-
-
Save voellm/1370e09f7f394e3be724 to your computer and use it in GitHub Desktop.
The following instructions walk you step-by-step in setting up a Google Compute Load Balancer benchmark that achieves 1,000,000 Requests Per Second. It is the code and step were used when writing a blog post for the Google Cloud Platform blog. You can find the Google Cloud Platform blog @ http://googlecloudplatform.blogspot.com/ This GIST is a c…
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
***NOTE --- This has been updated since the 1M RPS blog post to run on V1 of the GCE API | |
The scripts below are are broken up into: | |
1. Web Server In-Guest Startup Script used to create the Web Server Image. | |
Called webgueststartup.sh | |
2. Launch the Web Server machines using the previously created image. | |
Called webprovision.bash | |
3. Load Generator In-Guest Startup Script. | |
Called loadgengueststartup.sh | |
4. Launch the Load Generator machines using the Load Generator In-Guest Startup Script. | |
Called loadgen.bash | |
5. Process the Apache logs to determine the Request & Response Throughput. | |
Called computerps.py | |
Instructions on how to use the scripts and run the benchmark: | |
1. Create a project directory where all the scripts will reside. You will be | |
running these scripts from that directory. | |
$ mkdir gclb1m | |
2. Set the environment variable YOURPROJECT to the Google Cloud project you | |
are going to use. Here is an example: | |
$ export YOURPROJECT=voellm | |
3. You will want to enable both GCE and GCS for YOURPROJECT | |
4. Install gcutil (Google Compute Engine command line tool). Make sure it is in | |
your path and you have set up authentication for the project you plan to use. | |
5. Install gsutil (Google Cloud Storage command line tool). Make sure it is in | |
your path and you have set up authentication for the project you plan to use. | |
6. Copy all the scripts using the names provided into your project directory. | |
Make sure they are editable (chmod 755 *). Here is how my project directory | |
looked: | |
$ ls gclb1m | |
computerps.py loadgen.bash loadgengueststartup.sh webgueststartup.sh webprobision.bash | |
7. Now you are ready to setup the Web Server Image. For this we are going to | |
create an instance using the Web Guest Startup script called “webprovision”. | |
We will also use this same instance to gather results at the end so leave it | |
running. When asked for an image to boot choose the latest Wheezy7 image. If | |
you want to understand why certain instructions below are executed to create | |
the image read more about Creating GCE images (https://developers.google.com/compute/docs/images) | |
on the Cloud Platform Site. One last detail is to change “gs://voellm” to | |
your Google Cloud Storage Bucket. If you don't have one you should create one | |
on the Cloud Console (http://cloud.google.com/console). | |
$ cd gclb1m | |
$ gcutil --project=$YOURPROJECT addinstance webprovision --zone=europe-west1-a --machine_type=n1-standard-1 --metadata_from_file=startup-script:webgueststartup.sh | |
<.... choose an image ....> | |
$ gcutil --project=$YOURPROJECT ssh webprovision | |
<.... you are now running in the guest! ....> | |
$ ps -eaf | grep apache | |
<.... if apache is running go to the next step otherwise check again ....> | |
$ exit | |
<.... leaving the guest ....> | |
8. Download curl_loader to your project directory. Then follow the steps below | |
to build curl_loader in the webprovision machine. We do this to make sure we | |
don't hit any library issues. | |
$ gcutil --project=$YOURPROJECT push webprovision curl-loader-0.56.tar.bz2 . | |
$ gcutil --project=$YOURPROJECT ssh webprovision | |
<.... you are now running in the guest! ....> | |
$ sudo apt-get install bzip2 -y | |
$ sudo apt-get install build-essential -y | |
$ sudo apt-get install libssl-dev -y | |
$ tar jxf curl-loader-0.56.tar.bz2 | |
$ cd curl-loader-0.56 | |
$ make debug=0 optimize=1 | |
$ sudo make install | |
$ cp curl-loader .. | |
$ exit | |
<.... leaving the guest ....> | |
9. Now you are ready to create the image for the websever and loadgen VM's. | |
$ gcutil --project=$YOURPROJECT ssh webprovision | |
<.... you are now running in the guest! ....> | |
$ sudo gcimagebundle -d /dev/sda -o /tmp/ --log_file=/tmp/abc.log | |
<.... this takes a couple of minutes ....> | |
$ cd /tmp | |
$ ls # I am looking for a file with <numbers>.image.tar.gz | |
$ exit | |
<.... leaving the guest ....> | |
$ gcutil --project=$YOURPROJECT pull webprovision /tmp/<numbers>.image.tar.gz . | |
$ gsutil mb gs://voellm | |
$ gsutil cp <numbers>.image.tar.gz gs://voellm | |
$ gcutil --project=$YOURPROJECT addimage gclb1mimage gs://voellm/<numbers>.image.tar.gz | |
<.... you will be prompted for a kernel. Pick a latest with no-conn-track in the name ....> | |
10. Now you are ready to kick off your Web Server VMs | |
$ chmod 755 webprovision.bash | |
$ ./webprovision.bash start | |
11. Now you are ready to kick off the load generator. This is fun! | |
$ chmod 755 loadgen.bash | |
$ ./loadgen.bash start | |
12. As a sanity check you can make sure all the VMs are running. | |
$ gcutil --project=$YOURPROJECT listinstances | |
+--------------+----------------+---------+----------------+----------------+ | |
| name | zone | status | network-ip | external-ip | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw0 | europe-west1-a | RUNNING | 10.164.190.193 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw1 | europe-west1-a | RUNNING | 10.206.175.123 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw10 | europe-west1-a | RUNNING | 10.253.142.96 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw2 | europe-west1-a | RUNNING | 10.203.55.215 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw3 | europe-west1-a | RUNNING | 10.77.95.15 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw4 | europe-west1-a | RUNNING | 10.15.188.148 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw5 | europe-west1-a | RUNNING | 10.60.188.93 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw6 | europe-west1-a | RUNNING | 10.74.212.116 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw7 | europe-west1-a | RUNNING | 10.122.236.87 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw8 | europe-west1-a | RUNNING | 10.3.243.104 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| gw9 | europe-west1-a | RUNNING | 10.173.52.23 | | | |
+--------------+----------------+---------+----------------+----------------+ | |
| lg0 | europe-west1-a | RUNNING | 10.117.68.136 | 8.34.221.4 | | |
+--------------+----------------+---------+----------------+----------------+ | |
| lg1 | europe-west1-a | RUNNING | 10.73.139.229 | 8.34.221.11 | | |
+--------------+----------------+---------+----------------+----------------+ | |
| lg2 | europe-west1-a | RUNNING | 10.98.64.173 | 8.34.221.16 | | |
+--------------+----------------+---------+----------------+----------------+ | |
| lg3 | europe-west1-a | RUNNING | 10.3.142.86 | 8.34.211.6 | | |
+--------------+----------------+---------+----------------+----------------+ | |
| lg4 | europe-west1-a | RUNNING | 10.78.35.79 | 8.34.222.196 | | |
+--------------+----------------+---------+----------------+----------------+ | |
| webprovision | europe-west1-a | RUNNING | 10.184.150.3 | 8.34.222.70 | | |
+--------------+----------------+---------+----------------+----------------+ | |
13. Once this has run for a while you can stop it | |
$ ./loadgen.bash stop | |
14. Now we we need to capture all the result files to check the RPS achieved. | |
The first step is to get the public ssh key into the guest. Then we copy | |
the files from all the web servers. Then we copy them locally to be | |
processed. We have to use webprovision VM as an intermediary because the | |
web servers were started without a public IP address. | |
$ gcutil --project=$YOURPROJECT push webprovision ~/.ssh/google_compute_engine . | |
$ gcutil --project=$YOURPROJECT ssh webprovision | |
<.... you are now running in the guest! ....> | |
$ chmod 600 google_compute_engine | |
$ mv google_compute_engine ~/.ssh/ | |
$ for i in {0..199}; do scp -i ~/.ssh/google_compute_engine -o CheckHostIP=no -o StrictHostKeyChecking=no $USERNAME@gw$i:/var/log/apache2/access.log access$i.log; done | |
$ exit | |
<.... leaving the guest ....> | |
$ mkdir result1 | |
$ gcutil --project=$YOURPROJECT pull webprovision access* result1 | |
15. Now that the results are local you can process them to get the final | |
result.csv you can load into your favorite spreadsheet. The result.csv | |
file has “<time>,<requests per second>” one per line for all time | |
intervals computed. | |
$ cd result1 | |
$ ../computerps.py > result.csv | |
$ cd .. | |
16. Delete your web server and webprovision instance. | |
$ ./webprovision.bash stop | |
$ gcutil --project=$YOURPROJECT deleteinstance webprovision | |
######################################################## | |
## 1. Web Server Guest Startup Script - webgueststartup.sh | |
######################################################## | |
#!/bin/sh | |
apt-get -y update | |
apt-get -y install apache2 | |
rm /etc/apache2/sites-available/default-ssl | |
echo 'LogFormat "%{%s}t" time' >> /etc/apache2/apache2.conf | |
echo '<VirtualHost *:80> | |
ServerAdmin webmaster@localhost | |
DocumentRoot /var/www | |
<Directory /> | |
Options FollowSymLinks | |
AllowOverride None | |
</Directory> | |
<Directory /var/www/> | |
Options Indexes FollowSymLinks MultiViews | |
AllowOverride None | |
Order allow,deny | |
allow from all | |
</Directory> | |
ErrorLog ${APACHE_LOG_DIR}/error.log | |
LogLevel warn | |
CustomLog ${APACHE_LOG_DIR}/access.log time | |
</VirtualHost>' > /etc/apache2/sites-enabled/000-default | |
echo ' | |
<IfModule mpm_worker_module> | |
StartServers 2 | |
MinSpareThreads 25 | |
MaxSpareThreads 75 | |
ThreadLimit 64 | |
ThreadsPerChild 25 | |
MaxClients 400 | |
MaxRequestsPerChild 0 | |
</IfModule> | |
' >> /etc/apache2/apache2.conf | |
/etc/init.d/apache2 restart | |
dd if=/dev/zero of=/var/www/tiny bs=1 count=0 seek=1 | |
echo 'net.core.wmem_max=12582912' >> /etc/sysctl.conf | |
echo 'net.core.rmem_max=12582912' >> /etc/sysctl.conf | |
echo 'net.ipv4.tcp_rmem= 10240 87380 12582912' >> /etc/sysctl.conf | |
echo 'net.ipv4.tcp_wmem= 10240 87380 12582912' >> /etc/sysctl.conf | |
echo 'net.ipv4.tcp_window_scaling = 1' >> /etc/sysctl.conf | |
echo 'net.ipv4.tcp_timestamps = 1' >> /etc/sysctl.conf | |
echo 'net.ipv4.tcp_sack = 1' >> /etc/sysctl.conf | |
echo 'net.core.netdev_max_backlog = 5000' >> /etc/sysctl.conf | |
sysctl -p | |
######################################################## | |
## 2. Launch the Web Servers - webprovision.bash | |
######################################################## | |
#! /bin/bash | |
# Usage webprovision.bash [start | stop] | |
# start means to create the web serving environment, LB, etc | |
# stop mean to teardown the web serving environment. LB, etc | |
/bin/date | |
STARTUP_SCRIPT=lb_startup.sh | |
MACHINE_TYPE=n1-standard-1 | |
PROJECT_ID=$YOURPROJECT | |
IMAGE=gclb1mimage | |
INSTANCE_NAMES1=() | |
INSTANCE_NAMES2=() | |
INSTANCE_NAMES3=() | |
INSTANCE_NAMES4=() | |
for i in {0..49} | |
do | |
INSTANCE_NAMES1=("${INSTANCE_NAMES1[@]}" "europe-west1-a/instances/gw$i") | |
done | |
for i in {50..99} | |
do | |
INSTANCE_NAMES2=("${INSTANCE_NAMES2[@]}" "europe-west1-a/instances/gw$i") | |
done | |
for i in {100..149} | |
do | |
INSTANCE_NAMES3=("${INSTANCE_NAMES3[@]}" "europe-west1-a/instances/gw$i") | |
done | |
for i in {150..199} | |
do | |
INSTANCE_NAMES4=("${INSTANCE_NAMES4[@]}" "europe-west1-a/instances/gw$i") | |
done | |
INSTANCE_NAMES_SS1=$( IFS=$' '; echo "${INSTANCE_NAMES1[*]}" ) | |
INSTANCE_NAMES_CS1=$( IFS=$','; echo "${INSTANCE_NAMES1[*]}" ) | |
INSTANCE_NAMES_SS2=$( IFS=$' '; echo "${INSTANCE_NAMES2[*]}" ) | |
INSTANCE_NAMES_CS2=$( IFS=$','; echo "${INSTANCE_NAMES2[*]}" ) | |
INSTANCE_NAMES_SS3=$( IFS=$' '; echo "${INSTANCE_NAMES3[*]}" ) | |
INSTANCE_NAMES_CS3=$( IFS=$','; echo "${INSTANCE_NAMES3[*]}" ) | |
INSTANCE_NAMES_SS4=$( IFS=$' '; echo "${INSTANCE_NAMES4[*]}" ) | |
INSTANCE_NAMES_CS4=$( IFS=$','; echo "${INSTANCE_NAMES4[*]}" ) | |
ZONE="europe-west1-a" | |
REGION="europe-west1" | |
if [ "$1" == "start" ] | |
then | |
# Create the LB | |
time gcutil --project=$PROJECT_ID addfirewall gw-firewall --allowed=tcp:80 | |
time gcutil --project=$PROJECT_ID addtargetpool gw-pool --region=$REGION | |
time gcutil --project=$PROJECT_ID addforwardingrule gw-rule --region=$REGION --port_range=80 --target=gw-pool | |
# Boot the machines | |
time gcutil --project=$PROJECT_ID addinstance $INSTANCE_NAMES_SS1 --image=$IMAGE --zone=$ZONE --machine_type=$MACHINE_TYPE --external_ip_address=none & | |
time gcutil --project=$PROJECT_ID addinstance $INSTANCE_NAMES_SS2 --image=$IMAGE --zone=$ZONE --machine_type=$MACHINE_TYPE --external_ip_address=none & | |
time gcutil --project=$PROJECT_ID addinstance $INSTANCE_NAMES_SS3 --image=$IMAGE --zone=$ZONE --machine_type=$MACHINE_TYPE --external_ip_address=none & | |
time gcutil --project=$PROJECT_ID addinstance $INSTANCE_NAMES_SS4 --image=$IMAGE --zone=$ZONE --machine_type=$MACHINE_TYPE --external_ip_address=none | |
# Add the machine to the LB | |
time gcutil --project=$PROJECT_ID addtargetpoolinstance gw-pool --region=$REGION --instances=$INSTANCE_NAMES_CS1 | |
time gcutil --project=$PROJECT_ID addtargetpoolinstance gw-pool --region=$REGION --instances=$INSTANCE_NAMES_CS2 | |
time gcutil --project=$PROJECT_ID addtargetpoolinstance gw-pool --region=$REGION --instances=$INSTANCE_NAMES_CS3 | |
time gcutil --project=$PROJECT_ID addtargetpoolinstance gw-pool --region=$REGION --instances=$INSTANCE_NAMES_CS4 | |
else | |
if [ "$1" == "stop" ] | |
then | |
time gcutil --project=$PROJECT_ID deleteforwardingrule -f gw-rule --region=$REGION | |
time gcutil --project=$PROJECT_ID deletetargetpool -f gw-pool --region=$REGION | |
time gcutil --project=$PROJECT_ID deleteinstance -f --delete_boot_pd $INSTANCE_NAMES_SS1 --zone=$ZONE | |
time gcutil --project=$PROJECT_ID deleteinstance -f --delete_boot_pd $INSTANCE_NAMES_SS2 --zone=$ZONE | |
time gcutil --project=$PROJECT_ID deleteinstance -f --delete_boot_pd $INSTANCE_NAMES_SS3 --zone=$ZONE | |
time gcutil --project=$PROJECT_ID deleteinstance -f --delete_boot_pd $INSTANCE_NAMES_SS4 --zone=$ZONE | |
time gcutil --project=$PROJECT_ID deletefirewall -f gw-firewall | |
else | |
echo "$0 [stop | start]" | |
fi | |
fi | |
/bin/date | |
######################################################## | |
## 3. Load Generator In-Guest Startup Script - loadgengueststartup.sh | |
######################################################## | |
#!/bin/sh | |
IP=`/sbin/ifconfig eth0 | grep 'inet addr:' | cut -d: -f2 | awk '{ print $1}'` | |
echo "BATCH_NAME= test | |
CLIENTS_NUM_MAX=1000 | |
CLIENTS_NUM_START=1000 | |
CLIENTS_RAMPUP_INC=10 | |
NETMASK=32 | |
INTERFACE =eth0 | |
IP_ADDR_MIN=$IP | |
IP_ADDR_MAX=$IP | |
CYCLES_NUM= -1 | |
URLS_NUM= 1 | |
########### URLs SECTION ####################### | |
# Change URL to be xxx.xxx.xxx.xxx/tiny where the x’s are the | |
# ip of the forwarding rule of the load balancer | |
URL=FORWARDINGIP | |
URL_SHORT_NAME="tiny" | |
REQUEST_TYPE=GET | |
TIMER_URL_COMPLETION = 0 | |
TIMER_AFTER_URL_SLEEP = 0" > /tmp/test.conf | |
######################################################## | |
## 4. Launch The Load Generators - loadgen.bash | |
######################################################## | |
#! /bin/bash | |
# Usage loadgen.bash [start | stop] | |
# start the load generators instances | |
# stop the load generator instances | |
LG_SCRIPT=loadgengueststartup.sh | |
PROJECT_ID=$YOURPROJECT | |
LG_MACHINE_TYPE=n1-standard-4 | |
IMAGE=gclb1mimage | |
ZONE="europe-west1-a" | |
REGION="europe-west1" | |
LG_NAMES=() | |
for i in {0..63} | |
do | |
LG_NAMES=("${LG_NAMES[@]}" "lg$i") | |
done | |
LG_NAMES_SS=$( IFS=$' '; echo "${LG_NAMES[*]}" ) | |
LG_NAMES_CS=$( IFS=$','; echo "${LG_NAMES[*]}" ) | |
if [ "$1" == "start" ] | |
then | |
export IP=`gcutil --project=$YOURPROJECT getforwardingrule gw-rule | awk -F '|' '{if (index($2, "ip ") != 0) print $3;}'` | |
sed "s/FORWARDINGIP/$IP/g" $LG_SCRIPT > temp.sh | |
gcutil --project=$PROJECT_ID addinstance $LG_NAMES_SS --image=$IMAGE --zone=$ZONE --machine_type=$LG_MACHINE_TYPE --metadata_from_file=startup-script:temp.sh | |
# Compile curl-loader v.56 with debug off and optimize on and distribute the binary | |
# to each load generator | |
# Start curl-loader on each load generator (test.conf should point to wherever the startup | |
# script put the conf file) | |
for i in {0..63} | |
do | |
gcutil --project=$PROJECT_ID ssh ${LG_NAMES[i]} "sh -c 'nohup sudo ./curl-loader -w -r -t 16 -f /tmp/test.conf > nohup.out 2>&1 &'" & | |
done | |
else | |
gcutil --project=$PROJECT_ID deleteinstance -f --delete_boot_pd $LG_NAMES_SS --zone=$ZONE | |
fi | |
######################################################## | |
## 5. Process the Apache logs to determine RPS - computerps.py | |
######################################################## | |
#!/usr/bin/env python | |
import glob | |
file_names = glob.glob("*.log") | |
files = [] | |
time = float("inf") | |
next_lines = dict() | |
for f in file_names: | |
log = open(f, 'r') | |
files.append(log) | |
next_lines[log] = log.readline() | |
time = min(time, int(next_lines[log])) | |
linesleft = True | |
while linesleft: | |
total = 0 | |
for log in files: | |
while len(next_lines[log]) != 0 and time > int(next_lines[log]): | |
next_lines[log] = log.readline() | |
if len(next_lines[log]) == 0 or time < int(next_lines[log]): | |
continue | |
count = 0 | |
next_line = log.readline() | |
while len(next_line) != 0 and int(next_lines[log]) >= int(next_line): | |
count += 1 | |
next_line = log.readline() | |
next_lines[log] = next_line | |
total += count | |
if total != 0: | |
print str(time) + "," + str(total) | |
if total != 0: | |
time += 1 | |
else: | |
next_time = float("inf") | |
for line in next_lines.values(): | |
if len(line) != 0: | |
next_time = min(next_time, int(line)) | |
time = next_time | |
linesleft = False | |
for line in next_lines.values(): | |
if line: | |
linesleft = True | |
break | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
the for loop seems not necesary: