We are constantly being bombarded by the benefits of cloud computing.
New trends usually start out with a thought leadership hype cycle, even while most pragmatic technology platform users admit that cloud computing platforms need more maturity to be trusted.
So, I raised my eyebrows when I saw Glenn Mattson’s contrarian post on Seeking Alpha about why Rackspace’s business model is flawed.
I claim that this superficial analysis of the cloud computing space misses many essential points about the industry, and here are many details why.
The main assertion is that Rackspace is essentially doing a sale-lease-back of IT infrastructure so that companies can rent storage and compute resources rather than buy them, and that on its face this transfers unacceptable utilization risk and capex outlays to Rackspace.
This simple assertion ignores several structural cost advantages leveraged by well-scaled Infrastructure As A Service (IAAS) companies like Rackspace and Amazon’s AWS.
- Compute and Storage Costs
Large IAAS companies build their own servers from commodity parts, do not pay ongoing maintenance and support contracts for these resources, and have been demonstrated to achieve a 5x cost advantage over medium-scale enterprise costs for storage and compute.
- Power Costs
Large IAAS companies house their servers in locations that have cost-effective wholesale power generation options, such as hydroelectric power in Eastern Washington (such as Amazon Web Services, Google, and others) or inexpensive generation costs in North Carolina (such as Apple’s recent $750m iCloud datacenter). Power costs in these areas can range from 3 to 4.5 cents per kilowatt-hour (kWh), compared to 8-16 cents/kWh retail power costs in most of the United States. Overseas, the advantage is even more stark, with European power costs ranging as high as 30 cents/kWh, and power in Singapore north of 22 cents/kWh.
- Bandwidth Costs
Large IAAS companies located in carrier-neutral hubs like Equinix have the leverage to broker multi-carrier Internet bandwidth agreements in large (multi-10Gigabit) quantities. Such leverage, compared to companies buying at the Gigabit level in less-connected buildings can make a 7:1 difference in bandwidth cost. Bandwidth is how all those servers can serve all those users, so whether a company uses its own server rooms, or uses the Cloud, the need is still there.
- Software and Administration Overhead
Large IAAS companies provide customers with infrastructure automation dashboards that are extremely simple and easy to use. This can allow a company to maintain 1,000 servers for the same cost of managing 100-150 servers using a more traditional enterprise approach. Even companies that are adopting private clouds using automation and virtualization platforms like VMWare (which is still an excellent and revolutionary approach for Enterprises, and has worked great for our CIO here at Equinix) have to pay licensing costs that large IAAS companies do not, because they are built on open source platforms such as OpenStack.
- Utilization Sharing
Finally, because of the scale of IAAS across the needs of hundreds of customers, companies like Rackspace can share peak load demands and time frames across many customers, allowing much greater aggregate utilization of resources than each company doing this alone.
There are many excellent articles such as “Above The Clouds: A Berkeley View of Cloud Computing” that provide a more detailed analysis of the cost structure of Cloud and IAAS platforms at scale. I invite you to read them.
But the bottom line is that there are inherently large cost and resource efficiencies that overshadow any marginal sale-leaseback amortization issues.
Companies such as Rackspace, Amazon, Microsoft Azure, Virtacore, GoGrid, SoftLayer and many others stand a lot to gain by providing these efficiencies to companies gearing their IT platforms for today’s challenging economy.