virtualization Economics

In normal Internet economics, bandwidth is the main commodity of interest. However, in network virtualization, virtual nodes are equally important. InPs are the producers here, while the SPs are the buyers/sellers as they have to pay the InPs for using their platform and also sell to the end users.

In virtual networking, there are two types of markets; centralized and decentralized markets. Centralized markets are efficient but they are rigid and not scalable. Decentralized markets, on the other hand, are scalable but are vulnerable to threats to their open nature. To strike a balance between these two marketplaces becomes a dilemma.

References

  • N.M. Mosharaf Kabir Chowdhury, Raouf Boutaba, “A Survey of Network Virtualization,” University of Waterloo Technical Report CS-2008-25, Oct. 2008.
  • Kilian Rausch, Network Virtualization – An Overview; Department for Computer Science, Technische Universität München
  • George N. Rouskas, Network Virtualization: A Tutorial; Department of Computer Science North Carolina State University; March 2012.
  • Dr. Qingni Shen, Introduction to Virtualization; Peking University