Archive

Posts Tagged ‘Amazon’

The Last Big Thing – Cloud Computing

There is always a “Big Thing” in computing that captures the attention of the industry press, consultants, and conference sponsors.

It was recently Cloud Computing, and the term was used to describe a couple of different but complementary scenarios.   In the first scenario, cloud computing was used to describe applications like Salesforce.com, and NetSuite that were running in remote data centers managed by those companies with access to the applications done via a web browser.   These companies were able to make a compelling argument that it easier and less costly for the enterprise to have applications like customer relationship management and enterprise resource planning hosted in a data center run and maintained by a third party.   For the most part they’ve been right.

In the second scenario, cloud computing was used to describe the virtualization of the data center wherein applications and databases can be moved between servers and scaled up or down (using more or less of a server’s resources) according to the demand of the organization so that overall computing resources can be shared and better utilized.  How much better?   In a presentation by George Slessman the CIO at IO Data Centers at the 2012 Uptime Symposium he makes the assertion that roughly 65% of data center resources are not being utilized at any particular time.   In other words, you are paying for capacity in your data centers that is sitting idle most of the time.    There is a strong argument here for either using outsourced data center services on an as needed basis, or for building your own virtual data centers that expand and contract based on computing needs.

One of the characteristics of analytics – the analysis of your corporate data for insights and for modeling current and future scenarios – is that we can’t always predict just how much computing resource will be required to do the job.   This fluctuating demand for computing resources means some weeks we may need all the computing resources the data center can muster to crunch the numbers, while at other times computing resources will sit idle while we pour over the results.

It was natural for us to consider Cloud Computing – or virtual data center services – as a way to accommodate the fluctuating demand for computing.  If this is done within the enterprise, it means better data center utilization as different parts of the organization share resources (remember timesharing?).

Some entrepreneurial organizations also realized they might avoid capital expenditures for computing resources by using Cloud Computing vendors like Amazon to host their analytics needs.  In other words, why not let Amazon deal with managing timesharing?

Unfortunately for the large enterprise, it isn’t always so easy to just ship your data off to Amazon – or one of the many cloud analytics providers that host at services like Amazon – so you can run your analysis at will without having to worry about capacity planning.  There is plenty of discussion on the web about the potential security issues associated with sending data off site and potential conflicts with licensing your favorite analysis tool.  Most of the cloud service providers have become very adept at managing security and licensing.

In my experience, the biggest limitation to using cloud services for analytics is the difficulty in physically transferring large volumes of data from your location to the cloud provider’s servers.  There are three aspects to this.   First, you will have to convince your IT staff to perform the extract job to acquire and prepare the data according to the cloud vendor’s specifications.   Second, they will have to compress the data in preparation for transfer with an agreement on the service provider’s side regarding what they will use to decompress the data.   Third, your IT team will be held responsible for scheduling, tracking, and assuring that the work on their end gets done in the appropriate time window, AND tracking the service providers performance because at the end of the day your analysts and business users are likely to expect them to explain and fix ANY problem that keeps them from their analysis, even if it is beyond their control.

The bottom line is that pushing your analytics into a cloud service may or may not be practical and IT will still need to be involved.  There will be enhancements to the transfer and security management of data that will make outsourcing analytics easier. One factor to consider that may offset  any of these new enhancements is the explosion in the volume and velocity of data that analysts now want to include in their analysis and modeling.   This brings us to the current big thing – Big Data.

Categories: Uncategorized Tags: ,