Plan Your IaaS Environment for Optimal Performance
Download as PPTX, PDF•0 likes•660 views
The document outlines a webinar focused on optimizing IaaS environments for performance, highlighting significant common performance issues such as disk latency, memory oversubscription, and packet loss in cloud setups. It emphasizes the importance of establishing a performance baseline and provides strategies for data gathering, benchmarking, and proactive management to improve cloud infrastructure performance. Additionally, it mentions upcoming webinars and RISC Networks' services aimed at enhancing IT analytics and operational efficiency.
Plan Your IaaS Environment for Optimal Performance
- 1. Cloud Webinar Series Plan your IaaS Environment for Optimal Performance January 15, 2013 Speaker: Jeremy Littlejohn, Founder and Chief Analyst RISC Networks
- 2. Agenda 1. 2. 3. 4. Understand common performance issues How to establish a performance baseline Keys to your success Our Analysis
- 4. Cloud is Still Virtualization – Same Rules Apply • • • • Disk Latency Hypervisor Memory Oversubscription Packet-loss on the network Hypervisor CPU Oversubscription
- 5. Disk Latency 55% of organizations have storage latency of over 50ms
- 6. Hypervisor Memory Oversubscription 53% Organizations that have VMware Servers with performance degradation due to Memory Swap In / Swap Out. Learn More: Performance Troubleshooting for VMware vSphere
- 7. Packet Loss 64% 26% 64% of organizations have receive packet loss occurring on virtual switch ports and 26% have transmit packet loss occurring on virtual switch ports
- 8. CPU Oversubscription 33% Of organizations have at least 1 VMware Hypervisor with CPU Oversubscription Learn More: VMware recommends less than 5% ready time for all VMs’
- 9. Here is the Kicker! All of these challenges exist in cloud environments, both Public and Private. The question is, how can you build a baseline that helps you prepare to deal with them?
- 11. Infrastructure Performance Statistics 10 In order to establish an effective baseline, you have to focus on more than ONE dimension. 9 8 7 6 5 4 3 2 1 Infrastructure Performance Statistics CPU Usage
- 12. Cross-Functional Analysis Data Center 54% Vmware Perf Problems 33 % Oversubscribed Line Cards 47 % Storage Perf Problems 73 % Network Perf Problems 1. 2. 3. 4. 5. Don’t Focus on CPU usage Analyze Disk I/O Analyze Network I/O Capture Network Traffic Look at the network relations and dependency
- 14. Collect (at least) the following data • CPU • Usage % • % Steal time OR % Ready Time • Memory • Swap In/Out Rate (hypervisor if virtual) • CPU % Wait Time (can show disk slowness for swapping) • VM Swap Wait • Disk • IOPS (I/O Per Second) • I/O Block Size • Latency and Queue Length • Network • Packet Loss % • Delay
- 15. 140 120 100 80 60 40 20 0 Max IOPS 1/10/2014… 1/9/2014… 1/9/2014… 1/9/2014… 1/9/2014… 1/9/2014… 1/9/2014… 1/9/2014 8:44 1/9/2014 6:20 1/9/2014 3:56 1/9/2014 1:32 1/8/2014… 1/8/2014… 1/8/2014… 1/8/2014… 1/8/2014… 1/8/2014… 1/8/2014 9:55 1/8/2014 7:34 1/8/2014 5:39 1/8/2014 3:16 1/8/2014 0:55 1/7/2014… 1/7/2014… 1/7/2014… 1/7/2014… 1/7/2014… 1/7/2014… 1/7/2014 9:21 1/7/2014 7:00 1/7/2014 4:36 1/7/2014 2:13 1/6/2014… 1/6/2014… 1/6/2014… 1/6/2014… 1/6/2014… A quick example – Planning for Disk Performance in Amazon AWS • Amazon EBS (Basic) Supports 100 IOPS (@16KB) • Step 1: Collect I/O statistics from your vCenter • Sum the Reads/Writes/Sec for each hour Max IOPS
- 16. A quick example – Planning for Disk Performance in Amazon AWS • Step 2: If you are < 100 IOPS directly measured, also determine the Write / Read Size to see if you need to adjust to Provisioned IOPS (KB per Write/16) *Writes/Sec<100 = OK (KB per Write/16)*Writes/Sec>100 = Need Provisioned IOPS
- 17. Benchmark your Disk Latency • If you have latency > 50ms, you are already having problems • If you are under 50ms, then you want to be sure that you don’t introduce latency to the applications as you move them. • You can track queue length on your current system, and then compare that to AWS EBS VolumeQueueLength metric to verify performance
- 18. Keys to Success
- 19. Keys to your success 1. Gather Data 2. Benchmarking 3. Best Practices 4. Proactive Management
- 20. Gather Data • General usage info for CPU, Memory, Disk • Current usage and impact on performance • Current network performance information • Virtualized environment analysis
- 21. Benchmarking • Clearly identify specific areas of opportunity • Set performance expectations • Validate infrastructure assumption • Monitor company performance and manage change
- 22. Best Practices - Amazon AWS Performance Information - http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/EBSPerformance.html - Performance and Troubleshooting Guide - http://www.vmware.com/resources/techresources/10179 - Network Performance Statistics Information - http://www.slac.stanford.edu/comp/net/wanmon/thru-vs-loss.html
- 23. Proactive Management • Compare statistics to earlier benchmarks • Work on Lead, not Lag metrics • Test routinely – Traffic Simulation, etc • Capture Traffic so that you can validate Network Performance
- 24. Our Analysis
- 25. 5 Our Analysis Simulate and analyze cloud infrastructure traffic 2 Cloud Provider Collect and aggregate component-level data using secure data encryption. 1 SaaS delivered analysis and reporting portal for anywhere engagement management (*) Physical Appliance may be required during network traffic analysis, depending on the required infrastructure configuration 3 RISC Network’s only listens during infrastructure analysis and during network/cloud traffic analysis. Only during network traffic simulation is there active data being put onto the network 4 Virtual Appliance* Securely Installed in as little as 15 minutes
- 26. Our Analysis by the #’s 97,000 Virtual Machines VMware Virtual Machines Analyze
- 27. Our Analysis by the #’s 4,000+ Business Networks Analyzed
- 28. Our Offering
- 29. Cloud Readiness Analytics • • • • • • Software-as-a-Service (SaaS) Pricing Model Analysis Network Performance Analysis Move Group Planning and Analysis Server & App Dependency Visualization Cloud Performance Testing and Validation To learn more about our analytics go to: http://www.riscnetworks.com/cloudscape
- 30. Upcoming Webinars • Subscribe to our BrightTALK™ Channel: https://www.brighttalk.com/channel/10539 • • • • February 5th – Why Network Testing is Critical for Cloud Success February 19th – Top 5 Network Infrastructure Performance Issues March 5th – Be Confident That Your Cloud Solution is Working Well March 12th – Performance Benchmarking for Private and Hybrid Cloud
- 31. Thank you
- 32. About Us At RISC Networks we imagine a world where IT leaders are not constrained by the complexities of their IT tools. We have a perspective, a different way of thinking about IT, a simpler way! We believe you should have more meaningful data analysis; information that actually helps guide you in solving your problems, tools that help you research and take immediate action. IT – As simple as it should be™ Based in Asheville, North Carolina, RISC Networks is a Actionable IT Analytics firm that focuses on developing research and technology that helps IT organizations improve their IT operations. More information is available at: http://www.riscnetworks.com/. Follow RISC Networks on Twitter @Risc_Networks or on Facebook at http://www.facebook.com/RiscNetworks. 60+ Partners Including (but not limited to):