Information Management: Strategies for Gaining a Competitive Advantage with Data (2014)

Chapter 13. Cloud Computing

On-Demand Elasticity

Nothing will change the way IT operates like cloud computing. The cloud can be any combination of deployment model, cloud service, and application domain. No two companies will approach it the same way, but it’s important to take heads-up activity toward an enterprise cloud strategy for information.

Keywords

cloud computing; virtual data center; public cloud; private cloud; virtualization; internet-based services; utility computing; business intelligence; information technology

Everyone is talking about the cloud—and for good reason. Nothing will change the way IT operates like cloud computing. Those of us who have been deploying applications for a long time are used to being able to, at some point, perform maintenance and upgrades, or perhaps simply view, our physical servers. Those days are passing us by.

Server management is one of those functions that is not considered part of the core competencies in developing applications. It is increasingly being delivered by service level to applications, including information management applications that need servers to function, which is most of them.

Service levels will include things like up time, response time, data size, tiered performance levels, level of willing risk (vs. cost), resource management needs, etc. Though these may seem to constitute a quite reasonable range of the needs of an information management application, if there is any temptation to turn over platform selection to the cloud provider, let me dissuade you from that type of thinking. The information management owner needs to make those decisions, the decisions that this book is geared toward supporting.

The Changing Nature of IT

IT was once determined to be best fit in an organization as a central organization, providing all of the technical services that the organization requires. As of this writing, it still is in many organizations. However, many other organizations are taking a more balanced approach with IT. There is still a central IT, but there is latitude for distributed IT organizations (most likely not in name, but in function). Wherever the technical roles reside is what I mean when I reference “IT.”

Defining Cloud Computing

So what is “the cloud” and how does it fit into this discussion of turning server management over to service levels? The cloud is perhaps the most nebulous (or nebula ) term in technology. At one end of the spectrum (spectra ), the cloud is erroneously used as a term to simply mean the Systems Administration group no longer wants to allow you to see your server due to difficulties with security access, the questions that the visits raise, or as a means of emphasizing the split in responsibilities between Systems Administration and the application teams. This can be referred to as a “private cloud.” However, I will give some criteria later that I think a true private cloud adheres to—some good functionality that provides true value to an organization.

At the other end of the spectrum, you have established public clouds like Amazon Web Services, Softlayer (an IBM Company), Rackspace, GoGrid, etc. that provide exhaustive services that go beyond the basic cloud criteria I will give you later.

The best definition for cloud computing I’ve found is the one by the U.S. National Institute for Standards and Technology (NIST).

It defines cloud computing as “a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.”¹ NIST defines the 5 characteristics of cloud computing as:

• On-Demand and Self-Service

• Broad Network Access

• Resource Pooling

• Rapid Elasticity

• Measured Service

On-demand and self-service disintermediate IT and others from an application owner and her ability to provision servers. Broad Network Access means access to a wide variety of modern platforms—much more than one could get on its own. Resource Pooling is gaining the benefits of economies of scale. Rapid Elasticity is the “accordion effect” of resources growing and shrinking seamlessly as needs dictate. Measured Services refers to the reporting infrastructure of the cloud provider, providing detailed reports to you on usage.

We are all consumers of cloud services, whether we refer to it as such or not. We might use a version of Microsoft Office in the cloud (starting with Version 365), DropBox or Google Docs for file sharing, iCloud from Apple, etc. Public email services are cloud based. Many organizations use one of the first pieces of software ready for the enterprise from the cloud: salesforce.com (whose phone number is 1-800-NO-SOFTWARE). Salesforce.com software is for contact management and many see this as an enterprise function that can be easily isolated. However, the idea of offloading not only the servers referenced above, but also the software (installation, maintenance, upgrades, etc.), is catching on in many enterprise domains, including all of the ones related to information management—data storage, data integration, data access, etc.

The cloud is a metaphor for the internet and telephone networks.

You are completely hamstrung these days selecting information management software and hardware if you patently rule out the cloud option. In many areas, the cloud options are the rule, not the exception. You may end up with pockets of software, hardware, and data in many clouds—private and public—and your new dominant role in IT is to choose and integrate wisely.²

Information management is a frequent leading domain in bringing the cloud strategy to an organization or taking the strategy to new levels. The cloud is essential for the information manager to understand, which I why I must talk about it in this book on information management.

Moving into the Cloud³

The most elaborate public cloud that I’ve visited is the Switch SuperNAP in Las Vegas (www.switchlv.com). This is where companies like eBay, Google, DreamWorks, Sony, Cisco, Pixar, Joyent, HP, Vmware, Nirvanix, and many others call home for their server infrastructure.

The SuperNAP is the world’s most efficient, high-density data center in the world. It is (currently) 2,200,000 square feet of data center space for 520,000 square feet of operational data space with 500 megavolt amperes (MVA) of power capacity, 567 MVA of generator capacity, and 294 MVA of Uninterruptible Power Supply.

The highly redundant systems and service levels for on-call power delivery would seem to make it the last building to lose any power if the worst were to happen. Electrical power inside Switch facilities is delivered on System+System design so each primary and failover system is completely separated from the other. The fault-tolerant power design enables Switch to deliver 100 percent power uptime to its customers. As needed, the SuperNAP can be power self-sufficient with its inclusion of up to fifty 2.8 megawatt diesel generators onsite for a total of 140 megawatts of generator capacity.

The cooling system is a work of art itself. With 202,000 tons of cooling, it has 22,000,000 CFM (cubic feet per minute) of airflow. Depending on weather conditions, the system chooses which of the four ways it has at its disposal to cool. Each customer has a private secure caged area called t-scif™ that enables higher power and cooling densities per cabinet than any other data center in the world.

They also believe the location supports, rather than competes with the cooling system. As hot as Las Vegas is understood to be, 70% of the time the temperature is 72 degrees or less (i.e., nighttime). The temperatures are consistent. As well, Las Vegas is one of the few places without a real threat of natural disasters like floods, hurricanes, and earthquakes.

The cleanliness rivals the clean rooms where chips are made. Our guards said the physical aspects were the most important element to security. The armed escort guards at either end of our group certainly attested to that.

And in those over 31,000 cabinets reside the data that is supported by this infrastructure and made available in many forms, including collocation, connectivity, and cloud computing. This is data that many of us interact with daily. SuperNAP clients are built on data and have huge downsides if transactions are not collected, hospital records are compromised, or television content fails.

The SuperNAP is simply unattainable infrastructure for a company to build for themselves. Yet, it reflects the pseudo-requirements that I see all of the time. This is what it takes to provide true “99.9999%” availability. I was impressed with the decision making process across so many factors that the designers must have gone through.

Similarly, many companies are looking to build their own private clouds and evaluate public cloud providers and will need to determine their requirements. The SuperNAP sets the bar. While designs and public clouds state high availability, greenness, efficiency, security, performance (eBay’s numbers are outstanding), and elasticity, it’s the infrastructure that delivers.

FIGURE 13.1 The switch SuperNAP.