The second day of the Open Compute Project Summit V was graced by the presence of Peter Krey, President and Founder of Krey Associates, an independent consultant for about 11 years, with a previous 25-year experience in core infrastructure engineering and application infrastructure inside financial services.

As he disclosed, his clients consist of enterprise CIOs and CTOs plus their respective hands-on teams; also venture capital and private investors, serial entrepreneurs and start-ups.

He confessed to be on a constant hunt for emerging disruptors.

A self-professed fan of Open Compute, Krey stated that for him the concept of “open” absolutely rocks. He focuses on thought leaders and enabling disruptors and on start-up founding teams, basically the ones who create and make things.

What is cloud computing?

 

Responses vary.

“For external cloud there’s Amazon, Rackspace, Google, Azure and others. For internal cloud there’s 99 percent VMware. Back in 2011 you started hearing about OpenStack. From my experience, cloud is not traditional enterprise IT,” said Krey. “Whether internal or external, first time ever enterprise customers and developers have had choice, freedom and control. No IT people, no process, no bureaucracy, no business case, no governance, standards’ compliance, architecture reviews, 9-12 month delays, etc. Cloud is a dislocation of excess process and potential massive acceleration via 100 percent automation,” he explained.

With true 100 percent automated cloud, you get what you need in minutes, not months.

“What does that have to do with networking?” asked Krey. “Tons.”

Furthering his presentation, Krey attempted a comparison between the East Coast vs. West Coast, as far as the enterprise IT is concerned.

In Krey’s opinion, the East Coast approach is very manual and time consuming. Low technology, inconsistent, difficult to repeat in terms of speed and scale.

As Krey once put it, “Enterprise IT puts the NO in inNOvation.”

In contrast, the West Coast method is highly automated, repeatable and consistent, as well as highly receptive to adopting and advancing innovation. Lots of open source, standards based, vendor agnostic hardware, and tons of automation.

While on the East coast tens to hundreds of infrastructure devices are administered per admin, on the West Coast thousands to ten thousands are administered per admin.

“Again, what does that have to do with networking?” asked Krey. “Tons.”

Why cloud computing?

 

Because traditional IT is too darn slow.

Krey shared a couple of Enterprise IT CIOs’ quotes:

“It’s also difficult to be the long pole in the tent every time we want to go do something.” – Rob Carter, CIO at FedEx

“Now it takes days. It’s hard to underestimate the value of letting scientists work at their own pace …” – Michael Heim, CIO of Eli Lilly

“Our goal is to optimize the time from idea to solution … the idea is when the stopwatch starts. Really, it’s about the customer – when does their stopwatch start, not mine.” – Chris Perretta, CIO State Street

There’s been a long lasting impression that the cloud is not enterprise IT. Krey highlighted a couple of counter-arguments:

 .

1. Point. Click. Deploy. 100 percent Automated Everything!

• Delivered in Minutes … Not Months. Near Instant Time-to-Market

2. Developer & End-User Self-Service, API Direct Automation Access

• Web Portal Front-end to Underlying Automated Provisioning Processes
• Zero Management Engagement And/or Service Provider Interaction

3. Economics Driven By Multi-Tenant, Virtualized, Shared Infrastructure

• Virtualized Servers, Shared Storage & Networking and Other Core Infrastructure
• Managed High Utilization of Shared Resources

4. Developer As Customer & SysAdmin Direct Automation Access

• Command Line Interface (CLI) and ReST Multi-language Interface

5. Grid-Like, Scale-Out Infrastructure

• Enables Smoother Scalability Path
• H/W & S/W Vendor Independent, Volume Driven Economics
• Quasi Elasticity and Dynamic Scaling

 .

Krey’s presentation, “Curb your innovation II – OCP Open Source Networking Update” was continued by describing in detail the Cloud OSI-like layers:

 .

1. Infrastructure as a Service (IaaS)

• All Basic Computing Infrastructure Deployed & Delivered As 100% Fully Automated Services (e.g., AWS)
• Users & Developers Access Via Web-portal, CLI And/or REST Based Interfaces.
• Self-deploy and Run Arbitrary Software, Incl. OS & Applications Within VMs
• Infrastructure Platform Virtualization of Servers, OS As VM’s, Networking, and Storage
• Versus Purchasing, Building, and Operating Infrastructure, Utilized As Pay-as-Used Variable Service

2. Platform as a Service (PaaS)

• App Dev Platform to Create & Deploy New Applications (E.G., Tomcat / Spring, Google App Engine)
• Deploys Onto IaaS, Apps Created Using Prog. Languages, Dev Platform & Tools Supported by the PaaS
• Facilitates Deployment of Applications Without the Cost & Complexity of Buying and Managing Underlying Hardware and
• Software Layers

3. Software as a Service (SaaS)

• Web-Based Applications & Software Services (e.g., GMail. Google Apps, Salesforce.com, WebEx)
• Accessible From Various Client Devices Through a Thin Client Interface Such As a Web Browser
• Provided Over Internet, Eliminating the Need to Install & Run the
• Apps on the Customer’s Own Infrastructure, Simplifying
• Maintenance and Support

 .

Cloud computing is 100 percent about automation. If you don’t get the infrastructure right, everything else about it will be in jeopardy.

What is missing in enterprise cloud?

 .

“Networking is clearly missing and it has been missing for quite a long time. Our overall automation layer can only go as fast as the slowest part.

“Networking is very closed, very proprietary, very limited open developer API access. In a nutshell, the state of the art developers’ key is copy/paste…” Krey explained.

The left column presents the status of the industry in 2011, while the right column lists the solutions available right now. As Krey points out, “we have zero open source developers tools for networking and zero open APIs.”

“Another way to look at this is talk to COOs, CFOs and the people responsible for multimillion-dollar budgets in some big organizations; from a non-technical Opex and Capex perspective, there’s also a huge potential impact here,” said  Krey.

He added: “While it is difficult to get senior enterprise colleagues to disclose networking Opex and Capex percentages of their total cost structures, let’s make some conservative assumptions: it’s not hard to imagine that some of these enterprises consume 20-30 percent of their total Opex and Capex from networking. In many organizations is one of the top five line items.”

“It’s not just about the Opex and Capex save,” emphasized Krey. “Automation is also about orders of magnitude, reduction in time to deliver and orders of magnitude increase in consistency and quality of the configuration, quality of operations etc.”

Focusing his discourse on what the future might hold, Krey announced a couple of accomplishments:

• ONIE (Open Network Install Environment) is a common multi hardware switch provisioning software as a project. It only enables large scale automation and OS installation on tons of switchers.

Version 1.0 supports the legacy power pc switch control processors, and an upcoming ONIE version currently in development will also be supporting the x86 switch control processors.

 .

• Broadcom Open 1.0

.

• Mellanox MSX 1400-OCP

.

• Intel Open Switch Spec

For  Krey, the future holds: upcoming Open Switch Doc Spec reviews, additional Open Switch Spec contributions, Open Switch software collaborations and hopefully many more.