Managing the Dynamic Datacenter

Datacenter Automation

Subscribe to Datacenter Automation: eMailAlertsEmail Alerts newslettersWeekly Newsletters
Get Datacenter Automation: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn

Datacenter Automation Authors: Yeshim Deniz, Elizabeth White, Pat Romanski, Liz McMillan, Glenn Rossman

Related Topics: Cloud Computing, Virtualization Magazine, Datacenter Automation, Java in the Cloud, Big Data on Ulitzer


Big Data in the Non-Virtualized Cloud: Myth or Reality?

Cloud is typically approached as a combination of virtualized or bare metal infrastructure

By Jake Gardner - Cloud is typically approached as a combination of virtualized or bare metal infrastructure. At this stage of the game, to think about cloud is to think at some level about virtualization. Businesses are also looking at some form of orchestration. And the underlying hardware – CPU, RAM, and disk – is becoming commodity for the cloud. Nobody cares what the underlying processor on a server is that’s running in the cloud. They just care about the things that come after: performance, chiefly.

We assume in the cloud that there some basic level of automation occurring. Whether that’s template deployment, or utilizing some form of configuration management tool like Puppet or Chef, or a DCIM model (using OpsWare or BMC).

This is what we call the cloud today.

Big data performance

How do you max out performance goals for big data? Bare metal might be the key...

But there is a niche that can exist for companies seeking a different kind of cloud performance by putting layered orchestration over just bare metal. For the intersection of what we call High Performance Computing and cloud, the bare metal cloud has the potential to become a viable pathway either as a hosted or “run it in your data center” model.

Companies utilizing big data, for example, can put pieces of hardware behind a layer of orchestration that are capable, through software defined switching, of creating “virtual servers” on bare metal servers without being technically virtualized.

The benefit is higher performance. With your average hypervisor, there is typically a loss of some level of overhead (whether that’s a little memory, CPU or throughput on the network) – its not a zero footprint.

While it’s true that not all applications will benefit from this deployment architecture, but many forms of high performance computing (numerical number crunching or Hadoop) applications can benefit from the benefits of a cloudy approach. It’s nice to be able to say “spin up 10 Hadoop nodes” and have that happen quickly. You can then do your computing on them and then turn them off.

Issues arise when attempting to achieve this level of speed and resource building is attempted in a virtualized environment; your performance overhead on an application like Hadoop has a negative impact to its overall performance. This issue doesn’t exist with bare metal. Often companies take large computation projects like that and put them in non-virtualized environments.

The tradeoff, however, is that it’s not cloud, and achieves none of the cost and flexibility benefits that cloud can deliver. But if you utilize a bare metal traditional hosted server model and apply the cloud orchestration on top of it, then you can gain both the benefits of performance from the non-virtualized machines, as well as the cloudy qualities that enable a greater degree of automation and speed to the process.

More Stories By Gathering Clouds

Cloud computing news, information, and insights. Powered by Logicworks.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.