Since 2012 public cloud has been one of the hottest and most discussed topics in the IT industry. Headlines such as, “Cloud is the Future” can still to be found everywhere, but in 2018 public cloud is no longer the future, it is a reality.
Even though many organisations varying from small businesses to large enterprises already have some degree of experience with cloud, most won’t call their large-scale cloud projects a success. The main reason for this is the long-term effect of their cloud journey being less rewarding than initially planned and budgeted, and sometimes even more costly compared to their previous on-premise solution.
An organisation’s cloud bill is often the single highest expense of IT departments. Many organisations struggle to forecast cloud spend accurately enough so that the budgeted amount suffices to cover the project over the budgeted period, or just as often, fail to consume the entire amount committed to the cloud provider.
Here are some common pitfalls affecting organisations with experience in on-premise IT – on their journey to the cloud:
1) Negotiating a short-sighted contract
Cloud is evolving quickly and negotiating a contract without taking into consideration the extreme pace of change in the industry will result in higher spending.
Most cloud providers offer discounts on prepaid commitments to spend a certain sum over a period on any of the provider’s cloud products or monetary commitment to specific products (SKUs).
Larger organisations tend to make a large upfront commitment for 3-5 years and in return typically receive a fixed discount on the cloud products, as well as some free credits for the initial period (typically a year). To most procurement specialists this would sound like a great deal. As it would to the IT team, they will essentially be handed the keys to a promised land of unlimited possibility and freedom, to experiment without having to care about the costs.
Unfortunately, this almost always leads to IT administrators and developers becoming less responsible in managing the cost of their cloud resources. They tend to take the largest, quickest instance instead of the right one for the job. People leave projects running longer than needed or even forget about running a test environment for days, where only a short test was supposed to be conducted.
As a consequence, the free credits will disappear in no time, the monetary commitment will get spent far sooner than expected and the first invoice for overconsumption will arrive. In most cases this will catch an organisation by surprise and will force them to find the additional funds from other budgets.
Some organisations are more cautious about their cloud projects – they will conduct a study around which internal systems they can/should move to the cloud. These companies will negotiate with the premise of knowing how many and which instances they would use over the next few years. In such a case, providers will often offer to fix the price-list over the contract duration and discount some very specific cloud products (SKUs).
Again, this may sound like a great deal. Providers won’t be able to increase the cost of service over the contract duration and will give a discount to the products which are supposed to be used, right?
It is important to remember that the compute SKU in the cloud is not like the software you purchase a license for. You can’t just upgrade as long as you have active maintenance. It is hardware, a virtual machine instance type is simply the version (model) of the hardware, equipped with a specific processor/network card/GPU etc. Your long-term commitment to this SKU means that you intend to use it over your whole contract duration.
Processor manufacturers release more energy-efficient CPUs with higher core counts all the time. The main cost contributors of the cloud providers are space, electricity and cooling. To drive down their costs of service, they adopt new technology as quickly as possible, retiring their old components. It could be that yesterday, the instance type you were negotiating was state-of-the-art, however in a year, it may become outdated and you’d be better off using a new instance type at the list price, costing lower than your outdated instance type with a discount.
The cloud industry is a very competitive and major cloud providers are in constant competition with each other. Hence, they always look for new ways to lower their costs and decrease their prices, as opposed to increasing them. Which is why fixing a price list today and receiving a discount for the fixed price for three years may turn out to be more expensive than being able to benefit from price drops.
2) Moving too quickly, for the wrong reasons
Contract negotiations aside, some organisations get cloud as a free add-on to their licensing agreements or begin their journey because it is trendy and everyone it is doing it. Typically the board will decide to move to the cloud and will task IT with getting them there. IT departments main task is usually ‘keeping the lights on’ and they are prone to a lack of time for the careful planning of a migration. In most cases, to save precious time IT departments will choose to migrate their on-premise estate using a “lift and shift” migration method.
Although technically easier, less risky and quicker, such migrations are thought to be cheap. It may only take a few days to perform a big bang migration with the only bottleneck being the speed of a connection used to upload the data into the cloud.
Most organisations underestimate the magnitude of longer-term costs, which will start ramping up as soon as migration has taken place. For the same reason the IT department decide to go with the ‘lift and shift’ migration (due to lack of time), they see their job as done and won’t optimise the solution once the migration has occurred.
As a result, the resources will not be maximised and the organisation will end up paying a substantial amount for thin air. In fact, according to a survey conducted by Right Scale, about 30-35% of cloud solutions are wasted.
3) Not embracing the benefits of cloud
Cloud providers will often sell cloud as the possibility to save on IT costs. Many organisations fail to fully embrace the benefits of cloud technology, one of the biggest benefits being per minute/per hour accounting.
Normally a server is accessed by users between 8:00 AM and 8:00 PM, i.e. only 12 hours a day. Subtract the weekends and you get your requirement to run a virtual machine (VM) for roughly 276 hours per month vs. 732 hours if the VM were to run 24/7.
Automated start-up and shutdowns of VMs is something fairly easy to implement and won’t require any modification of the actual software. That is a potential 62% costs reduction, which is often wasted.
4) Reserving more capacity than necessary
Some business workloads are required to run 24/7. They are predestined for cost-saving vehicles such as reserved instances or committed usage discounts. They are often positioned by cloud providers as a cost management silver bullet – just buy them and start saving money.
Such capacity reservations typically include an upfront payment for a specific instance type, in a specific region. They can be scoped for a single account/subscription or they can be shared between accounts/subscriptions within the organisation. This means the number of instances as reserved instances won’t be billed or will be billed at a lower rate. The next virtual machine will be billed at the standard rate.
There is no 1:1 allocation of the instance to a virtual machine and organisations often struggle to purchase the right amount of capacity reservations to satisfy their needs, as reserved instances are applied on a first come, first served principle. Quite often, reservations are “snatched” by instances they weren’t intended for and a larger than required number of reservations are being purchased, ending up unused.
Requirements change and newer, cheaper instances providing better performance are being released by cloud providers constantly. Changing the instance type to a cheaper one frees up the reservation until another instance of such size is started. And if the instance type has become outdated, which is something to expect within the three years of reservation lifetime, the reservation may get rendered useless. Some cloud providers offer the possibility to return the reservation for a fee, but ultimately the money is wasted if the reservation becomes unused.
5) Putting all of your eggs in one basket
At the beginning of a cloud journey many organisations decide to use one cloud provider, assuming they would get longer-term benefits by investing in building skills for one specific cloud platform. This strategy is justifiable if an organisation is making use of the technology specific to this platform, such as serverless computing, artificial intelligence, big data etc. In reality, about 60-80% of the cloud services used today are Infrastructure as a Service (IaaS) and Platform as a Service (PaaS) which are provided by more than one cloud provider, differing only by price.