Editing Performance
Warning: You are not logged in. Your IP address will be publicly visible if you make any edits. If you log in or create an account, your edits will be attributed to your username, along with other benefits.
The edit can be undone.
Please check the comparison below to verify that this is what you want to do, and then save the changes below to finish undoing the edit.
Latest revision | Your text | ||
Line 1: | Line 1: | ||
− | == | + | == Response Time == |
− | + | === Benchmark Description === | |
+ | The aim of this benchmark is to measure how fast can application inside of virtual machine (VM) or operating system container (CT) react on external request under various conditions: | ||
+ | * Idle system and idle VM/CT | ||
+ | * Busy system and idle VM/CT | ||
+ | * Busy system and busy VM/CT | ||
− | + | Described benchmark case is common for many latency sensitive real life workloads. For example: high performance computing, image processing and rendering, web and database servers and so on. | |
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | |||
− | == | + | === Implementation === |
− | + | To measure response time we use well known netperf TCP_RR test. To emulate busy VM/CT we run CPU eater program (busyloop) inside of it. To emulate busy system we run several busy VM/CT (to eat all the host CPU time). Netperf runs in server mode inside of '''one''' VM/CT. On the separate physical host we run netperf TCP_RR test against selected VM/CT over the 1Gbit network. | |
− | * [ | + | |
− | * | + | === Testbed Configuration === |
+ | Server: 4xHexCore Intel Xeon (2.66 GHz), 32 GB RAM | ||
+ | |||
+ | Client: 4xHexCore Intel Xeon (2.136 GHz), 32 GB RAM | ||
+ | |||
+ | Network: 1Gbit direct server<>client connection | ||
+ | |||
+ | Virtualization Software: ESXi4.1upd1, XenServer5.6fp1, HyperV (R2), PVC 4.7 (RH6) 2.6.32-042test006.1.x86_64 | ||
+ | |||
+ | Guest OS: Centos 5.5 x86_64 | ||
+ | |||
+ | Software and Tunings: | ||
+ | * netperf v2.4.5 | ||
+ | * '''one''' VM/CT with netperf in server mode configured with 1 vCPU, 1 GB RAM | ||
+ | * '''six''' VMs/CTs (which needed to load server CPU - see testcases) configured with 4vCPU 1GB RAM | ||
+ | * netperf run string: | ||
+ | ** in VM/CT; netserver -p 30300 | ||
+ | ** on the client: netperf -p 30300 -H 172.0.1.1 -t TCP_RR -l 120 -- -r 128 -s 128 | ||
+ | * Firewall was turned off | ||
+ | * All other tunings were left at default values. | ||
+ | |||
+ | === Benchmark Results === | ||
+ | |||
+ | [[File:Response_time.png]] | ||
+ | |||
+ | |||
+ | '''In all the three cases (idle system and idle VM/CT, busy system and idle VM/CT, busy system and busy VM/CT) Virtuozzo Containers show the lowest overhead over all the tested virtualization solutions''' | ||
+ | |||
+ | |||
+ | == 10 Gbit Network Throughput == | ||
+ | |||
+ | === Benchmark Description === | ||
+ | |||
+ | In this benchmark we measure throughput over 10 Gbit network connection in two directions: | ||
+ | * from VM/CT to physical client | ||
+ | * from physical client to VM/CT | ||
+ | |||
+ | |||
+ | === Implementation === | ||
+ | |||
+ | To measure network throughput we use standard performance test '''netperf'''. Host with VM/CT and physical client connected are interconnected directly (without switches, etc.) | ||
+ | |||
+ | === Testbed Configuration === | ||
+ | Server: 4xHexCore Intel Xeon (2.66 GHz), 32 GB RAM, Intel 82598EB 10-Gigabit network card | ||
+ | |||
+ | Client: 4xHexCore Intel Xeon (2.136 GHz), 32 GB RAM, Intel 82598EB 10-Gigabit network card | ||
+ | |||
+ | Network: 10Gbit direct server<>client optical connection | ||
+ | |||
+ | Virtualization Software: ESXi4.1upd1, XenServer5.6fp1, HyperV (R2), PVC 4.7 (RH6) 2.6.32-042test006.1.x86_64 | ||
+ | |||
+ | Guest OS: Centos 5.5 x86_64 | ||
+ | |||
+ | Software and Tunings: | ||
+ | * netperf v2.4.5 | ||
+ | * '''one''' VM/CT with netperf configured with 4 vCPU, 4 GB RAM | ||
+ | * where it was possible, we set offloading & hardware checksumming (gro, gso,etc...) and jumbo frames (MTU=9000) features | ||
+ | * netperf run string: | ||
+ | ** Server: netserver -p PORT (5 instances) | ||
+ | ** Client: netperf -p PORT -HOST -t TCP_SENDFILE -l 300 (several instanes) | ||
+ | * Firewall was turned off | ||
+ | * All other tunings were left at default values | ||
+ | |||
+ | === Benchmark Results === | ||
+ | |||
+ | [[File:10gbit_throughput.png]] | ||
+ | |||
+ | === Summary === | ||
+ | |||
+ | *Parallels Virtuozzo Containers support near native 10Gbit network throughput: 9.70Gbit in receive and 9.87Gbit in send tests | ||
+ | *Parallels Virtuozzo Containers shows the best network throughput over all the solutions tested | ||
+ | *In Receive performance test (physical client-> VM/CT) Parallels Virtuozzo Containers shows great benefits over hypervisors: x2 Times faster than ESXi4.1 and x5 Times faster than XenServer5.6 |