As with cars, music and razors, the way we pay for computing power is changing. Software-as-a-service and ‘serverless’ hosting are picking up momentum based on promises that you only pay for what you use, but the opacity of pricing models and a lack of preparedness among customers can lead to costs ballooning. Here are some tips to help you maximise the value of cloud computing.
How we pay for goods and services is changing. Music is now rented, rather than bought. Nearly nine out of ten new cars driven off the forecourt are leased or bought on contract, not owned outright. Razors come through the post for a monthly fee.
This shift towards subscription and monthly contracts can be attractive to cash-strapped users. But it can also mask poor service and true costs. Streamed music is of lower audio quality than CD or vinyl, and the musicians and writers earn less. The leased car is often ultimately more expensive for the user than if it had been bought. The subscription razors would often be cheaper bought from the chemist when needed.
Dealing with the complexity of cloud pricing
The same shift to on-demand is happening in business. Most enterprises are pushing their digital services into the cloud where possible, where the economic model is typically ‘pay-as-you-go’ or ‘as-a-service’. But as with music, cars and razors, this does not automatically mean services will be better or cheaper.
Controlling spending in the cloud is far from simple. Large organisations are struggling to calculate the costs and risks involved. In fact, according to RightScale’s latest State of the Cloud Survey, cost-control has become the number one cloud concern for IT leaders, overtaking security and skills-availability.
There are several reasons cost management is such a challenge.
- Cloud pricing models are complex and rates can be changed without notice by vendors.
- Prices in public cloud are vulnerable to real-time exchange rate fluctuations.
- Enterprise business cases tend to be based on allocating a certain amount of capital to do a thing – cloud does not fit that model.
- Procurement teams are used to buying things in chunks, and the default bargaining tool is volume. Cloud prices do come down with volume, but the standard model is pay-as-you-go, rather than, say, paying a set price for three year’s worth.
- Enterprise agreements, where customer and vendor strike a bespoke deal, do exist in cloud, but they are not the norm. Arguably, they should be phased out in favour of using published, scalable, universal rate cards.
- Cloud computing is still new to most organisations, and services are expanding rapidly. In general, enterprises lack the skills and experience to design and deploy as confidently as they did with their legacy systems.
- It’s very easy to over-provision and overspend in cloud computing. Machines and services must be deployed very carefully to avoid needless run time. Every company on the learning curve has experience of leaving services running and paying more than necessary.
- Cloud encourages self-serve in the organisation. This means close controls must be introduced to ensure newly-empowered users behave responsibly.
- Moving an enterprise to the cloud is a major, multi-year undertaking. There will be an extended period of managing and paying for cloud services in parallel with on-premise services, which compounds the difficulties in cost management.
Can you haggle over the price of a cloud?
In public cloud, the lead vendors are diverging. Microsoft Azure is developing as a software-as-a-service company and still uses enterprise agreements. AWS is an infrastructure player, but is becoming a rich repository of ready-to-run software from multiple third-party vendors. AWS delivers, for the most part, against published rates.
Google’s mix of infrastructure and software options are a becoming a genuine alternative for productivity, storage, collaboration and data management for enterprise customers. Pricing is by consumption, but not necessarily easy to navigate. As Google pushes further into enterprise, big customers have room to negotiate.
Big-ticket enterprise vendors like Oracle and SAP are moving their existing business models to the cloud. Platform players like Salesforce are pure cloud and tend to operate on a monthly licence model, but generous discounts can be achieved if large volumes of high-level licences are booked.
Tips for managing cost control in the cloud
- Design to scale services as usage increases, rather than provision big just in case.
- Set up the tightest possible cost monitoring tools from the outset. The big cloud vendors offer sophisticated capping and alerting systems as standard, but they have to be actively set and switched on.
- Keep architecture as simple as possible, and design for the cloud. Migrations of existing architecture can work, but the move to the cloud is an opportunity to reconfigure services.
- Set roles carefully, so that only appropriate members of staff can scale services and increase costs.
- Avoid enterprise agreements – they tend to be win-win for the vendor. Instead, choose the appropriate tariff for your needs and build in flexibility from the start.
- Each new cloud service should be designed according to user need. Do not just take the recommendation of the dominant supplier, or be tempted by a bundled service – make competitive comparisons.
- Where possible, create a central cost dashboard. This may be impossible in a giant enterprise, but there should always be high-level controls to avoid a productive self-serve environment turning into a free-spending Wild West.
- Take a deep look into the cost-saving opportunities in cloud computing; it is reasonably easy to reduce spend significantly by judicious use of block-booking and off-peak usage.
- Stay close to the vendors and their partners. So intricate is the science of spend control in the cloud that expert knowledge is required to ensure good value.
About Jon Holt
Jon is a partner at NotBinary, where he runs the cloud practice. He advises multinational organisations on technology investment, cloud migration, architecture, serverless design, data lake creation and supplier relations.