Why should organisations be using the cloud over on-premise hosting? Many organisations are facing unexpectedly high costs in the cloud today and some organisations have made the strategic choice to retrench back to on-premise data centres. If cloud is really better than on-premise, why are some companies choosing to leave or pursue hybrid approaches?
When organisations first started to adopt cloud, it was naturally infrastructure teams that took up the challenge. These teams had a long history of understanding, specifying and ordering and commissioning hardware on-premise.
Hardware delivery takes weeks and it makes little sense to tune the ordering and commissioning processes to reduce latency. These same, relatively slow processes, were used to deliver cloud assets masking one of the cloud’s raison d'etre, the near instant provisioning and deprovisioning of assets. The assets considered were typically Infrastructure as a Service (IaaS), that is network, storage, virtual machines and later containers.
The infrastructure teams had less experience metering and delivering other capabilities and the result was organisations typically received an “on-premise in the cloud” experience.
In response to this, engineering or development teams were sometimes given and sometimes they took the role of cloud management. Developers were naturally quick to embrace software defined infrastructure and proceeded to unlock a good number of benefits of cloud. However, up until this point developers had always worked in a hardware constrained environment.
Some determination of the necessary on-premise hardware had been made early in a project and the software was written and tuned to fit that constraint. In selecting the necessary hardware a formal or informal process ensured that the cost matched in some way, the value of the solution.
With cloud, developers were no longer constrained, every log message could be stored, new instances could be brought online and more memory was just a click away. This resulted in the cloud cost bomb that organisations are trying to defuse today. According to the 2022 State of Cloud Cost Intelligence report, “only 3 out of 10 organisations know exactly where their [cloud] spend is going” *. Those that do tend to find it being spent on inefficient or wasteful solutions. It is this cost bomb that is driving many organisations to reevaluate their cloud strategy in favour of building on-premise solutions. However, it is bad execution which is creating the high costs not the strategy.
In most organisations, the strategy to move to cloud is driven by cost, agility and increasingly, the unique capabilities of cloud.
PAY ONLY FOR WHAT YOU USE
The cloud's lower cost potential comes from only paying for what you use. This requires a change in the way organisations think about, and use IT capabilities. For compute resources, you only need peak capacity at peak demand (or maybe slightly before). If a solution can be built to scale up and down effectively with demand there are significant cost savings available. Additionally, environments that are not required full-time, such as development or test rigs, can be turned off up to three quarters of the week.
The dynamic capacity argument also bears fruit in storage. On-premise, typically storage is budgeted, purchased and provisioned for the year. Anticipated storage needs for the year are purchased up front. The cloud can be cheaper in three ways:
1. Storage is only rented when it is required and not months in advance, saving money*.
2. Storage is charged for usage not capacity i.e. you don’t pay for free disk space on a drive.
3. If the storage is no longer needed, it is no longer rented.
Again, the dynamic capacity extends to the network. On-premise, you size your internet connection requirement based on anticipated need and then enter a multi-year deal where you typically under utilise capacity in the early years and are constrained in the latter years. In the cloud, you pay for what is used, not for what might be provisioned.
It is extremely hard to compare cloud and on-premise costs. The on-premise solution benefits from an unattributed share of the data centre(s) costs including:
Capital costs: building acquisition, power supply, initial network connection, cooling and fire suppression installation;
Operating Expenses: ongoing power consumption, internet network connection, security personnel, building maintenance and equipment-related costs.
Network costs: Includes costs for internal and external networking, cabling and network staff.
Hardware management: server setup, commissioning, testing, patching, upgrades, decommissioning, cleaning and disposal.
Support and maintenance: Expenses such as after-hours support, replacement equipment, servers and storage, software licences and their management.
The cloud always includes capabilities that may not be required such as data being duplicated, networks are segregated and redundant, logging always built in and so on.
Cloud costs are more complete and more easily tracked and attributed to different projects, solutions and services. This clearer view of spending and return on investment equips the organisation to make better decisions.
Agility is the ability to react easily to changing organisational demands to seize new opportunities or resolve emerging problems. Paying only for what you use enables agility and encourages the organisation to try things with no ongoing commitment, unlock learning through experimentation, rapidly iterate solutions and test ideas for better decision making. The flexibility of cloud computing fully supports agile development where spikes of activity with specific hardware can be supported and changes of direction don’t carry a cost overhead.
Cloud flexibility also enables compute operations that are not practical on-premise. For example, an entire solution could be temporarily replicated to investigate a subtle issue or examine a “what if” question. Large amounts of training data could be used for a one-off training of an AI model. Large amounts of data could be rapidly migrated using many machines in parallel. A full scale performance test could be run on an entire solution. In each case, these needs could be easily satisfied by provisioning cloud resources which are released at the end of the work.
These transient projects have significant compute needs and are time criticality and are just not practical on-premise. The agility unlocked by this capability is significant, speeding up work, encouraging progress, reducing risk and improving information.
The flexibility also ensures that the right tool is employed for each problem rather than accepting the challenge of bringing together a collection of disparate “spare” kit. The servers are always the current generation with enough memory, accelerator cards for AI are available when required, the network is segregated and low latency and near infinite storage is instantly accessible.
The specific technology can be chosen for the specific problem without the constraint of existing licences or prior purchases. If another technology choice turns out to be better the organisation can simply migrate with minimal cost from the original choice. This again promotes action by reducing the penalties of wrong decisions.
The rise of the dominant cloud providers has led to a standard set of Azure, AWS and GCP skills in the market. Skilled individuals joining the organisation on a temporary or permanent basis have less to get up to speed on. They are already familiar with the infrastructure, administration and other key capabilities in the cloud. This improves productivity and raises the quality of solutions as individuals bring best practice from other organisations.
Whilst cost and agility have been significant drivers from the inception of the cloud the unique capabilities of the cloud are becoming increasingly compelling. This includes normal engineering capabilities synonymous with cloud computing like databases and LLMs. It also includes the often overlooked built-in capabilities such as remote administration, monitoring and logging, security tooling and geographically diverse hosting.
It is no longer common for solutions to be built only using physical or virtual machines. Modern solutions require a plethora of managed capabilities from relation to graph databases, from directory services to content delivery networks, from AI services to API management and more. These managed functionalities play a crucial role in solution engineering by simplifying and speeding up the development process. They abstract the complexity and drudgery of installing and maintaining the infrastructure underpinning a capability allowing delivery teams to be productive quickly.
Cloud computing offers the ability for solutions to connect to niche services such as mobile text, mobile notification services or AI Large Language Models (LLMs) with minimal effort. These services are sometimes only available in the cloud and often the most up-to-date offerings available.
While these services can be accessed from on-premise there is both a latency and cost penalty. The cloud is likely to continue to be the focus of compelling future capabilities as cloud providers vie to attract users to their platform.
Administration capabilities are at the core of the value add of cloud providers. Remote administration, logging, monitoring and instrumentation makes infrastructure, platform, and software as a service possible.
During the Covid-19 pandemic, organisations continued to manage their cloud assets from home without a second thought. Today it supports distributed teams and external collaborators from anywhere in the world.
Cloud providers offer a shared security model where the application builder is responsible for a portion of the security. Additionally, the cloud provider secures the physical infrastructure and, depending on the service, the operating system and platform software. Network segregation and sensible defaults promote secure by default deployments. Tools are provided that support both proactive and reactive security monitoring. Few organisations can hope to do better than outsourcing a defined portion of their security responsibilities to cloud providers.
The cloud also offers the ability to host all of these capabilities in the widest range of geographic locations. This supports the most extreme service continuity geography requirements, allows network latency to critical users to be minimised and data sovereignty restrictions to be satisfied.
All these functionalities improve productivity, enhance workforce flexibility and remove the need to purchase and manage alternative services. In other words, they also represent a cost saving. At its core, the cloud enables organisations to build solutions that are just not practical or in some cases possible on-premise.
Many organisations are facing high cloud costs and believing a cloud strategy is flawed are returning on-premise, to some extent at least. However, cloud strategies driven by cost, agility and capability are well grounded.
Costs are hard to compare with many on-premise costs hidden and absorbed rather than attributed to projects and solutions. Paying only for what you use in the cloud offers the potential for large scale removal of cost.
This pay per use model is at the heart of the agility of cloud computing. The flexibility allows organisations to do things that simply are not practical or even possible on-premise. The cloud also ensures that the optimum solution components are available for use with abundant skills in the market to move quickly. This cloud agility encourages organisations to iterate and experiment gaining insights leading to earlier and better decision making.
However, it is increasing unique cloud capabilities that drive organisations from on-premise to the cloud. A wide range of management/administration, solution and security capabilities that can be deployed anywhere in the world enable organisations to deliver for their customers.
Those that can make the move encounter a world of convenience supporting incredible flexibility, access, visibility and delegated access. This world reduces complexity and noise for an organisation allowing it to focus on their customers and differentiating aspects of their organisation. That is, after all the point, there are more important things to do than run an environment for IT.
While there are many advantages to the cloud, perhaps the most important is simply “someone else is taking care of it”. Organisations are freed up to focus their time and energy on their customers and stakeholders.