Cloud Computing Technology Fundamentals
Domain scope
This domain tests the technology under the services: virtualization, storage, networking, containers, and high availability. Questions often ask why a design works, not which product name to pick.
Core rules
- A Type 1 hypervisor runs on bare metal; a Type 2 hypervisor runs on a host OS. Production clouds use Type 1 for performance and isolation.
- Compute virtualization schedules vCPUs on physical cores; overcommit raises density but risks contention. Memory virtualization uses ballooning and swap.
- Distributed storage keeps multiple replicas; with three replicas, raw capacity is roughly one-third usable before overhead.
- VXLAN encapsulates Layer 2 frames in UDP/IP to extend virtual networks across Layer 3; it supports large tenant ID spaces.
- Containers share the host kernel and isolate processes; VMs have their own guest kernel. Containers start faster but have a different isolation boundary.
- High availability reduces downtime; backup restores data after loss; disaster recovery survives a site failure. They are different objectives.
Worked example
A storage pool has 30 TB raw and uses three replicas. Usable capacity is about 10 TB before metadata and reserve overhead, so planning for 12 TB of data is unsafe. Reason: each object is written to three replicas, so every 1 TB of logical data consumes about 3 TB raw. If the design needs more usable capacity, use erasure coding where appropriate or add disks; do not assume RAID alone gives geo-redundancy.
Traps
A snapshot is not a backup if it lives in the same failure domain and depends on the same storage. A container is not a lightweight VM; it does not boot its own kernel. Overcommit is not free capacity. VXLAN is not encryption.