Making a Case for Network Virtualization

cloud380

Image copyright Johannes Kornelius

At the beginning of this century, server virtualization burst onto the IT scene and changed the way modern IT organizations think about server hardware. The basic concept behind server virtualization is that multiple “virtual” servers can be run on a single physical server. This consolidation of servers resulted in much higher utilization of physical servers, which dramatically reduced the capital expenditure (capex) costs required to provide IT services. Using fewer physical servers also lowered power, cooling and datacenter space requirements. Other indirect benefits were increased agility, since the IT department could provision a server in minutes or hours versus the weeks and months it used to take to procure, set up and turn on new physical servers.

By all accounts, server virtualization adoption has been extremely successful. One look at VMware’s revenue numbers and there’s no doubting this fact. In the early 2000s, the predominant question that many in IT asked was, “Why should I virtualize?” Today, the predominant question you’ll hear is, “Why can’t the servers be virtualized?”

The rise of self-service IT

We’re going through a similar paradigm shift today with self-service IT. Internal business units want the breadth of services and speed of provisioning that they can get outside of the firewall through cloud service providers like Amazon Web Services. In response, forward-thinking IT departments have been changing their traditional role into one that looks more like a service provider and have begun offering a full menu of solutions to their constituents. One of the staples on the menu is a private Infrastructure-as-a-Sevice (IaaS) cloud offering. A few concrete business drivers underlying this offering are:

  • Fine-grained control over the infrastructure which lowers risk and increases ability to deal with compliance concerns
  • A lower cost when compared to external services like Amazon
  • Lowered operational expenditures with regard to provisioning resources
  • A much faster provisioning speed than internal IT typically offers
  • Better disaster recovery options
  • Increased application availability

Elements of an IaaS cloud

There are four elements needed to build an IaaS cloud: a cloud management system, compute (also known as the hypervisor), storage and networking. The cloud management system handles all the provisioning and orchestration of the underlying compute, storage and network components. Examples of such systems are OpenStack, Citrix CloudStack, Eucalyptus and VMware’s Vsphere product.

For compute, storage and networking, cloud architects look for solutions that linearly scale out (adding new capacity incrementally) rather than scaling up (buying bigger devices). This approach keeps costs low by consistently maximizing utilization. Even if cost weren’t an issue, this scale out approach is highly favored because it increases availability and reduces service interruptions of your cloud. In a well-thought distributed IaaS design, a single large device would never be an integral component of your cloud. Adhering to distributed design philosophies is a key reason why cloud service providers can consistently achieve very high levels of availability.

Another item cloud architects look for are products that can integrate with the cloud management systems so that they are fully automated. Scaling out and automating compute is a known problem and all the cloud management systems solve it with ease. As for cloud storage, there are now great distributed options like Ceph, SolidFire and OpenStack Swift that linearly scale out and can be easily automated.

Networks are hard to provision and scale

These newly minted cloud architects are beginning to realize something that those in the cloud service provider business have known for a while. Network devices weren’t designed to be automated, and they definitely weren’t designed to be provisioned at the granularity and high-churn rate than IaaS clouds demand of them. Also, some network devices, instead of linearly scaling out as demand increases, tend to adopt a scale up model.

Networks aren’t flexible enough for cloud requirements

A common use case for an IaaS cloud is disaster recovery, which often requires the recreation of complex network topologies. This can be problematic because that typically would require the physical network to be purpose built for that specific disaster recovery scenario, thereby eliminating the cost benefits and general purpose nature of the IaaS cloud. Another very common use case is migrating existing applications to the cloud. Many applications are reliant on very specific network design patterns. These apps would pose problems if they were moved to the cloud and might even have to be rewritten to fully operate in a cloud environment

Enter overlay-based network virtualization

Overlay-based network virtualization is a technology that allows cloud users to provision virtual network devices such as virtual switches, virtual routers, virtual firewalls and virtual load balancers. These virtual network devices can then be connected to VM’s as well as other virtual network devices to create complex network topologies. Since these virtual devices live in software, the underlying network (a.k.a. the physical network) only needs to be an IP network which allows all the compute hosts to see each other. Two leading examples of overlay-based network virtualization solutions are Midokura’s MidoNet and Nicira’s Network Virtualization Platform. These particular solutions have an added benefit that they are designed to be fully distributed; that means the scaling model is linear and can be scaled out incrementally as demand increases. They are also integrated with cloud management solutions so that virtual network device provisioning is automated. Those who spent their lives deploying production clouds think of overlay-based network virtualization as the best way to handle networking for cloud environments.

Predictions and prognostications

Now it’s time for me to put on my Nostradamus hat. Server virtualization adoption has grown at an extremely fast pace since its debut and has fundamentally changed the IT landscape. The next phase is widespread self-service IT adoption, and consequently, the proliferation of IaaS clouds. These concepts, as well as the technology behind them, will become essential to how the modern enterprise will deliver IT services. Because overlay-based network virtualization solves the very real problems stated above, it will soon become the preferred method of handling cloud networking. Now is a great time to start researching overlay-based network virtualization to better understand how it will fit within your IT future.

Ben Cherian is a serial entrepreneur who loves playing in the intersection of business and technology. He’s currently the Chief Strategy Officer at Midokura, a network virtualization company. Prior to Midokura, he was the GM of Emerging Technologies at DreamHost, where he ran the cloud business unit. Prior to that, Ben ran a cloud-focused managed services company.

Must-Reads from other Websites

Panos Mourdoukoutas

Why Apple Should Buy China’s Xiaomi

Paul Graham

What I Didn’t Say

Benjamin Bratton

We Need to Talk About TED

Mat Honan

I, Glasshole: My Year With Google Glass

Chris Ware

All Together Now

Corey S. Powell and Laurie Gwen Shapiro

The Sculpture on the Moon

About Voices

Along with original content and posts from across the Dow Jones network, this section of AllThingsD includes Must-Reads From Other Websites — pieces we’ve read, discussions we’ve followed, stuff we like. Six posts from external sites are included here each weekday, but we only run the headlines. We link to the original sites for the rest. These posts are explicitly labeled, so it’s clear that the content comes from other websites, and for clarity’s sake, all outside posts run against a pink background.

We also solicit original full-length posts and accept some unsolicited submissions.

Read more »