Welcome!

Big Data Journal Authors: Elizabeth White, Liz McMillan, Roger Strukhoff, Adrian Bridgwater, Pat Romanski

Related Topics: SDN Journal, Java, .NET, Virtualization, Cloud Expo, Big Data Journal

SDN Journal: Blog Feed Post

Virtual Apostasy

When all you have is a hypervisor, everything looks like it should be virtualized

When all you have is a hypervisor, everything looks like it should be virtualized.

Yes, I'm about to say something that's on the order of heresy in the church of virtualization. But it has to be said and I'm willing to say it because, well, as General Patton said, "If everyone is thinking the same...   someone isn't thinking."

The original NFV white paper cited in the excellent overview of the SDN and NFV relationships "NFV and SDN: What’s the Difference?" describes essentially two problems it attempts to solve: rapid provisioning and operational costs.

The reason commodity hardware is always associated with NFV and with SDN is that, even if there existed a rainbow and unicorns industry-wide standard for managing network hardware there would still exist significant time required to acquire and deploy said hardware. One does not generally have extra firewalls, routers, switches, and application network service hardware lying around idle. One might, however, have commodity (cheap) compute available on which such services could be deployed.

Software, as we've seen, has readily adapted to distribution and deployment in a digital form factor. It wasn't always so after all. We started with floppies, moved to CD-ROM, then DVD and, finally, to neat little packages served up by application stores and centralized repositories (RPM, NPM, etc...).

Virtualization arrived just as we were moving from the physical to digital methods of distribution and it afforded us the commonality (abstraction) necessary to enable using commodity hardware for systems that might not otherwise be deployable on that hardware due to a lack of support by the operating system or the application itself. With the exposure of APIs and management via centralized platforms, the issue of provisioning speed was quickly addressed. Thus, virtualization is the easy answer to data center problems up and down the network stack.

But it isn't the only answer, and as SDN has shown there are other models that provide the same agility and cost benefits as virtualization without the potential downsides (performance being the most obvious with respect to the network).

ABSTRACT the ABSTRACTION

Let's abstract the abstraction for a moment. What is it virtualization offers that a similar, software-defined solution would not? If you're going to use raw compute, what is it that virtualization provides that makes it so appealing?

Hardware agnosticism comes to mind as a significant characteristic that leads everyone to choose virtualization as nearly a deus-ex machina solution. The idea that one can start with bare metal (raw compute) and within minutes have any of a number of very different systems up and running is compelling. Because there are hardware-specific drivers and configuration required at the OS level, however, that vision isn't easily realized. Enter virtualization, which provides a consistent, targetable layer for the operating system and applications.

Sure, it's software, but is standardizing on a hypervisor platform all that different from standardizing on a hardware platform?

We've turned the hypervisor into our common platform. It is what we target, what we've used as the "base" for deployment. It has eliminated the need to be concerned about five or ten hundred different potential board-level components requiring support and provided us a simple base platform upon which to deploy. But it hasn't eliminated dependencies; you can't deploy a VM packaged for VMware on a KVM system or vice-versa. There's still some virtual diaspora in the market that requires different targeted packages. But at least we're down to half-a-dozen from the hundreds of possible combinations at the hardware level.

But is it really virtualization that enables this magical deployment paradigm or is it the ability to deploy on common hardware it offers that's important? I'd say its the latter. It's the ability to deploy on commodity hardware that makes virtualization appealing. The hardware, however, still must exist. It must be racked and ready, available for that deployment. In terms of compute, we still have traditional roadblocks around ensuring compute capacity availability. The value up the operational process stack, as it were, of virtualization suddenly becomes more about readiness; about the ability to rapidly provision X or Y or Z because it's pre-packaged for the virtualization platform. In other words, it's the readiness factor that's key to rapid deployment. If there is sufficient compute (hardware) available and if the application/service/whatever is pre-packaged for the target virtualization platform then rapid deployment ensues.

Otherwise, you're sitting the same place you were before virtualization.

So there's significant planning that goes into being able to take advantage of virtualization's commoditization of compute to enable rapid deployment. And if we abstract what it is that enables virtualization to be the goodness that it is we find that it's about pre-packaging and a very finite targeted platform upon which services and applications can be deployed.

The question is, is that the only way to enable that capability?

Obviously I don't think so or I wouldn't be writing this post.

COMPLACENCY is the GREAT INHIBITOR of INNOVATION

What if we could remove the layer of virtualization, replacing it instead with a more robust and agile operating system capable of managing a bare metal deployment with the same (or even more) alacrity than a comparable virtualized system?

It seems that eliminating yet another layer of abstraction between the network function and, well, the network would be a good thing. Network functions at layer 2-3 are I/O bound; they're heavily reliant on fast input and output and that includes traversing the hardware up through the OS up through the hypervisor up through the... The more paths (and thus internal bus and lane traversals) a packet must travel in the system the higher the latency. Eliminating as many of these paths as possible is one of the keys*** to continued performance improvements on commodity hardware such that they are nearing those of network hardware.

If one had such a system that met the requirements - pre-packaged, rapid provisioning, able to run on commodity hardware - would you really need the virtual layer?

No.

But when all you have is a hypervisor...

I'm not saying virtualization isn't good technology, or that it doesn't make sense, or that it shouldn't be used. What I am saying is that perhaps we've become too quick to reach for the hammer when confronted with the challenge of rapid provisioning or flexibility. Let's not get complacent. We're far too early in the SDN and NFV game for that.

* Notice I did not say Sisyphean. It's doable, so it's on the order of Herculean. Unfortunately that also implies it's a long, arduous journey.

** That may be a tad hyperbolic, admittedly.

*** The operating system has a lot - a lot - to do with this equation, but that's a treatise for another day

Read the original blog entry...

More Stories By Lori MacVittie

Lori MacVittie is responsible for education and evangelism of application services available across F5’s entire product suite. Her role includes authorship of technical materials and participation in a number of community-based forums and industry standards organizations, among other efforts. MacVittie has extensive programming experience as an application architect, as well as network and systems development and administration expertise. Prior to joining F5, MacVittie was an award-winning Senior Technology Editor at Network Computing Magazine, where she conducted product research and evaluation focused on integration with application and network architectures, and authored articles on a variety of topics aimed at IT professionals. Her most recent area of focus included SOA-related products and architectures. She holds a B.S. in Information and Computing Science from the University of Wisconsin at Green Bay, and an M.S. in Computer Science from Nova Southeastern University.

Latest Stories from Big Data Journal
When one expects instantaneous response from video generated on the internet, lots of invisible problems have to be overcome. In his session at 6th Big Data Expo®, Tom Paquin, EVP and Chief Technology Officer at OnLive, to discuss how to overcome these problems. A Silicon Valley veteran, Tom Paquin provides vision, expertise and leadership to the technology research and development effort at OnLive as EVP and Chief Technology Officer. With more than 20 years of management experience at lead...
IoT is still a vague buzzword for many people. In his session at Internet of @ThingsExpo, Mike Kavis, Vice President & Principal Cloud Architect at Cloud Technology Partners, will discuss the business value of IoT that goes far beyond the general public's perception that IoT is all about wearables and home consumer services. The presentation will also discuss how IoT is perceived by investors and how venture capitalist access this space. Other topics to discuss are barriers to success, what is n...
Cisco on Wedesday announced its intent to acquire privately held Metacloud. Based in Pasadena, Calif., Metacloud deploys and operates private clouds for global organizations with a unique OpenStack-as-a-Service model that delivers and remotely operates production-ready private clouds in a customer's data center. Metacloud's OpenStack-based cloud platform will accelerate Cisco's strategy to build the world's largest global Intercloud, a network of clouds, together with key partners to address cu...
BlueData aims to “democratize Big Data” with its launch of EPIC Enterprise, which it calls “the industry’s first Big Data software to enable enterprises to create a self-service cloud experience on premise.” This self-service private cloud allows enterprises to create 100-node Hadoop and Spark clusters in less than 10 minutes. The company is also offering a Community Edition via free download. We had a few questions for BlueData CEO Kumar Sreekanti about all this, and here's what he had to s...
Labor market analytics firm Wanted Analytics recently assessed the market for technology professionals and found that demand for people with proficient levels of Hadoop expertise had skyrocketed by around 33% since last year – it is true, Hadoop is hard technology to master and the labor market is not exactly flooded with an over-abundance of skilled practitioners. Hadoop has been called a foundational technology, rather than ‘just’ a database by some commentators – this almost pushes it towards...
The cloud provides an easy onramp to building and deploying Big Data solutions. Transitioning from initial deployment to large-scale, highly performant operations may not be as easy. In his session at 15th Cloud Expo, Harold Hannon, Sr. Software Architect at SoftLayer, will discuss the benefits, weaknesses, and performance characteristics of public and bare metal cloud deployments that can help you make the right decisions.
Technology is enabling a new approach to collecting and using data. This approach, commonly referred to as the “Internet of Things” (IoT), enables businesses to use real-time data from all sorts of things including machines, devices and sensors to make better decisions, improve customer service, and lower the risk in the creation of new revenue opportunities. In his session at Internet of @ThingsExpo, Dave Wagstaff, Vice President and Chief Architect at BSQUARE Corporation, will discuss the real...
Where historically app development would require developers to manage device functionality, application environment and application logic, today new platforms are emerging that are IoT focused and arm developers with cloud based connectivity and communications, development, monitoring, management and analytics tools. In her session at Internet of @ThingsExpo, Seema Jethani, Director of Product Management at Basho Technologies, will explore how to rapidly prototype using IoT cloud platforms and c...
Amazon, Google and Facebook are household names in part because of their mastery of Big Data. But what about organizations without billions of dollars to spend on Big Data tools - how can they extract value from their data? Ion Stoica is co-founder and CEO of Databricks, a company working to revolutionize Big Data analysis through the Apache Spark platform. He also serves as a professor of computer science at the University of California, Berkeley. Ion previously co-founded Conviva to commercial...
Due of the rise of Hadoop, many enterprises are now deploying their first small clusters of 10 to 20 servers. At this small scale, the complexity of operating the cluster looks and feels like general data center servers. It is not until the clusters scale, as they inevitably do, when the pain caused by the exponential complexity becomes apparent. We've seen this problem occur time and time again.