‘Shared Servers’ is not ‘Shared Services’
Doug Hadden, VP Products
Data centre consolidation is not easy pickings
Data centre consolidation is seen as the ‘low hanging fruit’ for shared services in government. Economies of scale, lower costs, better performance. There are shared servers projects planned by the federal governments of Canada, the UK and the United States.
Beware the low hanging fruit
Sharing servers does not provide the benefits of shared services (standardization, risk mitigation, governance etc.) And, shared servers can cost governments more than the alternative: keeping data centres as they were.
Data centre consolidation is the plumbing for shared services in government. Failure to meet cost objectives in shared servers does not bode well when governments attempt more complex and valuable software shared services. That’s because of some recent shared services failures in the UK and Western Australia where shared services cost more money than non shared services. For example, the UK, “the project was due to be completed by December 2009 at a cost of £79m but actually only finished its development phase in March 2011, at nearly double that price tag – £130m.”
The potential higher cost of server farms
Where are these hidden data consolidation costs?
- Setup costs – a recent analysis by Price WaterhouseCoopers found that data centre consolidation “could eventually save between $45 million and $293 million” but at a cost of “between $145 million and $278 million in one-time costs.” Therefore, the return on investment is as low as 116% and as high as 202%.
- Management Layer - management and organizational communications costs increase to coordinate among so many government organizations.
- Personnel Costs – there can be significant re-training costs to update personnel on newer systems and support more complex network and systems management tools. The Price WaterhouseCoopers report suggested that major savings would accrue thanks to a reduction in IT staff: “the business case for this strategy, to achieve its maximum benefits, will have significant labour implications.” So, there will also be compensation packages.
- Timeliness – the time required for consolidation in any large government means an opportunity cost. New technology is rapidly being introduced. Operating systems and databases systems are being upgraded. And, new low-cost open source middleware is becoming more available. There is a risk of consolidating old technology. And, there is the further risk that many of the services contemplated to operate on the shared services can be securely and less expensively deployed on the cloud, as we see in the United States.
- Formalized SLA and QoS – the need to maintain formal service level agreements and Quality of Service requires equipment and tools. It’s one thing for a departmental server to crash. It’s another thing altogether if the shared service isn’t consistently reliable, as we have seen with failures in Google Mail, Twitter (fail whale), Salesforce and Blackberry. Customers demand reliability – even for free services. So, this means redundancy, back-up servers, back-up centres and fault tolerance. (There was a Forrester Research report a years ago that explained the huge incremental cost to achieve “five 9’s” availability.)
- Distance – data centre consolidation extends the distance between user and server, particularly in big countries. Servers in Canada or the United States could be in different time zones than users. This often means increasing bandwidth or adding regional centres to overcome the increase in reliability and performance related to distance.
- Peak of peak activities – data centres need to be planned based on expected activity. This means the need to deploy servers to handle the highest volume. Governments operate on the same fiscal period. Therefore, an exclusive government data centre needs to be provisioned based on peak activities around government year-end. Therefore, any advantage by balancing the load across multiple servers can be negligible when the data centre is provisioned to handle peaks of 5 or 10 times the nominal volumes.
- Platforms – governments operate many unique applications because of mandates or so many “Lines of Business”. This means that any data centre consolidation requires the support of many technology platforms. This can include supporting different versions of different platforms with different fixes, security vulnerabilities, scalability profiles, and upgrade complexity.
- Virtualization isn’t magic – not all applications and platforms play nice within virtualized environments. Virtualization is not magic to balance loads across a server farm. Switch load balancing, technology capacity planning is also required. The complexity increases with numbers of platforms and apps.
- Network and Systems Management – there is likely a need for more sophisticated network and systems management tools to monitor health across the data centre requiring the creation of a Network Operations Centre.
- Root cause analysis – it can be more difficult to uncover performance and reliability root causes because of last six points. It’s easy to reboot the IIS server in the server in the wiring closet to see what happens. It’s another thing to diagnose server problems across a farm of racks filled with blades.
- Power and reliability – one of the key benefits to data centre consolidation is reduced power consumption. Except that these data centres require operating 24/7 and special power provisions and back-up generators.
Bad fruit doesn’t make good wine
I’m not saying that these ‘shared server’ initiatives in government will fail. My point: there is significant risk. And that risk pales in comparison to shared services.
Data centre consolidation success is a good sign on the way to shared services. Consolidating office systems like electronic mail, word processing and collaboration is the next stage up the vine. Although, I wonder whether deploying these as a private cloud within government because of security concerns can be justified. Gartner analyst Andrea DiMaio wonders whether this will stall movement to the public cloud when those services are secure and cost less. And the move to operational financial and human resources systems presents more risk as I described in a recent FMI presentation.