Original URL: http://www.theregister.co.uk/2011/05/11/cloud_scale_cost_considerations/

Trevor Pott's guide to pricing up the cloud

Does scale equal less cost?

By Trevor Pott

Posted in Cloud, 11th May 2011 07:00 GMT

As a sysadmin for a small-ish business, I lack the resources to keep an in-house expert who is intimately familiar with the intricacies of each technology on my network. There are so many technologies from so many vendors: it is simply overwhelming for two sysadmins to try to know everything between them.

So why not outsource everything, engage an appropriate consulting firm and put all IT in the hosted cloud? After all, the marketing message of recent years is clear: the sheer scale of clouds ensures lower costs.

In reality it is more complicated to measure the suitability of hosted clouds.

For some, the cost savings are clear, the maths as simple as those of an internal cloud. For most, hosted versus private clouds are so close that the make or break points lie in the intangibles; items for which no ledger entry exists.

So, I invite you to consider two important questions before asking if clouds – hosted or private – cost less. Just exactly how much are the time and focus of in-house IT staff worth? How much would be freed up by embracing cloud services? When you can answer these, you can answer how much cloud computing is worth to you.

In-house knowledge

My case for the value of in-house IT staff is that they are indoctrinated into the corporate culture. In-house sysadmins know the ins and outs of a particular workplace’s corporate politics and they understand what it takes to get things done. This is vital for support and also informs which approaches to new services are the proper fits.

In-house staff have incentive to address the issues that matter to the people who actually work there. It is their job to understand what is important to an individual, a department and to the company as a whole. Being in tune with the company is priceless; it can take longer to explain to a consultant why X must be done in Y fashion than to simply do it yourself.

But in-house staff are not the best fit for everything. When you have more services than can reasonably be expertly maintained by your current level of staff you are faced with an age-old question: hire more people or explore outsourcing?

Smaller companies in particular lack the resources to retain an individual on payroll covering every aspect of every service their business requires. From Banking and legal through real estate and cleaning, running a lean business requires SMEs to outsource any services not related to core competencies.

For many SMEs this list of services includes IT.

Service mentality

In today’s terminology, we can group many outsourced IT services under the umbrella term of Cloud Computing.

Cloud computing can be defined, broadly speaking, as tech services and applications delivered - mostly - over the internet. Typically, in a public cloud, customers are charged on a utility basis - everything is wrapped into a monthly per user fee.

Today’s most commonly accepted interpretation is that cloud computing refers to pools of computing resources, to “serve multiple consumers using a multi-tenant model, with different physical and virtual resources dynamically assigned and reassigned according to consumer demand”. This has the virtue of simplicity for customers - there is no up-front capital expenditure, no heartache over software management and licenses, and no server rooms or data centres to maintain. The potential for cost savings here is obvious

Very briefly, we can characterise cloud providers under three “as-a-service” categories: Software as a Service (SaaS), Platform as a Service (PaaS), Infrastructure as a Service (IaaS) and so forth.

IaaS

It is increasingly easy to realising value from hosted cloud services. Not having your own IT plant removes direct costs; equipment, licensing, power, cooling, rent and other overheads.

IaaS, or infrastructure-as-a-service, enables sysadmins to refocus efforts away from keeping the lights on; allowing them instead to focus on R&D that moves the business forward by using technology to streamline other business elements.

IaaS is the ability to take your existing infrastructure – typically, virtual machines – and host it in someone else’s data centre. The reason to do this is simple; large IaaS providers have redundant bandwidth, power, cooling and global distribution unavailable to smaller organisations.

IaaS providers can also give their customers the ability to rapidly provision servers and bandwidth to deal with spikes in demand, say, or to roll-out services much more quickly. (The Open Group explores the ROI of Cloud and elastic services more fully, here.)

Depending on circumstances IaaS may very well allow organisations to get by with fewer sysadmins. This is an important consideration, as payroll is part of the burden companies are looking to the cloud to help them avoid.

But users of IaaS will still typically require at least one sysadmin. Hosted infrastructure is still infrastructure; management of applications is not normally part of the deal.

Software as a Service

SaaS is the easiest category of cloudy offering to understand.

A developer writes an application, hosts it somewhere on the internet and you pay for it. A subscription model is typical, but by no means a requirement. SaaS applications are typically delivered through a browser, but not always. Hosted Exchange is one example offering full functionality via both traditional and browser-based email clients.

For SMEs, cloud-based software gives them access to software that they could neither afford to build themselves or buy in, or it could free up resources for business-critical work.

For instance, The Register, home to about 50 staff, has three software developers, who concentrate on the home-grown content management system, improving the websites and reader accounts systems (and a bit of sysadmin work.) CRM, accountancy software, office apps and email are all contracted out to SaaS providers.

Do companies really need to run their own email systems? Hasn’t this technology been around for long enough that we can trust someone else to take care of it without problems? What about the corporate instant messenger, spam filtering, malware protection, web hosting, CRM or dozens of other “basic services"?

These are mundane items; no individual service consumes an inordinate amount of time. Yet in my environment, these “minor” services occupy dozens of virtual machines on several servers. Maintenance overhead alone is at least a month’s worth of my time every year. This is before taking into consideration of testing and replacing the relevant hardware each refresh cycle.

Integration of these services is where the real time is consumed. Cloud providers with multi-application “stacks” can deal with some of this as well.

For those applications sourced outside of a stack, integrating them into existing – or planned – infrastructure is rarely more difficult than integrating those applications would be were they run locally. In the same vein, if you have a stack of applications running locally that are all tightly integrated, breaking this up to host elements elsewhere makes little sense.

Vendor scale

Elsewhere in The Register Cloud Channel we write about risk, compliance and security issues ( - for instance). Here we simply record the argument that scale economies enable vendors to deliver service improvements in these three areas.

Over to Reg reader 'richard.cohn', who points out:

"Looking at "Software As A Service"-Style cloud vendors like Salesforce and Google (apps), they do indeed have quite strong incentives (read: funding, staffing) to "do security properly". Security is part of their "core business" and not just part of that "support function IT".

A single highly competent (read: expensive) security expert can secure millions of Salesforce users, while an inhouse-system will never get that attention."

Note, that vendor scale does not necessarily translate into lower costs for customers here, but that surely is a secondary consideration.

Platform as a Service

PaaS is a booming cloud offering targeted at developers. A PaaS provider offers up APIs allowing applications to transparently tap into feature sets, redundancy and sheer scale developers would otherwise be unable to front. A small developer can go from “my first smartphone app” to a SaaS offering living in a dozen data centres around the world overnight.

Software development studios will find the greatest benefit here; building their offerings on top of a PaaS stack is a quick route to marketing a reliable SaaS application. Developers only have to focus on the code; everything else is taken care of for them.

End users benefit from PaaS offerings as well. SaaS on PaaS applications are more robust, reliable and scalable than SaaS on custom infrastructure. SaaS on PaaS also enable small and mid-sized developers to provide higher quality application at a lower cost than do-it-yourself options.

The downside is lock-in; a trap developers and end users alike are justifiably terrified of. Once your application has been coded to the API of a specific PaaS provider, porting it can be very difficult. Should that PaaS provider alter their terms of service in an unacceptable manner, fail to deliver on their SLA or go bankrupt, the developer and all their clients are out of luck.

Workarounds are beginning to emerge; the Eucalyptus project famously enjoys close compatibility with Amazon’s APIs. Windows Azure, Microsoft’s PaaS, is also interestingly positioned. Other vendors offer projects which are designed to allow multiple providers to create offerings based on a common API. Hope for future interoperability is also present amongst the various cross-provider cloud API projects.

Conclusion

This as-a-service model is an important shift in how outsourced IT services are marketed. Admittedly, there are some basic differences with the outsourced services of yesteryear, especially in pricing models, but at their heart, most cloud computing services are outsourced services.

And at a minimum, each service outsourced increases administrative overhead. The latest trend in IT outsourcing – the hosted cloud – also offers up questions about SLAs, portability, privacy and ownership of data.

Despite this, I remain optimistic. Internal clouds have already demonstrated their value. The major kinks have been worked out, the ROI proposition is clear. The hosted cloud offers the possibility of spending less time with maintenance. If it can deliver, then the opportunity exists - with the right services-split - to greatly reduce the workload of existing staff. ®

Trevor Pott is a sysadmin, based in Edmonton, Canada