Cloud
Cloud Computing
What is the cloud?
- Rent servers and storage instead of owning them
- Use and pay only for what you need
- IT infrastructure basics
What the cloud really changes is less the technology than how you spend money and how fast you can make decisions. Buying a server means filing an asset purchase request worth tens of millions of won, followed by depreciation, server room space, and maintenance contracts. In the cloud, the same resources become a single monthly bill, so capital expenditure (CAPEX) shifts to operating expense (OPEX). What used to be a purchase order that took months becomes a console task that takes minutes, and since a failed experiment ends when you return the resources, the cost of experimenting itself drops.
Breaking it into layers makes the boundaries of responsibility clear. With IaaS, you rent only servers and networking and handle everything up to OS patching. With PaaS, you hand off the runtime environment and just upload code. With SaaS, you get a finished product through an account. The higher up you go, the easier it gets, but the less you can control. The key point not to miss is the shared responsibility model: the provider secures 'the cloud itself,' while the customer is responsible for 'security in the cloud,' such as account permissions, data, and configuration.
Deployment is not one-size-fits-all either. Instead of putting everything on the public cloud, many Korean manufacturers and financial firms use a hybrid setup: payment and personal-data systems stay on-premises, and only the front-end web tier runs in the cloud. The region (data center zone) you choose affects both response speed and data residency requirements. Pricing also splits into three models: on-demand, billed by the hour; reserved, which lowers the unit price with a 1–3 year commitment; and spot, which is cheap but can be reclaimed at any time. Even for the same specs, the unit price varies widely by contract type, so you first need to decide which workloads go on which pricing plan.
The belief that breaks most often in practice is 'moving to the cloud makes it cheaper.' Copying your existing server setup as-is and running it 24/7 actually costs more. In particular, outbound data transfer fees and dev/test servers that are never shut off inflate the bill. Without caps on auto-scaling or budget alerts, charges can multiply over a single weekend. One more thing: backups and access management are still the customer’s job, so leaks happen when access keys get pushed to GitHub or storage is left open to the public. For regulated workloads, check the region and security certification before you plan the migration.
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”The concept is old, but it went mainstream in 2006 when Amazon commercialized cloud services such as EC2 and S3 (AWS). It has since become core infrastructure for digital transformation (DX).
Consider a hypothetical case. Company A, a pet supplies online store, has 40 employees and annual revenue of about KRW 12 billion (≈ US$8,978,000). It ran six servers in its own data center, and during every holiday and major sale, checkout backed up within 30 minutes of the start and orders stopped coming in. When the orders lost to outages in one year topped KRW 100 (≈ US$1) million, management decided to complete a cloud migration within three months.
- When the CTO opened a year of outage logs, 78% of downtime was concentrated in the first 30 minutes of each event. Adding servers took six weeks from order to delivery, so the hardware arrived after the event every time.
- Infrastructure manager Kim first moved just two dev/test servers, ran them for four weeks, and reported back with a single table comparing the cloud’s monthly cost of KRW 870,000 (≈ US$650) against the data center’s rent and electricity of KRW 2.1 million (≈ US$1,570).
- The order database was migrated at 2 a.m., the lowest-traffic hour, which held actual service downtime to 3 hours 40 minutes. Two days earlier, the support center received advance-notice wording and a refund-handling script.
- Manager Kim set up an architecture that automatically scales web servers from 4 to 12 with traffic and drops them to 2 between 2 and 6 a.m., cutting the staff needed for manual response on event day from 3 to 1.
- The CFO said, "Not knowing what this month’s bill will be is what worries me most," so we set a KRW 15 million (≈ US$11,200) monthly budget alert and department-level tag billing, so each team’s name appears on the invoice.
- At that year’s Chuseok (Korean harvest festival) event, concurrent users spiked to 3.2 times normal, yet there were zero outages. Monthly infrastructure costs fell from KRW 23 million (≈ US$17,200) to KRW 14.5 (≈ US$1) million, and new feature releases dropped from six weeks to two days.