We all talk about moving to the Cloud…
…but few understand the financial implications of transforming from an oversized but underutilized CapEx driven operational model, to a consumption driven OpEx model. Some claim that moving to cloud saves money, while others comfortable with on-premise enterprise IT, claims that building and operating IT locally is much cheaper. Who is right, and who is wrong?
Traditional on-premise IT is a mix of CapEx and OpEx. Capital expenditure, or CapEx, are funds used by a company to acquire or upgrade IT infrastructure, platform and software, while OpEx is the total cost of operating the CapEx investment. In Cloud IT the method is entirely based on OpEx, because both capacity and operations is included in the service.
I’m often told, that one of the major drawbacks on moving IT into the cloud, is the lack of transparency on the total cost, which is surprising due to the fact that every asset in Cloud Computing is priced very clearly, while the cost of on premise IT is almost impossible to calculate.
What is the true cost of on-premise IT?
Calculating the true cost of on-premise IT is a daunting and difficult task, often neglected or misinterpreted when comparing to cloud IT. This is due to a variety of different of things.
Complexity
Who can truly price an on-premise SQL database with 99.99% uptime guaranteed including backup and disaster recovery? You would have to factor in parts of: data centre investments and operations, hardware investments and operations, software investments and operations, backup, disaster recovery, service-desk, database administrators, security and all expenditures around IT in general? The cost of a database would be (Total IT costs * X% of in-direct IT costs + direct database investments & database operations) / (number of active databases). Complex? The truth is that most cannot calculate the true cost of on-premise IT services due to the complexity.
Hidden costs
The way on-premise IT is funded is also a complex part of the calculation. Often business divisions who gets funding for IT projects, pays for establishing the new software or pays for a storage expansion unit with project funding, because internal IT is underfunded, and ceases the chance to upgrade or renew parts of their platforms. Project funded infrastructure expansions or upgrades are hardly ever calculated as a part of the IT budget, but rather a regular business activity. In some companies, the housing costs or power costs are on different financial accounts, not visible in cost of operations, while in other companies, people in business divisions spending time on IT are not calculated as part of the IT investments or IT operational costs.
Oversized, under-utilized
With on-premise IT, you are punished if you purchase and provision IT with too few resources, thus the typical investment is huge (over capacitated) because it must last for 3-5 years at least, and coming back for more money every year makes you look like an idiot at capacity planning. Virtual servers are provisioned and oversized, to ensure the business doesn’t complain about performance, resulting in overprovisioned servers, that are underutilized. If a business case calculates the server cost at i.e. € 50 per VM per month, the cost is often 3-5 times more because of overprovisioning and underutilization.
Technical debt
Another cost factor is depreciation, where companies in technical debt, often tend to run their hardware or software renewal cycles for 4-10 years to prolong their IT platforms, hence putting the company at risk. This is most often due to lack of time, resources or funds to renew, something a cloud service provider would never risk, and thus would likely comply to 3 year depreciations. Technical debt makes on-premise IT look cheaper but is it really?
Cost of building & operating IT vs. consuming IT
When choosing a Cloud SaaS solution i.e. Office 365, the price per user and the SLA is quite clear. Capacity and operations is included in a price per user per month. But what is the cost on building and operating on-premise IT? It depends. If calculating cost of a virtual server, everyone involved through to the hypervisor level is involved. If calculating the cost of SharePoint, everyone involved through to SharePoint including the SQL DBAs. When calculating the cost of an on-premise Office 365 solution, everyone involved from operating the datacentre through to SharePoint, Exchange, Skype for Business and other included services is involved. Most enterprises do not have resources with specialist knowledge on-duty 24×7. The cost of building and operating IT is considerable, and often not of a quality comparable to cloud vendors that have IaaS, PaaS and SaaS as their core business.
Cost of downtime
Most on-premise administrators are overworked, reactive to operational incidents, and when systems fail, they either take too long to fix it, or they hire external consultants to find the root cause, while the business suffers from downtime or bad user experience. A cloud service provider delivering services most often delivers a SLA uptime guarantee and is punished financially if they fail to deliver as promised. Therefore, the amount of focus and investment dedicated to a 99.99% uptime guarantee by cloud service providers is enormous, and not easily comparable to on-premise SLAs or “best effort” guarantees. When IT services are down, calculating the cost is difficult and individual for each company depending on the service affected, but no doubt downtime costs money – sometimes a lot of money.
“Soft” costs
The soft costs of on-premise IT are often overlooked as they are very difficult to put a price tag on, but generally these costs are often the highest.
Costs on lack of innovation
Business demands more innovation. Innovation requires the ability to try out a lot of ideas, get rapid feed-back, move forward with the best ideas and quickly kill ideas that will not work. On-premise IT often follows the traditional gated governance model. It starts with an idea, turned into a business case, then a requirement specification, a security model, a system design (with expected sizing and scale of a full-blown production platform), and then the provisioning of IT resources executed as a project in the project pipeline. The lack of dynamics and speed means that there is no room for experimentation and innovation, which in the long-term could cause a company to fall behind competition. The cost on lack of innovation is there, and it certainly isn’t cheap.
Costs of waiting
On-premise IT often takes longer time to deliver than cloud IT, because someone must build it first, then someone must provision it, make a firewall opening, ensure backup and monitoring as well as register licenses and update CMDB, which is often at a very low level of automation. The cost of employees or external consultants waiting for IT is difficult to calculate, but it is there, and the cost of people waiting for IT is also not cheap.
Costs of poor time-to-market
People talk about going digital and the importance of time-to-market. There are lots of examples to be found where a company identifies an opportunity but are then too slow to get a solution to market or alternatively they cannot release new solutions fast enough. Losing months in the provisioning of on-premise IT resources or tedious manual/semi-automated test and release management processes could cost valuable time in the harsh competition for customers. The cost of time-to-market being too long could be fatal for a company and in turn becomes very expensive when competitors move first.
Cost of security
Most enterprises have a security function, responsible for the protection of the company’s data and assets. The security team is responsible for overseeing complex solutions with limited time to do so, and often have limited specialist knowledge of the individual components in a complex Enterprise world. Hackers with evil intentions never take vacation or time off. Thus, all enterprises are fighting an invisible army with unlimited resources and funding, who attempts to destroy or steal corporate data through penetration of system security holes. Ensuring a 24×7 security protection team with thousands of people, like most cloud service providers have, is very costly and often neglected due to “we’ve never been hacked” attitudes. There are numerous examples of companies that have lost everything due to lack of security awareness and level of protection. Thinking on-premise internet-connect IT is “secure” is cheap, obtaining GDPR, ISO27001, ISO27002, PCI, SOC1/2/3 compliance like most cloud service providers have, is costly. If you think investing in security is expensive, you should see the cost of not investing in security.
Cloud Economics
Cloud economics is the discipline of minimizing the cost of consuming cloud services. Because cloud is often payed for by actual and active consumption, comparing on-premise cost to in-cloud costs, a 1:1 comparison doesn’t make sense.
Rightsizing
Rightsizing is the discipline of provisioning IT resources based on the required performance and capacity typical average on-premise sizing is 40% above required resource requirements.
Because cloud is billed based on resource consumption, continuously monitoring resource requirements to lower consumption is required. By scanning and resizing resources, based on the last 30 days peak consumption of CPU, Memory and Disk IO, you reduce cost and maintain smooth operations.
Utilization
Utilization optimization is the discipline of ensuring that IT resources are shutdown when not in use. Typically, less than 40% of on-premise IT is production (the rest is Development, Test, QA and Staging), and only 50% of production requires 24×7 operations. Because Cloud is billed based on resource consumption by the minute, shutting down resources when not in use is required. By configuring automation for shutting down resources when not used, savings are significant.
Bring your own license
Some cloud providers allow for you to bring your own licenses to the cloud.
License optimization is the discipline of ensuring the optimal usage of existing licensing agreements.
I.e. companies with Microsoft Enterprise agreements and usage rights for Windows, can save up to 43%, depending on size, when moving servers into the Azure cloud. Furthermore, cloud providers have special licensing for development and test systems further reducing cost. By evaluating existing licensing models, and applying best practice usage of hybrid use benefits and reduces cost.
Reserved Instances
Cloud Reserved Instances (RI) provide a significant discount (up to 72%) compared to On-Demand pricing and provide a capacity reservation when used in a specific availability zone.
Reservation is a commitment for the specific VM and zone for a period of 1-3 years. This is typically used for production workloads, which you know will be running for a specified number of months or years. When combining bringing own licenses with reserved instances, a saving of 82% compared to on-demand pricing is possible.
Source
Dynamic Scale
When provisioning resources in PaaS services, the true dynamics of cloud can be used to your advantage. Let’s say you create a web-application that is used for a retail business who’s opening hours are every day from 8 am to 8 pm.
You could scale-up and create two large instance Web Applications with load balancing and high-availability like you would on-premise, or you create one-small instance web-application that automatically scales out, and spins up new instances when load reaches 80% and scale down when load falls below 20%. With true scale-out dynamics, you utilize the benefits of Cloud IT, and only pay for what you need, when you need it.
The value of true dynamic scale also minimizes cost for pre-production environments, and test-environments, that doesn’t have to consume the same resources as production, because of the built-in ability to scale dynamically.
Transformation Optimization
Transformation of on-premise servers (IaaS), to in-cloud solutions based on PaaS, reduces operational cost and technical debt. Most businesses don’t need another server to patch, monitor, secure, upgrade etc., but rather need a database, a web-application or data warehouse.
By evaluating candidates for moving from infrastructure servers to platform services or software services typically reduces cost by 30%
Conclusion
Determining the true cost of an on-premise IT structure, and avoiding comparing apples to oranges when comparing to Cloud (because of the difference in quality, uptime-guarantees, security, flexibility etc.) is extremely complicated, and it takes a trained eye to calculate all the different aspects that constitutes the true cost of on-premise IT, while at the same time knowing everything about cloud economics to utilize cloud the way it was meant, to keep costs down.