Learning Guide

How one growing firm uses Amazon's EC2

This is part one of a three-part series on working with Infrastructure as Service (IaaS) from Jo Maitland, Executive Editor of SearchCloudComputing.com.


When online investment community KaChing Group Inc. arrived on the scene about a year ago, the company found it difficult to plan business and IT growth based on unpredictable traffic patterns.

"We are not in the business of trying to figure out how to scale our architecture up and down depending on network traffic. That's what we have Xignite for," said Pascal-Louis Perez, the CIO of Redwood City, Calif.-based KaChing. Xignite Inc., an on-demand financial market data provider, hosts its application servers on Amazon's EC2 Infrastructure as a Service platform.

These servers in turn feed real-time financial data, such as stock quotes, Securities Exchange Commission (SEC) filings, currency rates and mutual fund performance, to customers such as KaChing, Forbes and Citigroup Inc.

KaChing's computation engine, which also resides on Amazon's cloud, takes this real-time data to update the fantasy stock portfolios of its 400,000 registered users. Over time the company, which is a registered SEC investment adviser, will become a place where users follow site member trading experts to make real hedge fund investments.

Pay-as-you-go pricing works
Unlike the monthly or even hourly contracts companies can sign up for with many on-demand service providers, Xignite customers typically sign an annual contract and pay a monthly fee for server capacity and network bandwidth. But given that customers such as KaChing often reside in the same cloud on Amazon's ecosystem, network bandwidth costs are minimal, said Leo Chan, the CTO of Xignite in San Mateo, Calif.

"We don't like charging for [server capacity] overages. So when we see that happening, we sit down with the customer to elevate their contract to the next [capacity] tier," said Chan, who would not disclose the company's pricing.

On a small scale, charging per instance makes sense, in that you pay for the minimum capacity you need as you go. Experts argue that pricing models based on usage do not scale well for large businesses with heavy transaction volumes. But if you are a startup more concerned with gradual and affordable growth, such on-demand pricing models make sense.

"The pay-as-you-go [pricing] model allowed us to grow little by little and use incremental [Xignite] services without up-front infrastructure investments," KaChing's Perez said. "When we built the prototype, [our servers] were only doing a hundred calls a day to Xignite's service, and we had 1,000 users. We're now at 400,000 users, so you can imagine the amount of scaling we would have had to do for that."

 

Public vs. private cloud infrastructure:
Infrastructure as a Service (IaaS) comes in two forms:
  1. Pay-per-use hosting of virtual servers at an external, or public, cloud service provider.

  2. By operating an internal, or private, cloud, where your IT department offers virtual servers as a service.

Cambridge, Mass.-based Forrester Research Inc. found the following indicators of likely buyer interest in, and adoption of, these two forms of IaaS:

  • About 25% of all enterprises plan to adopt IaaS via an external service provider.

  • Firms are slightly less interested in internal clouds than they are in external IaaS.

  • Large-business respondents report more awareness, interest and adoption of external IaaS than do small businesses. The figures are the same for internal clouds.

These results are contrary to conventional wisdom regarding the initial demand for cloud services. Enterprises are leading the adoption, not small and medium sized businesses (SMBs). Moreover, they have different technology preferences and comfort levels with virtualization. Forrester also believes that early adopters of IaaS service offerings are driven by the instant provisioning of servers and the pay-per-use pricing model.

Further, unlike developer buyers, enterprise IT operations buyers may want to integrate their on-premise infrastructure with anything they deploy to a service provider, either temporarily or permanently. VMware, Citrix and others now push the idea of "bursting" from a private cloud into a public cloud for additional capacity as needed, but this is still at the very early stages, and neither company can provide a customer that is bridging from a private to a public cloud yet.


TABLE OF CONTENTS

Infrastructure as a Service: How to maintain control
How one growing firm uses Amazon's EC2
Cloud computing concerns slowing widespread adoption
Understanding cloud computing pricing

ABOUT THE AUTHOR:
Jo Maitland is an executive editor in the Data Center and Virtualization media group at TechTarget. This article features additional reporting by Christina Torode, a senior news writer in the CIO media group and TechTarget.

Editors' note: This chapter on Infrastructure as a Service is the fourth part of an e-book on cloud computing that also includes chapters on CIO strategies for the cloud, development for the cloud and Software as a Service.

This was first published in January 2010