Infrastructure Considerations Flashcards

Availability is one of the CIA triad pillars: systems must be accessible when authorized users need them. The security nuance: availability must be paired with access control — systems should be available to authorized users only. A system accessible to everyone (including attackers) is not a security success; a system offline to legitimate users (due to DDoS) is an availability attack. Availability is expressed as an uptime percentage: 99.999% ("five nines") = ~5 minutes maximum downtime per year. Organizations invest in monitoring, redundant systems, and failover mechanisms to maximize uptime — while ensuring access is restricted.

The primary resilience metric: the average time from when a failure is detected to when the system is fully restored to operational status. Lower MTTR = better resilience = less downtime. Factors that drive MTTR: (1) Root cause identification time; (2) Hardware replacement / software patch availability; (3) Redundant system readiness. Related metrics: MTTF (Mean Time to Failure) = how long before the next failure; MTBF (Mean Time Between Failures) = system reliability. Formula: MTBF = MTTF + MTTR. A high MTBF + low MTTR = highly resilient system. MTTR is reduced through pre-positioned spares, pre-built images, documented runbooks, and trained response teams.

Scalability: the ability to increase or decrease capacity to match changing demand. A scalable system can handle 1,000 or 50,000 users without redesign. Elasticity: the cloud-specific capability for automatic, real-time scaling — expanding and contracting multiple times per day in direct response to demand, without manual intervention. Why not always provision for peak? Because idle capacity costs money — elasticity matches cost to actual usage. Security requirement: security monitoring must scale with the infrastructure. Auto-scaling that provisions new compute instances without deploying SIEM agents and endpoint monitoring creates unmonitored attack surfaces.

The time between a user submitting a request and receiving a complete response. Critical for interactive applications — humans are sensitive to delays above ~100–200ms. Responsiveness is a composite of all components in the request path: network latency + application processing + database query time + caching. The weakest link determines overall responsiveness — a fast web server with a slow database query still delivers a slow response. Security appliances (WAF, IPS) that inspect every request add processing latency that degrades responsiveness. End-to-end measurement is required; optimizing one fast component while a bottleneck exists elsewhere has no user-visible benefit.

Modern applications have many components (web server, database, caching, load balancer, firewall, monitoring agents) — deploying them manually is complex and error-prone. Orchestration defines the entire deployment as code (Terraform, CloudFormation, Ansible) that provisions all components in the correct order, with the correct configuration, automatically and repeatably. Security benefit: manual deployment steps that are forgotten (firewall rules, monitoring agents, certificates) become vulnerabilities. Orchestrated deployments execute the same configuration every time — security settings are peer-reviewed, version-controlled, and identical across development, staging, and production. Deployment decisions must be addressed during the engineering phase, not improvised at deployment time.

Risk transference shifts the financial consequences of a risk to a third party — accepting that the risk may occur, but ensuring financial impact is partially absorbed by someone else. Cybersecurity insurance covers: business interruption losses, incident response costs, ransom payments (in some policies), legal defense and settlements, and breach notification costs. Grown significantly with the rise of ransomware. Critical distinction: insurance is not a security control — it does not prevent incidents or reduce their probability. It only transfers financial consequences. It complements security controls; it does not replace them. Insurers now require minimum security controls (MFA, EDR, backups) before issuing policies.

The speed and simplicity of restoring systems after an incident — a design criterion that must be planned before failures occur. The classic comparison: reloading from original installation media (~90 minutes per system) vs. deploying from a pre-built corporate image (~15 minutes per system). For 50 systems: 75 hours vs. 12.5 hours — a significant operational cost difference. Recovery procedures must be designed during the engineering phase, not improvised during a crisis. Ease of recovery directly determines the financial and operational impact of any incident: time offline = money lost. Modern approaches: corporate images, cloud snapshots, containerized restarts (seconds).

The organized process of identifying, testing, and deploying software updates to fix security vulnerabilities and bugs. The first task after deploying a new system: check and apply all patches released since the installation media was created. Standard cycle: vendor releases patch → evaluate severity → test in non-production → deploy to production. Microsoft releases patches on the second Tuesday of each month ("Patch Tuesday"). Test before production — a patch that breaks a critical application is itself an availability incident. Speed vs. thoroughness tradeoff: critical zero-day vulnerabilities require accelerated emergency patching where the risk of delay exceeds the risk of abbreviated testing.

Some systems cannot be updated through standard patch management: embedded systems (HVAC controllers, time clocks, badge panels), FDA-regulated medical devices, legacy industrial equipment, and end-of-support software. These systems have known vulnerabilities that will never be fixed. Compensating controls applied at the network boundary (not the device): (1) Network isolation on a dedicated VLAN — the vulnerability cannot be exploited if there is no network path; (2) Firewall permitting only required traffic; (3) Physical access control; (4) Network-level monitoring for exploitation attempts. The long-term solution is device replacement; compensating controls are the interim mitigation.

Power is the invisible foundation of all IT infrastructure — without it, availability = 0%. Two backup power layers work together: UPS (Uninterruptible Power Supply): battery-backed device that provides immediate power when utility power fails — bridges the gap from outage to generator start (seconds to minutes). Also conditions power quality. Generator: diesel or natural gas backup power for extended outages — fuel-dependent duration. Together: UPS provides instant power → generator starts (10–30 seconds) → generator takes over → UPS recharges. Without UPS, servers lose power during the generator start window. Data centers also use dual utility feeds for redundancy at the source.