PHDays Fest · May 22, 2025
Reliability and Security: Foundation or Optional?
Building systems where reliability and security aren't an afterthought
/ Reliability & Security · PHDays Fest 2025
Building systems where reliability and security aren't an afterthought
Building systems where reliability and security aren't an afterthought
What's at stake: downtime, breaches, trust
Emergent properties of systems and processes
ATAM, risk, and threats as architecture process
Google/Netflix/AWS cases, Spirit, takeaways
What's at stake when we talk reliability and security
Both hit revenue, reputation, and compliance
Hyper-connected: cloud, microservices, IoT
Emergent behavior: a bug cascades
Rapid change: hundreds of deploys a day
Threats: more frequent and sophisticated
Reliability and security are whole-system properties, not features
From process and code to system and product
Not plug-and-play — not a product
Whole-system: code, infra, process, people
Cross-cutting: spans every layer
Foundational, not 'bolt it on later'
Redundancy and graceful degradation
Circuit breaker, bulkhead, backpressure
SLOs for uptime/latency from the start
Chaos and load tests before release
Shift left: the earlier you find it, the cheaper
Formalizing analysis: ATAM, risk, threats
Quality attributes → analysis → risks and sensitivity points
Risk Matrix and FMEA turn worries into data
STRIDE, PASTA, DREAD, MITRE ATT&CK, and more
Google, Netflix, AWS — how the leaders do it
Benjamin Treynor Sloss, Google
SRE and SLOs/error budgets by default
Design reviews with reliability/security
Production Readiness Review before launch
Blameless postmortems, learning culture
Chaos Monkey kills instances in prod
Designs assuming things will break
Team autonomy + responsibility
Redundancy, fallback, load shedding
Werner Vogels, Amazon CTO
Security before every other priority
Encryption and strict IAM by default
No launch with a known security issue
Well-Architected: security + reliability
Security is everyone's responsibility
Reliability is a feature (SLOs, budgets)
Design for failure: assume it breaks
Leadership and culture set the tone
continuous improvement is a journey, not a one-time project
The Spirit platform: PaaS, observability, incidents, DevEx
One portal, the whole service lifecycle underneath
A unified search engine, alerting, and UI over the signals
From detection to postmortem and releases
Nestor — AI assistant in IDE
Safeliner flags vulnerabilities in code
Explanation + suggested fix
Security review inside the dev flow
Feedback loops, cognitive load, flow state
Satisfaction, Performance, Activity, Communication, Efficiency
Strategy, culture, architecture, DevSecOps
Build
Reliability/security in OKRs
Blameless security champions
ATAM, threat modeling, DevSecOps
Avoid
Security later
No incidents means fine
InfoSec owns it
AI ops helps, adds risk
Zero trust as a security foundation
Serverless/containers reshape resilience
Regulators mandate 'secure by design'
What to remember and where to start on Monday
Reliability and security are built-in
Mix architecture + risk + culture
Learn from leaders, fit your context
It's a journey: threats and complexity grow
building securely and reliably lets the business move faster
Assess — evaluate your current state
Quick wins — threat model + chaos
Set targets — SLOs for critical services
Plan — 12–18 month roadmap
polomodov.tech
All slides and links — in the Telegram channel
Alexander Polomodov, Technical Director & Fellow, T-Technologies
@book_cube