Hope Is Not A Strategy: Automating Efficient Resource Utilization for SREs
Download as PPTX, PDF0 likes227 views
The document discusses the importance of efficient resource utilization in Site Reliability Engineering (SRE) and presents a solution to combat challenges such as developer burnout and cloud waste. It highlights the role of automation and machine learning in optimizing application performance, resource allocation, and cost-effectiveness. The StormForge platform is introduced as a tool for continuous optimization, showcasing its ability to improve resource utilization without performance degradation.
Hope Is Not A Strategy: Automating Efficient Resource Utilization for SREs
- 1. Hope is Not a Strategy June 3, 2021 Automating efficient resource utilization for SREs Welcome! We’ll get started in a few minutes...
- 2. Today’s StormForge presenters ● Product mktg professional, former developer ● 30 years industry experience, including Dynatrace, Altiris, EDS ● Based in Ann Arbor, MI RICH BENTLEY Sr. Director, Product Marketing ● Sales Engineer, former IT leader ● 24 years experience: Finserv, Biotech, Juniper Networks, Tintri ● Based in the Bay Area, CA ERWIN DARIA Principal Sales Engineer
- 3. What does SRE stand for? Simply Restart Everything A Senior Rebooting Engineer B Sleep Rarely Ever C Software Ruins Evening D Seriously Regretting Everything E
- 4. Your job is incredibly difficult Pressure Stakes are high Uncharted waters Massive complexity
- 5. Cloud efficiency trade-offs Application Performance Time & Effort Cost/Resource Utilization
- 6. Achieving peak efficiency means... Application Performance Time & Effort Cost/Resource Utilization ...improving SLI performance ↑ Throughput ↑ Availability ↓ Latency ↓ Error rate To consistently meet or exceed SLOs... ...at the lowest possible cost... ...with the least amount of effort.
- 7. What is the impact of poor performance for your business? Revenues War rooms Customer sat Customer retention Conversions Slowed migration/ transformation Brand/reputation Delayed initiatives Productivity Loss: dev, ops, business users Abandonment
- 8. Nearly half of cloud spend is wasted Source: StormForge Cloud Waste Survey https://www.stormforge.io/survey- report/cloud-waste-survey-findings/
- 9. Developer burnout is a big problem Source: D2IQ, Kubernetes in the Enterprise: Uncovering Challenges & Opportunities 38% of developers and architects claim their work makes them feel extremely burnt out 51% of developers and architects say building cloud native applications makes them want to find a new job
- 10. Why is it so hard? 10 APPLICATION PERFORMANCE COST/RESOURCE UTILIZATION AVAILABILITY Container 1 CPU Memory Requests Limits Requests Limits Replicas Container 2 CPU Memory Requests Limits Requests Limits Replicas Application Settings JVM Heap size Garbage collection Container 3 CPU Memory Requests Limits Requests Limits Replicas Shuffle file buffer Reducer max size
- 11. What Tools Are You Using Today? 11 TRIAL-AND-ERROR PERFORMANCE & LOAD TESTING K8S AUTO-SCALING (HPA) MONITORING/ OBSERVABILITY TOOLS Easiest way to get started! Great way to waste engineering time K8s expertise is hard to find! Important step for validating non-functional requirements Helps to understand scaling behavior of app Doesn’t tell you how to address issues Dynamically scales # of pods Still requires configuration to optimize Doesn’t account for app- specific configuration settings Good for identifying and troubleshooting problems in production Reactive by nature - by the time you find a problem it’s already impacted your users
- 12. 12 So, how do we get there?
- 13. Automating efficient resource utilization 13 SRE approach Applying SW engineering principles to optimization Automated Leverage machine learning Empowered dev teams Focused on innovation Managed risk Understand the trade-offs TO THIS: FROM THIS: Sysadmin approach Operating and maintaining Manual Resource-intensive manual trial-and-error Dev teams on their own Stuck in the weeds of K8s Blind risk No sense of trade-offs Proactive Build efficiency into the release process Reactive Running a service & responding to events
- 14. “My goal is to automate myself into my next job.”
- 15. SRE = Software Ruins Evening? 15
- 17. Shift Optimization Left Build continuous optimization into your CI/CD pipeline 17
- 18. The StormForge platform APPLICATION OPTIMIZATION PERFORMANCE TESTING Create tests quickly with performance testing as a service Automate testing as code in CI/CD workflow Execute realistic tests with open workload model ML-powered Rapid Experimentation Engine No upfront training data required Automatically implement optimal config based on your goals Identify high-risk configs to avoid
- 19. APPLICATION PARAMETERS CPU, MEMORY, REPLICAS OUTCOMES LATENCY, THROUGHPUT, COST, RELIABILITY FEEDBACK LOOP CONTINUOUS OPTIMIZATION Test Case Business Objectives Your App in Your Environment Rapid Experimentation Engine Visualized Business Objective Trade-offs Optimal Configuration How does it work?
- 20. Product Demo
- 21. Large travel website achieves 50% cost-efficiency improvement with StormForge Initial config after extensive manual work Selected config reduced cost by 50% w. no impact on performance ● Team of 9 focused on manual optimization prior to StormForge ● Minimal dev and test environment due to private cloud resource constraints ● 50% resource utilization improvement with no performance degradation ● Free up private cloud resources for additional QA environments
- 22. See the cloud waste survey results: https://www.stormforge.io/cloud-waste-report Schedule a Demo: www.stormforge.io/request-a-demo/ Sign Up for Free: www.stormforge.io/get-started/ info@stormforge.io @StormForgeIO Next Steps Event Feedback? events@stormforge.io Find all these assets and more at www.stormforge.io/next-steps/