Every Cloud Has a Silver Lining

December 1, 2010

Last week, Netflix made headlines when the company announced that it was moving most of its operations to Amazon Web Services' Elastic Compute Cloud. The news was greeted with enthusiasm and was seen as further justification of the public cloud. Rightly so: the fact that Netflix generates up to 20% of US traffic during peak times, and that this traffic is moving to the public cloud would seem justification to me. This is a great piece of advertising for the cloud (much better than Microsoft's "to the cloud" campaign), and by proxy a great piece of advertising for SoftLayer.

So why did Netflix make the move? Economics - plain and simple. It is less expensive to move to the cloud than it is to continue supporting everything via internal Netflix DCs. In a cloud model, peak traffic loads dictate Netflix's economics - they pay for peaks, but only when they occur. When traffic drops off, Netflix enjoys the resultant cost savings, plus they relieve themselves of a considerable management burden. The argument is straightforward.

All of this has brought me back to consideration of the "private" cloud (which is arguably not the cloud at all, but I digress) and the value that it offers. The industry definition of a private cloud is a cloud implementation that is internal to (still owned and operated by) a single enterprise. SoftLayer defines a private cloud a little differently: SoftLayer remains the IAAS provider, but we ensure that a customer is on a single node (i.e. server). This conversation will stick with the industry definition.

So, what are the impacts of the private cloud across an enterprise?

In theory, a private cloud would give individual departments or discrete project teams the ability to better manage cost. As with a public cloud, a project team would be able to take advantage of the cost savings that come with paying only for what they use. However, this means that change is necessary in corporate accounting functions given that systems now need to be managed based on a "pay as you go model" versus a cost center model. This means a fundamental change in IT philosophy, as they now need to bill departments working on a variable use model - all of a sudden they have to think more like a business unit with a P&L to manage.

The cloud provides the ability to quickly spin up and scale. In the SoftLayer world, this translates to availability in anywhere from 2-4 hours. This should mean an increase in operational efficiency across departments using the cloud. Projects can start and end quickly without concern for lengthy implementation or teardown windows. That said I am not sure this increase in efficiency is meaningful when balanced against an IT department that must build and support a cloud infrastructure that has to account for the operation across the entire enterprise. The impact is potentially great at a micro level, but wanes when you consider the larger picture.

From a planning point of view, IT must now consider what a variable use model means in practical terms. Different departments will experience different peaks and valleys based on different workloads. In all likelihood, these peaks will not align on the calendar, nor will they be consistent month over month. In addition, my assumption is that deployment of a cloud will engender unanticipated usage patterns given the supposed cost and operational flexibility that the cloud delivers within the enterprise. The challenge will be to balance these needs against the delivery of a service that will adhere to QoS promises and associated internal service level agreements. ( And I think they will have to exist. If IT moves beyond a cost center, and internal organizations are trying to budget based upon forecasted compute use, it only makes sense that IT will be held up against external providers. Indeed, I would expect some rogue departments to go off the reservation to external providers based on cost considerations alone.)

My guess is that the IT response to planning will be predictable - over-engineer the private cloud to make sure that it is bullet proof. This might work, but it will be expensive and will paradoxically lead to underutilization of what is an over planned resource - something the cloud is supposed to mitigate. This approach is also likely to lead to IT bloat as capable internal resources are likely to be thin, driving a round of hiring to ensure expertise is on hand to manage the cloud.

In addition, I would assume that some applications will continue to be supported (for a variety of reasons - security, I/O challenges), thus adding more cost to the equation.

The arguments against the private cloud are numerous and ought to give the enterprise pause for thought. Regardless, I am willing to bet that private cloud implementations will accelerate in the enterprise. Many companies are supported by IT organizations that are strong and well entrenched within the corporate culture. Part of the fight will be based around losing budget, headcount and perceived power if the decision is made to go to an external IAAS provider. All of the usual rubrics will be used: security, quality of service, performance and on the list goes. In essence, these are the same arguments that have been made in the past when a decision to outsource anything has come to the fore. Does it make sense to outsource everything? The answer is a resounding no, but the argument for looking to the public cloud, or SoftLayer's private cloud is strong.

-@gkdog

Leave a Reply

Filtered HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • You can enable syntax highlighting of source code with the following tags: <pre>, <blockcode>, <bash>, <c>, <cpp>, <drupal5>, <drupal6>, <java>, <javascript>, <php>, <python>, <ruby>. The supported tag styles are: <foo>, [foo].
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.

Leave a Reply

Filtered HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • You can enable syntax highlighting of source code with the following tags: <pre>, <blockcode>, <bash>, <c>, <cpp>, <drupal5>, <drupal6>, <java>, <javascript>, <php>, <python>, <ruby>. The supported tag styles are: <foo>, [foo].
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.