1 / 43

vSphere vs. Hyper-V

Metron. vSphere vs. Hyper-V. Performance Showdown. Objectives. Architecture Available metrics Challenges in virtual environments Test environment and methods Results, conclusions, and caveats. vSphere Architecture. Small hypervisor footprint. Hyper-V Architecture.

varuna
Download Presentation

vSphere vs. Hyper-V

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Metron vSphere vs. Hyper-V Performance Showdown

  2. Objectives • Architecture • Available metrics • Challenges in virtual environments • Test environment and methods • Results, conclusions, and caveats

  3. vSphere Architecture Small hypervisor footprint

  4. Hyper-V Architecture • Windows OS required - Larger footprint

  5. vSphere Memory Management Features Transparent page sharing Memory borrowing Memory compression …and Paging

  6. Hyper-V Memory Management Feature Dynamic memory for enlightened Windows VMs

  7. Key Performance Metrics vSphere CPU Avg. CPU Usage in MHz CPU Ready Time Memory Avg. Memory Usage in KB Balloon KB Swap Used KB Consumed Active I/O Queue Latency Kernel Latency Device Latency Hyper-V CPU Hypervisor Logical Processor % Hypervisor Virtual Processor % Memory Dynamic Memory Balancer Dynamic Memory Pressure VM Vid Partition I/O Virtual Storage Device Virtual Network Adapter

  8. Challenges in Virtual Environments - Clock skew - NUMA (non-uniform memory access) scheduling - Pass-through I/O - Workload definition - 32 versus 64 bit OS and applications - Where to use solid state I/O devices - Storage tiers - Hidden overhead - etc...

  9. Test Environment • AMD Phenom II 3.3 GHz • 8 GB RAM • 1TB Hitachi 7200 RPM HD SATA 2 interface • 1GB Onboard network interface • vSphere 5 • Hyper-V role installed on Windows 2008 R2 SP1 • 2 x Windows 7 SP1 VM with integration services • 2 x CentOS 6.2 VM with integration services v3.2 • Simple custom benchmarks using ActiveState perl v5.14 • cpu.pl, disk.pl, mem.pl, net.pl

  10. Testing Methods Individual Tests - CPU w/1 process - CPU w/2 process - Disk on 1vCPU VMs - Mem on 1vCPU VMs - Net on 1vCPU VMs Combined TestsAll VMs running CPU and Mem. 2 vCPU VMs running Disk and Net.Each set of testing was run identically on the same host using both hypervisors Hyper-V Hostname Hyper30 Virtual Machines win_25, win_26 centos_11, centos_12 Win7 VMs configured with 512MB min, 64GB max dynamic memory. Centos VMs 2GB. VMs _25, _12 configured with 2 vCPU* No Pass-Through vSphere Hostname 192.168.0.99

  11. Results - Individual VM CPU Surprisingly Win7 on Hyper-V appears slower

  12. Results - Individual VM CPU Extra vCPU no help with one process

  13. Results - Individual VM CPU Twice the work in the same time

  14. Results - Individual VM CPU Summary of CPU results - Windows CPU performance on Hyper-V was significantly slower - Two vCPUs running a single process had little negative impact

  15. Results - Individual VM Disk Write a 512 MB file

  16. Results - Individual VM Disk Read the 512 MB file

  17. Results - Individual VM Disk Abysmal performance for Windows on Hyper-V

  18. Results - Individual VM Disk Summary of disk I/O results - Random I/O on a Hyper-V dynamic disk had terrible performance

  19. Results - Individual VM Memory No penalty for dynamic memory

  20. Results - Individual VM Network Slight advantage for vSphere

  21. Results - Combined Test And now… for the grand finale All workloads running at the same time on multiple VMs The winner is….

  22. Results - Combined Test Draw - except disk workload on Hyper-V did not finish

  23. CPU Details - Both platforms on the same chart

  24. Detail CPU Metrics • vSphere VM ready time • Hyper-V Guest run time

  25. CPU Details - vSphere CPU Ready Time

  26. CPU Details - Hyper-V Guest Run Time

  27. Detail Memory Metrics • vSphere memory consumed by VMs • vSphere memory ballooning • vSphere paging • Hyper-V memory balancer average pressure • Hyper-V memory current pressure • Hyper-V physical page allocation

  28. Memory Details - vSphere Ballooning

  29. Memory Details - vSphere Paging

  30. Memory Details - vSphere Consumed

  31. Memory Details - Hyper-V Memory Balancer Pressure

  32. Memory Details - Hyper-V Memory Current Pressure

  33. Memory Details - Hyper-V Pages Allocated

  34. Detail I/O Metrics • vSphere queue latency • vSphere device latency • Hyper-V disk throughput

  35. I/O Details - vSphere Queue Latency

  36. I/O Details - vSphere Device Latency

  37. I/O Details - Hyper-V Disk Throughput

  38. Interesting Observations • Hyper-V dynamic memory • Hyper-V dynamic disk device performance

  39. Dynamic Memory in action - High Pressure

  40. Dynamic Memory in action - Low Pressure

  41. Hyper-V - Random I/O on Dynamic Disks Chart from Hyper-V MSDN Blog by: Tony Voellm

  42. Conclusions, Caveats, and Final Thoughts Overall the combined results were surprising close Individual tests produced some interesting findings - Windows CPU performance on Hyper-V was significantly slower- Two vCPUsrunning a single process had little negative impact- Random I/O on a Hyper-V dynamic disk had terrible performance- Hyper-V dynamic memory worked great with no performance penalty Caveats- Workloads were very general and dependent on perl implementation- Many more variables could be taken into account- Result may be different on other hardware Running benchmarks in your own environment should be done to help make the best informed decisions.

  43. Thank you for attending vSphere vs. Hyper-V Charles Johnson Metron-Athene Inc. Charles@metron-athene.com

More Related