Insider's Warehouse

Clouds for rent

Cloud computing offers options for companies without a large data center.

Cloud computing is an emerging disruptive technology that is poised to change the structure of all data centers. In the past, IT only considered options to build or buy technology. Ironically, with cloud computing, components that have been around for many years are being brought together, and IT buyers are given a third option—renting these products.

At least 22 well-published definitions of cloud computing exist—a clue that the market is still developing. The National Institute of Standards and Technology Working Definition of Cloud Computing offers the best explanation of the technology’s key attributes. Clouds must provide:

figure 1

Click to enlarge

  • On-demand self-service. Dynamic provisioning
  • Broad network access. Internet and intranet
  • Resource pooling. Multi-tenant and virtualization of resources
  • Rapid elasticity. Quick scale-out and scale-in
  • Measured service. Chargeback and monitoring

Formations

Different kinds of cloud computing service models are available (see figure 1):

  • Software as a service (SaaS) provides applications running in the vendor’s data center, usually charging per-seat usage licenses. However, some SaaS is free. SaaS appeals to small and mid-size businesses (SMBs) that need enterprise-class applications but without the large data center. Additionally, large companies are exploring SaaS at the departmental level. Popular suppliers are Salesforce.com, NetSuite and WebEx. Well-known free-per-seat vendors include Google Mail, Zoho and AOL.
  • Platform as a service (PaaS) provides middleware, databases, Web application servers, Web services and tools. For example, eBay and PayPal offer Web services in the cloud for a monthly fee based on actual use. Equally important are the tools and middleware that are part of an application, such as Web application servers, portals, etc.
  • Infrastructure as a service (IaaS) provides hardware on demand, including servers, storage and networks. Public, private and hybrid clouds are the deployment models in this category.
Insiders_figure_tn

Click to enlarge

Public clouds

Public cloud vendors evolved from Web hosting in which servers and storage are available for rent outside the corporate firewall. Most cloud hardware consists of huge grids of commodity servers and storage area networks. Public cloud vendors bill the user for each hour of CPU time consumed, with fees per hour ranging from 10 to 80 cents, depending on the server’s CPU speed and memory size. Like your electric bill, you pay only for what you use. The ability to use servers on demand adds new dimensions of agility to technology infrastructures.

Disk storage is billed per gigabyte per month. At 15 cents per gigabyte per month, a terabyte of disk storage per month would be $150. Most vendors also charge for uploading and downloading data over the network. This may sound expensive over the long term until you consider that the public cloud vendor manages the hardware for you, pays the electricity bills, manages floor space, and provides cooling. Many CIOs are discovering that the cost of cooling and electricity per year is much higher than the cost of hardware.

The underlying technology of cloud computing has been maturing for years and now offers tangible benefits for cost reduction and corporate agility.

There are many other benefits to be realized. Corporate agility is derived from being able to quickly build up and shut down systems on demand in response to immediate business needs. CFOs often point to the elimination of large, up-front capital expense outlays. Many organizations also cite multiple reductions in:

  • Hardware requirements
  • Costs from the pay-per-usage model
  • Ratio of operations staff per blade server rack

However, public clouds also have drawbacks. Many companies worry that putting their data outside the firewall is a huge risk. Indeed, auditors often won’t allow it. Some public cloud vendors are high-risk, while others realize that their business survival depends on having good security.

Service levels among vendors also vary. Shared servers can slow down if another enterprise tenant launches several huge batch jobs on the server you are sharing. To get consistent performance, “reserved” hardware is available for additional fees. Still, because performance is not like a purpose-built configuration, public clouds are not used for mission-critical data warehousing. For example, availability (up time) isn’t guaranteed, even though fees are refunded if the server or storage is offline. Translation: It’s OK to put certain data marts in the cloud but not the enterprise data warehouse (EDW).

Private clouds

Private clouds provide the same benefits as public clouds but without many of the risks because the hardware and software are dedicated to one corporation and typically run on a grid inside the organization’s data center. (See figure 2.) Because these resources are maintained by the corporate IT staff, security of confidential data can be guaranteed, high-performance configurations can be built, and system availability can be carefully controlled to accommodate mission-critical data marts and data warehouses.

Private clouds operate inside the firewall and are built upon virtualization software such as VMware or Xen. These products let the DBA take a snapshot of an application and all of its components and put it in an image library. The image can then be loaded in minutes onto any server that is not too busy. To do this, the data must be in a storage area network or network-attached storage.

The biggest benefit of private clouds is server consolidation. Because most servers have low utilization, 50 of them can be collapsed into five large ones, each one containing 10 virtual images. That means 45 servers’ worth of floor space, electricity and cooling costs can be retired or used for other purposes. In the end, private clouds using virtualization benefit corporations by adding a dimension of on-demand provisioning.

Of course, there are a few downsides to private clouds when compared with public clouds. The CIO must spend capital expense dollars to buy the hardware up front; in addition, electricity, cooling and operations staff labor will incur ongoing costs.

Teradata in the clouds

Teradata Labs recently announced the Teradata Enterprise Analytic Cloud for both public and private clouds.

Private clouds

Teradata Labs is working closely with top architects at VMware Corp. to optimize both companies’ products for running analytic workloads in a private cloud. The first delivery, announced during the 2009 Teradata PARTNERS User Group Conference & Expo, is Teradata Express for VMware Player. This free test-and-development environment, which includes snapshot images of the Teradata Database bundled with Novell SLES 10, is ready to deploy using VMware Player. Teradata Express for VMware Player greatly simplifies development activities, especially when testing new patches, business intelligence (BI) tools, or extract, transform and load (ETL) compatibility. Using 64-bit x86 servers, the environment can support up to a terabyte of data—more than enough capability for a majority of developer tasks.

A second option is the Teradata Agile Analytics Cloud. Using a new Teradata Viewpoint self-service portlet, the business user can create elastic data marts, upload comma-separated values files and begin analysis in less than 15 minutes. Once tables are uploaded, the user can join it to the data warehouse to produce new reports or perform analyses. By leveraging the data warehouse, no additional ETL scripts are needed and data is not duplicated. Without any IT involvement, the user can test hypotheses, conduct research, develop requirements and discover new insights. In some cases, the elastic mart is promoted by integrating it into the data warehouse. Another benefit is the avoidance of data mart proliferation outside the IT organization. The key to all of this is that it is fast and free: There are no procurement cycles, justifications, planning or hardware purchases.

Public clouds

With top architects at Amazon Web Services, Teradata Labs is working to optimize public clouds for analytic workloads. As a first step, Teradata Express for Amazon EC2 is available for free to allow for test and development with up to a terabyte of data. Bundled with Novell SUSE Linux, this solution is good for data marts, proof-of-concept tests, analytic sandboxes, urgent projects and application development. The low cost of servers and storage from Amazon Web Services lets users expand capacity in minutes without having to buy and manage hardware, opening up myriad new ways to solve BI needs.

—D.G.

Clear vision

The underlying technology of cloud computing has been maturing for years and now offers tangible benefits for cost reduction and corporate agility. Many IT organizations are testing the various cloud formations, thinking through how best to take advantage of them. Even though there are some pitfalls, multiple benefits can be harvested. Renting data warehousing components not only offers cost savings that are attractive for many workloads but also provides the agility to quickly respond to business changes.

Visionaries are already using cloud computing while mainstream IT shops have heard the thunder loud and clear. It’s time to investigate analytics in the clouds.


Your Comment:
  
Your Rating:

Comments
Fuzzy Logix