Snapshot
Compute, virtualisation, containers, storage, database platforms, backup, digital workplace and lifecycle architecture.
Module status
- •Track: Field 1 — ICT infrastructure
- •Last verified: 2026-08-11
- •Scope: Independent EPSOHQ training mapped to the notice duties; official sources prevail.
Compute and virtualisation
Enterprise infrastructure separates physical capacity from logical services. Virtual machines package a guest operating system above a hypervisor. Containers isolate processes while sharing the host kernel. Orchestrators schedule workloads, maintain desired state and replace failed instances. The architecture choice must reflect isolation, portability, performance, licensing and operational competence.
Exam-ready rules:
- •A type 1 hypervisor runs directly on host hardware rather than on a general-purpose host operating system.
- •A virtual machine includes a guest operating system, while containers normally share the host kernel.
- •Horizontal scaling adds service instances, while vertical scaling adds resources to an existing instance.
- •NUMA awareness matters when memory access latency differs according to the processor socket serving the memory region.
- •Live migration moves a running virtual machine between hosts while aiming to minimise service interruption.
- •CPU overcommit allocates more virtual CPU capacity than physical capacity and therefore requires contention monitoring.
- •Immutable infrastructure replaces a deployed instance from a versioned image instead of repairing configuration drift in place.
- •An availability zone is an isolated location inside a region and should not be treated as an independent geographic region.
- •Container orchestration maintains declared workload state through scheduling, health checks and replacement of failed instances.
- •High-performance computing uses parallelism and high-throughput interconnects to execute tightly coupled or data-intensive workloads.
Capacity planning begins with demand, headroom, growth and recovery requirements. Average utilisation alone hides peaks and contention. Placement policies should prevent replicas and quorum members from concentrating in one failure domain. Firmware, hypervisor and base-image lifecycles need planned maintenance and rollback.
Storage and database services
Storage design starts with workload behaviour: latency, throughput, IOPS, consistency, capacity, durability and recovery objectives. Block storage exposes addressable blocks, file storage exposes a hierarchical namespace and object storage exposes objects through an API. None is universally superior.
Exam-ready rules:
- •Block storage exposes fixed-size addressable blocks and commonly underpins filesystems and database volumes.
- •Object storage keeps objects with metadata in a flat namespace and is normally accessed through an API.
- •File storage presents shared files and directories through a hierarchical namespace.
- •RAID can improve availability or performance but does not replace an independent backup.
- •A recovery point objective limits tolerable data loss measured in time, while a recovery time objective limits tolerable restoration time.
- •A crash-consistent snapshot preserves storage state at one instant but may still require application recovery on restart.
- •Write-ahead logging records a change before the corresponding data page is written, supporting transactional recovery.
- •Synchronous replication waits for acknowledgement from another replica before confirming a write and therefore affects latency.
- •Database connection pooling reuses established connections and protects the database from unbounded connection creation.
- •A restore test proves recoverability more directly than a successful backup-job status.
Database architecture distinguishes transaction processing from analytical workloads. Indexes accelerate reads but consume storage and add write cost. Replication helps availability and read scaling but creates consistency and failover decisions. Retention and deletion must propagate into backups according to policy and technical feasibility.
Digital workplace and application platforms
Workplace services combine device management, collaboration, identity, software distribution and support. Standard builds reduce variance. Conditional access uses signals such as identity, device posture, risk and location to apply policy. Application platforms need supported runtimes, dependency inventory, secrets handling, logging and patch ownership.
Primary references:
- •NIST SP 800-125, Guide to Security for Full Virtualization Technologies — NIST
- •NIST SP 800-190, Application Container Security Guide — NIST
- •Kubernetes concepts — Kubernetes documentation
- •NIST Cloud Computing Reference Architecture — NIST SP 500-292