Infrastructure Strategies: A Discussion Among Practitioners

Application Stacks or Hardware?


I’m getting pumped up for next week’s Wikibon Peer Incite (Tues 3/23 @ 12 noon EST). One of our guests is Omer Perra. Omer used to be the CIO of Aetna International and Joseph E Seagram. He, David Floyer and I have been doing some research these past few months with CIO’s, application heads and infrastructure leaders. We conducted some pretty in-depth interviews with these constituents along with Exchange and Oracle practitioners.    

We had three goals:    

  1. To understand how these constituents were making infrastructure decisions (i.e. what are the business drivers and objectives they are trying to meet).
  2. To explore ways in which organizations are deploying infrastructure function. Specifically, whether they are favoring placing storage function into application stacks (e.g. Oracle ASM or Microsoft DAG) or if they favor using a shared services model (e.g. SAN)
  3. Model the total cost factors of each of these approaches over a four-year period.

At this meeting we plan to discuss the research findings with the Wikibon community and get their feedback. Several of the practitioners that participated in the study plan to be on the call so we’re really looking forward to what they have to say.    

Here’s a brief summary of what we learned:    

  • Once CIO told us “infrastructure is evil.”
  • Application heads said they cared about performance, performance, availability, performance, recovery, SLAs, function, speed-to-deployment, performance, the end user experience…and if there’s any room left – yeah sure costs too.
  • Infrastructure leads told us this: When chargebacks are in place they can relatively easily promote a shared services model; when they’re not in place the application teams and businesses pretty much get what they want when it comes to infrastructure.

We explored two broad strategies with these practitioners, specifically place function in application stacks or in shared infrastructure (e.g. a SAN). We found that practitioners utilize the application-oriented approach to simplify storage infrastructure and streamline recovery. This was especially true in Oracle environments whereas in Microsoft shops the decision was more related to size of installation, influence of Microsoft and overall skill sets.    

Application-centric or Shared Infrastruture


What are Shared Infrastructure Services?    

In 2008, John Blackman put forth a model to the Wikibon community that showed infrastructure services as a three-dimensional construct (infrastructure components, data protection and performance tiers). The idea is to construct a menu of services where the intersection of these three dimensions can be priced. Here’s the picture John painted:    

Infrastructure Services Cube (Source: John Blackman and Wikibon)


Cost Modeling – Return on Assets    

We used a modeling technique focused on measuring return on assets (ROA). It’s similar to ROI but differs in that it attempts to capture the cost metrics of installed infrastructure over a life cycle. We modeled DAS versus SAN and measured only costs, not other business value.    

The following charts show a high level comparison for different strategic options with a drill down of cost detail at 200TB.    

Comparing DAS and SAN in Oracle and Microsoft Environments


*Oracle is virtually always more expensive due to licensing and maintenance costs.    

*DAS for Microsoft Exchange 2010 is cheaper in small installations but substantially more expensive at scale.    

*The main reason for SAN’s cost advantage relative to DAS at scale is that SAN is a re-usable asset across the application portfolio and this leverage improves utilization. It also has ripple effects on other infrastructure such as server count.    

Cost Drilldown at 200TB


Please join us on the call– we’re expecting a number of practitioners who participated in the study and we’d love your input.    

Additional Resources:    

Infrastructure Wars: The Battle Brewing in the Storage Industry— a high level blog post summarizing the research.    

Published Research    

To assist IT professionals in making better decisions, we’ve published several research notes and blogs. The following related research items can be used to support your goals:    

  1. Why Microsoft’s Head is up its DAS – a summary of how Google is forcing Microsoft’s hand and pressuring the company to do anything to lower the perceived cost of Exchange, including recommend DAS exclusively.
  2. How Google Apps Threatens Microsoft – the detailed research behind the blog post (#1) with a cost analysis of Exchange on-premise, Exchange On-line and Google Apps.
  3. Should Exchange 2010 use DAS or SAN? An economic analysis of DAS vs SAN in Exchange 2010 environments; including a view of where costs cross over.
  4. Should Storage Services Reside in Arrays or Application Stacks? An objective view that lays out the two broad strategies for placing storage function with a specific drilldown into Exchange 2010 and Oracle 11g.
  5. A Storage Services Approach will Streamline Infrastructure Delivery: A definition of a shared storage services model with a financial assessment comparing the total cost of ownership of shared storage vs. a One-off (bespoke) orientation.
  6. Comparing Return on Assets (ROA) to  Return on Investment (ROI) – The Wikibon ROA model underpins the financial analysis used in this research. The model specifically evaluates the value of installed assets over a period of time and is particularly useful in assessing asset utilization across an application portfolio.  

As always, we appreciate your participation as a Wikibon member to help your Peers make better IT decisions. We hope you find this information useful and we look forward to your feedback and assistance with future initiatives.