SaaS · PaaS · IaaS
Software as a Service / Platform as a Service / Infrastructure as a Service
What is SaaS?
- 3 cloud layers
- Divided by “how much someone else manages”
- The pizza analogy
The distinction between SaaS, PaaS, and IaaS isn't a technical classification chart. It's the dividing line for responsibility and billing. This line decides who has to get up when a server goes down at 3 a.m., who applies security patches, and who you call when an outage hits. If you choose IaaS, everything from OS patches to backup verification falls to your own infrastructure staff. If you choose SaaS, all you can do is watch the vendor's status page and wait. In other words, this choice settles 'how many people you need to assign' before it settles 'how much you pay.'
In practice, three criteria guide the decision. First, does the logic embody what sets your business apart? If it is your competitive edge, like a quote-calculation algorithm, build it yourself on PaaS or IaaS. Routine work that looks much the same everywhere, like payroll or electronic approval, is better handed to SaaS. Second, staffing. Without two or three people to run infrastructure full time, IaaS tends to leave abandoned servers piling up about a year after launch. Third, regulation. In finance and healthcare, network separation and domestic data-storage requirements still rule out SaaS in some cases. For reference, in Korea’s public sector, CSAP certification is effectively the first gate.
The most common misconception is the belief that 'SaaS needs no management.' Account management, permission design, revoking accounts of departed employees, and data backup are still the customer's job. Companies often get burned after canceling a SaaS contract because they can't get their data back, simply because they never checked the contract's data return clause and export formats. Another common situation is departments each paying for their own SaaS and then losing millions of won a month on accounts nobody uses. Cloud is less a cost that 'shrinks when you use less' than a cost that 'grows when you stop watching.'
Seungtae ParkHands-on AX instructor who teaches AI business development and marketing automation by building real systems
John KimInstructor John Kim, CEO of Nanum Consulting and author of “B2B Marketing Design”| Category | IaaS | PaaS | SaaS |
|---|---|---|---|
| What the provider covers | Infrastructure (servers, storage) | Up to the development environment | Finished software |
| Management burden | High | Medium | Almost none |
| Analogy | Buy ingredients and bake | Heat up a ready meal | Ordering delivery |
| Yes | Cloud server | App deployment platform | Gmail, Slack |
A classification of cloud computing service models that became standard as clouds like AWS and Azure went mainstream.
Consider a hypothetical case: Company A, an auto parts manufacturer with 320 employees and annual revenue of around KRW 90 billion (≈ US$67,335,000). It had run its ERP and production management systems on 14 servers in its own server room. When the quote to replace the aging servers came in at over KRW 400 (≈ US$1) million, it began considering a move to the cloud. But its IT team had only 3 people, including the team lead.
- The IT team lead first proposed moving everything to IaaS and continuing to manage it in-house. The finance director rejected it, asking, “Are you saying the three people holding this together with overtime now will keep doing that in the cloud?”
- In the end, they sorted the 11 systems into three groups. Six systems, including groupware, HR and electronic approval, moved to SaaS. The MES, which receives equipment data from the production floor, is heavily customized, so it moved to IaaS as is.
- The new partner ordering portal went onto PaaS. The development partner’s CEO said, “We spent three weeks just setting up servers, and in that time we could have built ten more screens,” so the company switched to a managed database and deployment environment.
- When they audited SaaS licenses six months after the switch, 27 accounts of former employees were still being billed. They then tied account revocation to offboarding in the HR system, plugging a leak of about KRW 1.9 million (≈ US$1,420) a month.
- A year later, total infrastructure cost was down about 18%. More importantly, after-hours server-failure calls to the IT team fell from an average of 6 a month to 1, freeing two people to move to production data analysis.