Cloud’s economic model is unique. Whether you are a cloud sceptic following a cloud-first mandate or are completely bought-in to the promise of agile infrastructures you must understand the sometimes counterintuitive cloud economics. Whatever you are concentrating on you will almost certainly have to justify to your CFO the increase in Cloud spending. While the better speed and agility will prove their own value how can you be sure that your internal users are employing responsible practices to keep costs efficient?
Here are ten facts about cloud economics that will help you identify how your organisation can optimise cloud usage.
1. VARIABLE COST AND SPEED MAKE PUBLIC CLOUD UNIQUE – AND INCONSISTENT UNTO ITSELF
Cloud isn’t the only solution but it does provide you with more sourcing options. Public cloud leverages many familiar concepts like standardisation and automation but the speed and variable pricing model are new and you are no longer locked into multi-year contracts or large server purchases. But beware, not every application benefits from this model. Variability isn’t always beneficial when you compare it with the discount you get with a longer commitment. Part of your cloud due-diligence will be properly understanding the incentive system of any model you select.
2. IT TAKES BEST USE TO LEVERAGE THE PRICING MODEL
Low costs per virtual machine aren’t what make cloud cheaper. Cloud infrastructure saves you money only when you aren’t using it. Best fit workloads are those with transient or dynamic properties. Buying new servers to accommodate a short burst in usage isn’t cost effective but using public cloud can and will let you scale as you need and pay-per minute / hour. But remember to properly analyse how long your ‘short burst’ is actually going to be, cloud may not be the most cost effective route.
3. BECOMING A CLOUD PROVIDER CHANGES THE MODEL
If you build a private cloud you become a cloud provider and the economics change. If you want to see savings you will need to build into your model: the high cost of software, high-end infrastructure, supporting performance expectations, maintaining excess capacity and meeting developer expectations. Follow in the footsteps of public cloud providers, focus on net-new services and build with standardised commodity components. Failing to do this leads to over-spending, documentation shortages and falling short on service level agreements.
4. INTERNAL SUPPORT COSTS WILL NOT DROP TO ZERO
Moving some or all of your business to the public cloud does mean someone else will be running the infrastructure but you are still responsible for managing, securing, monitoring and backing up cloud deployments. Facility management and hardware support will diminish but new governance and integration responsibilities take their place. Some cloud providers will offer you some help but remember that it might not be free and you still have to protect all your IT assets and ensure the performance and availability. If your public cloud ROI is dependent on headcount reduction you may well be disappointed.
5. SPEED, AGILITY AND PRODUCTIVITY ARE THE KEYS TO MEETING ROI
Unless your data centre contract is coming to an end your tech support is entirely consultant based your tech management costs are only going to go up with cloud. Cloud enables a long list of things that simply weren’t possible before it’s existence but now need tech support and implementation such as genomic processing and supplying resources for two-week marketing events.
The real ROI is the increased speed and agility which translates into faster, better customer engagement. Better cloud management helps to optimise cloud usage and spending meaning that the customer experience can be the priority. Remember that developers want speed and agility and the cloud makes it easier for them to circumvent your infrastructure to get it. You can deliver the autonomy they want via self-service portals and application programming interfaces and protect our healthy infrastructure policies and consistency with templates that abstract the details.
6. PUBLIC CLOUDS DONT FOLLOW BULK-BUY PRICING MODELS
Public cloud providers increase their margins by pushing average sustained utilisation rates as high as possible. Providers do this by moving around customer workloads to minimise the number of physical machines running.
A jar full of rocks always has room for the sand you pour into the remaining space. By encouraging customers to buy lots of small VMs cloud providers can improve their utilisation and so their margins. They will financially reward you for breaking apart your large apps into smaller components. Don’t be wooed by the price, many of your applications won’t adjust well to this seemingly minor change and you might find yourself spending money to save money.
7. HYPERSCALING PUBLIC CLOUD VENDORS ASSUME FAILURES WILL HAPPEN – SO MUST YOU
This is the classic pets-versus-cattle analogy. Cloud providers use commodity infrastructure knowing that they will have a higher VM failure rate. When this happens they terminate it and start another. This places the onus on the customer to design for application resiliency rather than infrastructure resiliency. You may be happy with this change but it might present significant issues for your existing applications. If an application is not built for the cloud (loosely coupled and highly scalable) an abundance of small VMs that frequently fail will mean poor performance. This is not a weakness of the providers it is a fundamentally different approach. If you choose to host workload in a public cloud you will need to calculate the cost of any work required to make this seamless.
8. EXISTING APPS WEREN’T BUILT TO LEVERAGE CLOUD EFFECTIVELY – REDESIGN OR REPLACE
You might be planning to move existing systems-of-record applications onto public cloud. This may well mean redesigning or rewriting your application and that’s not simple. Breaking applications into smaller components will allow for more granular control over scalability of each element. Components that need more capacity will include a set design template, automating the creation of new machines that can be run independently. You may need some duplication of controller information to remove bottlenecks and dependencies on one machine. The work may be onerous but it will be cost-effective; independent scaling, self-sustaining mini-components will help manage costs during peak usage and increase resiliency. Loss of a VM will now mean reduced capacity, not a systemwide failure.
9. VALUE-ADDING FEATURES AND SERVICES TRANSLATE TO LOCK-IN AND COST
You may well be tempted when you see that Amazon’s many additional services can eliminate thousands of labour hours from you app teams and yield a self-sustaining service with minimal maintenance. Be careful; the more you acclimatise to using high-order features and services, that seem so cheap individually, the harder it will be to transition to another provider who does not offer the same capabilities. You have to weigh the value of the services with the lock-in and potential migration costs against the long-term value of the cloud provider you are using. Determine the level of abstraction and the amount of choice required by your customers but keep in mind that every add-on service you adopt correlates to the amount you are locked in. Of course, lock-in may not be bad if your provider is giving you great value but don’t restrict your options. Revenue benefits may erode while switching costs rise.
10. BEWARE OF HIDDEN COSTS
It’s a cliche but there are always costs you didn’t expect and it is worthwhile making sure you have explored as many of them as you can. This is a short list you might benefit from considering: Software licences; Data-out charges; Mitigating latency; Direct connections; Onboarding error rates; Migration charges; Employee time (not just IT, remember finance, HR and legal); Backup and business continuity. Ask the cloud providers about them and see how it might alter their offer.
The cost benefits of moving to the cloud are real and worthwhile but make sure that you manage your expectations and do the work that’s needed to maximise your spend.
Thank you Forrester – https://go.forrester.com/