White Paper

Zeus Traffic Manager VA Performance on vSphere 4

Zeus. Why wait…

....................................................................... 6 Analysis ...................... along with the system under test...... The requests were generated by a program called Zeusbench 1................................ generated requests and sent them to the system under test............................. such as Apachebench........ 2 Test Setup ................. Why wait.............................................................com/articles/2009/03/26/introducing_zeusbench Z e u s Tr a f f i c M a n a g e r V A P e r f o r m a n c e o n V M w a r e v S p h e r e 4 Page 2 of 8 Zeus....................................................................... ..Contents Introduction .............................. Zeusbench reported the benchmarking results obtained on each of the clients.... which consisted of 64-bit dual-core CPU machines running Linux........................ The clients............................... 1 http://knowledgehub.. 2 System Under Test...................................................................................... Test Setup A series of clients and web servers.................................................................................................. switched Ethernet network using multiple 1Gbps NICs....................................................................................zeus...... Zeusbench produces similar results to those that would be achieved with other benchmarking tools. 3 Benchmarks ............ Zeus Traffic Manager was running on the system under test and load balanced the requests across the web servers............................................................... which consisted of 64-bit dual................................... 3 Virtual Appliance ...... a high-performance................................................................................. were connected to a flat................... 4 Results...0 and outlines the methods used to achieve these performance figures..... 7 Introduction This document details the performance figures obtained by the Zeus Traffic Manager Virtual Appliance (Zeus Traffic Manager VA) on VMware vSphere 4..................... vendor agnostic............................... These results were then collated to ascertain the overall performance of the system under test...... A comparison of the performance of various virtual appliance configurations relative to that of the native install of Linux on the same hardware is also provided... 3 Hardware ............................................................and quadcore CPU machines also running Linux....... HTTP benchmarking tool included with Zeus Traffic Manager............................................... 3 Native Software..........................

. Native Software The performance figures for the native hardware of the system under test were obtained by running a 64-bit Ubuntu 8. All non-essential services were stopped and the following Linux kernel tunings applied: # Increase local port range echo "1024 65535" > /proc/sys/net/ipv4/ip_local_port_range # Disable syncookies echo 0 > /proc/sys/net/ipv4/tcp_syncookies # Increase maximum syn backlog echo 8092 > /proc/sys/net/ipv4/tcp_max_syn_backlog # Increase maximum socket backlog echo 1024 > /proc/sys/net/core/somaxconn # Increase maximum available file descriptors echo 2097152 > /proc/sys/fs/file-max Virtual Appliance The performance figures for the various virtual appliance configurations were obtained using a pre-release version of Zeus Traffic Manager VA 5. Each of the four Intel Pro 1000VT network interfaces was configured on a separate virtual network and added to the virtual machine.System Under Test Hardware The hardware used to conduct the benchmarks consisted of a Dell PowerEdge 2950 containing an Intel® Quad-Core Xeon® E5450 processor. with one of the built-in interfaces used for the ESX service console and the other interface remaining unused. Why wait. When obtaining performance figures. the Zeus Traffic Manager Virtual Appliance under test was the only virtual machine installed on the vSphere server. No additional tuning was performed beyond that which already exists on the standard Virtual Appliance image.0 running on the system under test. VMware tools are preinstalled in the Zeus Traffic Manager VA image. The network interfaces are broken down into four Intel Pro 1000VT interfaces and two interfaces built in to the motherboard. The “vmxnet3” network device was used for each of these virtual network interfaces.. Only the Intel Pro 1000VT interfaces were used to conduct the benchmarks.1r2 virtual appliance configured on vSphere 4.. The virtual machine was allocated 512 MB of RAM and the “vmxThroughputWeight” setting for the ESX server was set to 128.10 Live CD on the machine and installing on it the Linux x86_64 build of Zeus Traffic Manager version 5.1r1. 8GB of RAM and 6 network interfaces. Z e u s Tr a f f i c M a n a g e r V A P e r f o r m a n c e o n V M w a r e v S p h e r e 4 Page 3 of 8 Zeus.

each on a new connection. SSL session IDs are not reused.Benchmarks The intention of the benchmarks was to obtain performance figures for a 64-bit virtual machines running with 1. Clients send a series of requests to Zeus Traffic Manager. Each client sends all its requests on the same connection to Zeus Traffic Manager. SSL connections per second: The rate at which Zeus Traffic Manager can decrypt new SSL connections. each on a new connection.. A short response is returned from the web server for each request. The methodology used for the HTTP requests per second benchmark is used here. SSL throughput: The throughput Zeus Traffic Manager can sustain whilst performing SSL decryption. The test measures the performance of the cipher used to encrypt and decrypt the data passed along the SSL connection. Why wait. Z e u s Tr a f f i c M a n a g e r V A P e r f o r m a n c e o n V M w a r e v S p h e r e 4 Page 4 of 8 Zeus. Layer 4 requests per second: The rate at which Zeus Traffic Manager can handle short requests and responses. The benchmarks conducted consisted of the following tests: Layer 4 connections per second: The rate at which Zeus Traffic Manager can handle new TCP connections. each on a new connection. Zeus Traffic Manager parses the requests and forwards them on to a web server using an established keepalive connection. so each connection requires a full SSL handshake. The web server sends back a response for each request. 2. Zeus Traffic Manager processes each request and forwards it on to a web server over another keepalive connection. 1024 bit RC-4 encryption is used. Each client sends its requests on an existing keepalive connection and SSL session IDs are reused. Clients send a series of HTTPS requests. The web server sends back a response for each request. HTTP connections per second: The rate at which Zeus Traffic Manager can process new HTTP connections. 3 and 4 virtual CPUs. Each client sends its requests down an existing keepalive connection. HTTP throughput: The throughput that Zeus Traffic Manager can sustain when serving large files via HTTP. For each request. These figures could then be compared with those obtained for the native system to discover how performance is affected by running Zeus Traffic Manager inside a virtual machine. Zeus Traffic Manager opens a new connection to a web server and forwards the request to it. HTTP 2K requests per second: The rate at which Zeus Traffic Manager can handle requests for a 2 KB file. HTTP 8K requests per second: The rate at which Zeus Traffic Manager can handle requests for an 8 KB file. The web server responds with a zero sized file. however the files requested are 1 MB in size. which forwards them over a persistent connection to a web server. HTTP requests per second: The rate at which Zeus Traffic Manager can handle HTTP requests. for a zero sized file.. . Clients send a series of HTTP requests.

. Cache throughput: The throughput that Zeus Traffic Manager can sustain when serving large files from its cache..Cache requests per second: The rate at which Zeus Traffic Manager can serve small files directly from its cache. Z e u s Tr a f f i c M a n a g e r V A P e r f o r m a n c e o n V M w a r e v S p h e r e 4 Page 5 of 8 Zeus. . Why wait.

99 2 vCPU 28700 128000 38900 95300 64300 38200 3.Results The following table presents the raw performance figures for the native hardware running Zeus Traffic Manager and each of the virtual machine configurations running Zeus Traffic Manager VA: Test Name L4 conn/s L4 req/s HTTP conn/s HTTP req/s HTTP 2k req/s HTTP 8k req/s HTTP throughput (GBits) SSL conn/s SSL throughput (GBits) Cache req/s Cache throughput (GBits) Native 28800 134000 43400 110000 75200 44800 3..01 4400 1. .97 7790 2.98 6540 2.23 8300 2. Why wait.63 220000 4.73 2350 0.01 1 vCPU 20600 106000 26800 67200 47400 30200 2.99 4 vCPU 30200 123000 42600 104000 63000 37000 2.933 118000 3.99 The following chart compares the performance of each of the virtual machine configurations against the native hardware performance: Performance as a percentage of native 140% 120% 100% 80% 60% 40% 20% 0% 1 vCPU 2 vCPU 3 vCPU 4 vCPU Z e u s Tr a f f i c M a n a g e r V A P e r f o r m a n c e o n V M w a r e v S p h e r e 4 Page 6 of 8 Zeus.31 202000 3.71 180000 4 3 vCPU 35300 140000 51100 113000 69200 38500 2..35 247000 3.

Available at [http://www. so if the necessary workload can be met with a single vCPU then it could improve the overall performance of the system when compared with running a multiple-vCPU appliance to cover the same workload. with the average performance the same as native.vmware. Performance would normally be expected to decrease when configuring fewer vCPUs for the virtual machine.com/pdf/vi_performance_tuning. however. This is due to the hypervisor using some CPU time to manage the network traffic and virtual machines. and the need for the vSphere network layer to consume some resource. The configurations vary by the number of virtual CPUs configured. with the service console requiring some CPU time..pdf] Z e u s Tr a f f i c M a n a g e r V A P e r f o r m a n c e o n V M w a r e v S p h e r e 4 Page 7 of 8 Zeus.0 hypervisor. the 3 vCPU configuration obtained the best performance.5 update 4. the results obtained using vSphere 4. Performance for some configurations and benchmarks. Typically with previous generations of hypervisor there would be a performance hit running under a virtual environment. was faster than running on a native installation of Linux. Why wait. Overall. . with 25% fewer CPUs it is therefore expected that the performance drops by approximately 25%. VMware’s performance tuning guide for ESX2 highlights that running a single-vCPU virtual appliance can have performance benefits over running a multiple-vCPU virtual appliance. in particular with 3 vCPUs.Analysis The chart presented above shows the difference in performance between benchmarks for each virtual machine configuration. The performance of the 1-vCPU appliances in these benchmarks was quite strong. The figures obtained. show that 3vCPUs mostly obtained higher figures than the 4vCPU configuration.0 are approximately 25% better than results obtained on the same System under Test (SUT) using ESX 3. However these numbers show the advantage of the highly optimized vSphere network stack coupled with a tuned Virtual Appliance. and SSL connections/second showing the lowest performance at 78% of native. This is likely to be caused by co-scheduling restrictions. With the exception of SSL connections/second. On production systems running other virtual machines performance figures will scale differently from those obtained here. The SSL connections/second test has a relatively small amount of network traffic and is highly dependent on CPU speed.. This was the maximum network bandwidth available on the SUT. 2 “Performance Tuning Best Practices for ESX Server 3” (VMware). This shows the huge performance improvements in the current vSphere 4. Cache throughput tests show results consistently at 4Gb/second. The results obtained do not show perfectly linear scaling with the number of vCPUs assigned to the virtual machine. using a 10Gb network or more 1Gb network cards cache throughput would be higher.

For further information.com © Zeus Technology Limited 2009.zeus. Zeus Technology.com Zeus Technology. All other brands and product names may be trademarks or registered trademarks of their respective owners. the Zeus logo.com/downloads Technical support is also available during your evaluation.com or visit www. TrafficScript.com or twitter. Zeus Traffic Manager and Cloud Traffic Manager are trademarks of Zeus Technology. California 94402 United States of America. Zeus Technology Limited (UK) The Jeffreys Building Cowley Road Cambridge CB4 0WS United Kingdom Sales: +44 (0)1223 568555 Main: +44 (0)1223 525000 Fax: Web: +44 (0)1223 525100 www. Inc.zeus.S.zeus.com Email: info@zeus. please email: info@zeus.com Stay in touch with Zeus by following: blog.zeus. Phone: 1-888-ZEUS-INC Fax: Web: 1-866-628-7884 www. .zeus. Simply visit our website: www.com/ZeusTechnology Try before you buy. Zeus.com Email: info@zeus. Zeus Web Server. All rights reserved.) 1875 South Grant Street Suite 720 San Mateo. (U.