Komiser:Detect potential cost savings on GCP

I’m super excited to annonce the release of Komiser:2.1.0 with beta support of Google Cloud Platform. You can now use one single open source tool to detect both AWS and GCP overspending.

Highlights



With the GDPR becoming real in EU, logging and storage of (potentially) personally identifiable information now need to be reduced in many organizations. Komiser allows you to analyze and manage cloud cost, usage, security, and governance in one place. Hence, detecting potential vulnerabilities that could put your cloud environment at risk.

It allows you also to control your usage and create visibility across all used services to achieve maximum cost-effectiveness and get a deep understanding of how you spend on the AWS, GCP and Azure.



Usage

Below are the available downloads for the latest version of Komiser (2.1.0). Please download the proper package for your operating system and architecture.

Linux:

1
wget https://cli.komiser.io/2.1.0/linux/komiser

Windows:

1
wget https://cli.komiser.io/2.1.0/windows/komiser

Mac OS X:

1
wget https://cli.komiser.io/2.1.0/osx/komiser

Note: make sure to add the execution permission to Komiser chmod +x komiser and update the user’s $PATH variable.

Komiser is also available as a Docker image:

Docker:

1
docker run -d -p 3000:3000 --name komiser mlabouardy/komiser:2.1.0

Note that we need to provide the three environment variables AWS_DEFAULT_REGION, AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY set in the container such as that the CLI can automatically authenticate with AWS.

Create a service account with Viewer permission, see Creating and managing service accounts docs.

Enable the below APIs for your project through GCP Console, gcloud or using the Service Usage API. You can find out more about these options in Enabling an API in your GCP project docs.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
appengine.googleapis.com
bigquery-json.googleapis.com
compute.googleapis.com
cloudfunctions.googleapis.com
container.googleapis.com
cloudresourcemanager.googleapis.com
cloudkms.googleapis.com
dns.googleapis.com
dataflow.googleapis.com
dataproc.googleapis.com
iam.googleapis.com
monitoring.googleapis.com
pubsub.googleapis.com
redis.googleapis.com
serviceusage.googleapis.com
storage-api.googleapis.com
sqladmin.googleapis.com

To analyze and optimize the infrastructure cost, you need to export your daily cost to BigQuery, see Export Billing to BigQuery docs.

Provide authentication credentials to your application code by setting the environment variable GOOGLE_APPLICATION_CREDENTIALS:

1
export GOOGLE_APPLICATION_CREDENTIALS="[PATH]"

That should be it. Try out the following from your command prompt to start the server:

1
komiser start --port 3000 --dataset project-id.dataset-name.table-name

If you point your favorite browser to http://localhost:3000, you should see Komiser awesome dashboard:



The versioned documentation can be found on https://docs.komiser.io.

Komiser is written in Golang and is MIT licensed — contributions are welcomed whether that means providing feedback or testing existing and new features.


https://komiser.io

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Komiser:Optimize Cost and Security on AWS

Over the last decade, the cost of Amazon Web Services (AWS) has become a primary concern of businesses. That’s no surprise: AWS has many services that offer a range of IT resources — from IT infrastructure and bandwidth to analytics tools and machine learning — and each affects the total cloud bill.

While AWS offers many fully-managed services like CloudWatch, CloudTrail, Trusted Advisor, etc to help you detect potential cost savings. Understanding and managing cloud costs isn’t simple with AWS.

That’s why, I came up one year ago, with an open source tool called Komiser to help reduce your AWS infrastructure cost based on custom recommendations.

After 1 year of intense development, I’m thrilled to announce the fresh new release of Komiser: 2.0.0 with support of new AWS services:


AWS Services supported by Komiser

Highlights



With the GDPR becoming real in EU, logging and storage of (potentially) personally identifiable information now need to be reduced in many organizations. Komiser allows you to analyze and manage cloud cost, usage, security, and governance in one place. Hence, detecting potential vulnerabilities that could put your cloud environment at risk.

It allows you also to control your usage and create visibility across all used services to achieve maximum cost-effectiveness and get a deep understanding of how you spend on the AWS, GCP and Azure.



Usage

Below are the available downloads for the latest version of Komiser (2.0.0). Please download the proper package for your operating system and architecture.

Linux:

1
wget https://cli.komiser.io/2.0.0/linux/komiser

Windows:

1
wget https://cli.komiser.io/2.0.0/windows/komiser

Mac OS X:

1
wget https://cli.komiser.io/2.0.0/osx/komiser

Note: make sure to add the execution permission to Komiser chmod +x komiser and update the user’s $PATH variable.

Komiser is also available as a Docker image:

Docker:

1
docker run -d -p 3000:3000 --name komiser mlabouardy/komiser:2.0.0

If you point your favorite browser to http://localhost:3000, you should see Komiser awesome dashboard:



The versioned documentation can be found on https://docs.komiser.io.

Komiser is written in Golang and is MIT licensed — contributions are welcomed whether that means providing feedback or testing existing and new features.


https://komiser.io

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Ingest Data from RDS MySQL to Google BigQuery

In analytics, where queries over hundreds of gigabytes are the norm, performance is paramount and has a direct effect on the productivity of your team: running a query for hours means days of iterations between business questions. At Foxintelligence, we needed to move from traditional relational databases, like Postgres and MySQL to columnar database solutions. While RDBS like MySQL is great for normal transactional operations, it has significant drawbacks when it comes to real-time analytics on large amount of data. We found Google BigQuery to deliver superior results significantly for usability, performance, and cost for almost all our analytical use-cases, especially at scale.

Both Amazon RedShift and Google BigQuery provide much of the same functionalities, there are some fundamental differences between how these two operate. So you need to pick the right solution based on your data and business.

Once we decided which data warehouse we will use, we had to replicate data from RDS MySQL to Google BigQuery. This post walks you through the process of creating a data pipeline to achieve the replication between the two systems.

We used AWS Data Pipeline to export data from MySQL and feed it to BigQuery. The figure below summarises the entire workflow:



The pipeline starts based on a defined schedule and period, it launches a spot instance that will copy data from MySQL database to CSV files (split by table name) to an Amazon S3 bucket and then sending an Amazon SNS notification after the copy activity completes successfully. Following is our pipeline that accomplishes that:



Once the pipeline is finished, CSV files will be generated in the output S3 bucket:



The SNS notification will trigger a Lambda function, it will deploy a batch job based on a Docker image stored on our private Docker registry. The container will upload CSV files from S3 to GCS and load data to BigQuery:

You can use Storage Transfer Service to easily migrate your data from Amazon S3 to Cloud Storage.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
#!/bin/bash

echo "Download BigQuery Credentials"

aws s3 cp s3://$GCP_AUTH_BUCKET/auth.json .

echo "Upload CSV to GCS"

mkdir -p csv
rm tables

for raw in $(aws s3 ls s3://$S3_BUCKET/ | awk -F " " '{print $2}');
do
table=${raw%/}
if [[ $table != "" && $table != df* ]]
then
echo "Table: $table"
csv=$(aws s3 ls s3://$S3_BUCKET/$table/ | awk -F " " '{print $4}' | grep ^ | sort -r | head -n1)

echo $table >> tables

echo "CSV: $csv"

echo "Copy csv from S3"
aws s3 cp s3://$S3_BUCKET/$table/$csv csv/$table.csv

echo "Upload csv to GCP"
gsutil cp csv/$table.csv gs://$GS_BUCKET/$table.csv
fi
done

echo "Import CSV to BigQuery"

python app.py

We have written a Python script to clean up raw data (encoding issues), transform (map MySQL data types to BQ data types) and load CSV file to BigQuery:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
import mysql.connector
import os
import time
from mysql.connector import Error
from google.cloud import bigquery

bigquery_client = bigquery.Client()

def mapToBigQueryDataType(columnType):
if columnType.startswith('int'):
return 'INT64'
if columnType.startswith('varchar'):
return 'STRING'
if columnType.startswith('decimal'):
return 'FLOAT64'
if columnType.startswith('datetime'):
return 'DATETIME'
if columnType.startswith('text'):
return 'STRING'
if columnType.startswith('date'):
return 'DATE'
if columnType.startswith('time'):
return 'TIME'

def wait_for_job(job):
while True:
job.reload()
if job.state == 'DONE':
if job.error_result:
raise RuntimeError(job.errors)
return
time.sleep(1)

try:
conn = mysql.connector.connect(host=os.environ['MYSQL_HOST'],
database=os.environ['MYSQL_DB'],
user=os.environ['MYSQL_USER'],
password=os.environ['MYSQL_PWD'])
if conn.is_connected():
print('Connected to MySQL database')

lines = open('tables').read().split("\n")
for tableName in lines:
print('Table:',tableName)

cursor = conn.cursor()
cursor.execute('SHOW FIELDS FROM '+os.environ['MYSQL_DB']+'.'+tableName)

rows = cursor.fetchall()

schema = []
for row in rows:
schema.append(bigquery.SchemaField(row[0].replace('\'', ''), mapToBigQueryDataType(row[1])))


job_config = bigquery.LoadJobConfig()
job_config.source_format = bigquery.SourceFormat.CSV
job_config.autodetect = True
job_config.max_bad_records = 2
job_config.allow_quoted_newlines = True
job_config.schema = schema

job = bigquery_client.load_table_from_uri(
'gs://'+os.environ['GCE_BUCKET']+'/'+tableName+'.csv',
bigquery_client.dataset(os.environ['BQ_DATASET']).table(tableName),
location=os.environ['BQ_LOCATION'],
job_config=job_config)

print('Loading data to BigQuery:', tableName)

wait_for_job(job)


print('Loaded {} rows into {}:{}.'.format(
job.output_rows, os.environ['BQ_DATASET'], tableName))

except Error as e:
print(e)
finally:
conn.close()

As a result, the tables will be imported to BigQuery:



While this solution worked like a charm, we didn’t stop there. Google Cloud announced the public beta release of BigQuery Data Transfer. This service allows you to automates data movement from multiple data sources like S3 or GCS to BigQuery on a scheduled, managed basis. So it was a great use case to test this service to manage recurring load jobs from Amazon S3 into BigQuery as shown in the figure below:



This services comes with some trade-offs such as Google BigQuery cannot create tables as part of data transfer process. Hence, a Lambda function was used to drop the old dataset, and create the destination tables and their schema in advance of running the transfer. The function handler code is self-explanatory:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
func handler(ctx context.Context) error {
client, err := bigquery.NewClient(ctx, os.Getenv("PROJECT_ID"))
if err != nil {
return err
}

err = RemoveDataSet(client)
if err != nil {
return err
}

err = CreateDataSet(client)
if err != nil {
return err
}

uri := fmt.Sprintf("%s:%s@tcp(%s)/%s",
os.Getenv("MYSQL_USERNAME"), os.Getenv("MYSQL_PASSWORD"),
os.Getenv("MYSQL_HOST"), os.Getenv("MYSQL_DATABASE"))

db, err := sql.Open("mysql", uri)
if err != nil {
return err
}

file, err := os.Open("tables")
if err != nil {
return err
}
defer file.Close()

scanner := bufio.NewScanner(file)
for scanner.Scan() {
tableName := scanner.Text()
fmt.Println("Table:", tableName)

columns, _ := GetColumns(tableName, db)
fmt.Println("Columns:", columns)

CreateBQTable(tableName, columns, client)
}

if err := scanner.Err(); err != nil {
return err
}
return nil
}

func main() {
lambda.Start(handler)
}

The function will be triggered by a CloudWatch Event, once the data pipeline finishes exporting CSV files:



Finally, we created Transfer jobs for each table on BigQuery to load data from S3 bucket to BigQuery table:



Using Google BigQuery to store internally hundreds of gigabytes of data (soon terabytes) with the capability to analyse it in few seconds give us a massive push toward business intelligence and data-driven insights.

Like what you’re read­ing? Check out my book and learn how to build, secure, deploy and manage production-ready Serverless applications in Golang with AWS Lambda.

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

CI/CD for Android and iOS Apps on AWS

Mobile apps have taken center stage at Foxintelligence. After implementing CI/CD workflows for Dockerized Microservices, Serverless Functions and Machine Learning models, we needed to automate the release process of our mobile application — Cleanfox — to deliver features we are working on continuously and ensure high quality app. While the CI/CD concepts remains the same, its practicalities are somewhat different. In this post, I will walk you through how we achieved that, including the lessons learned and formed along the way to boost your Android and iOS application development drastically.



The Jenkins cluster (figure below) consists of a dedicated Jenkins master with a couple of slave nodes inside an autoscaling group. However, iOS apps can be built only on macOS machine. We typically use an unused Mac Mini computer located in the office devoted to these tasks.

We have configured the Mac mini to establish a VPN connection (at system startup) to the OpenVPN server deployed on the target VPC.



We setup an SSH tunnel to the Mac node using dynamic port forwarding. Once the tunnel is active, you can add the machine to Jenkins set of worker nodes:



This guide assumes you have a fresh install of the latest stable version of Xcode along with Fastlane.

Once we had a good part of this done, we used Fastlane to automate the deployment process. This tool offers a set of scripts written in Ruby to handle tedious tasks such as code signing, managing certificates and releasing ipa to the app store for the end users.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
default_platform(:ios)

platform :ios do

lane :tests do
scan(
scheme: options[:scheme],
clean: true,
skip_detect_devices: true,
build_for_testing: true,
sdk: 'iphoneos',
should_zip_build_products: true
)
firebase_test_lab_ios_xctest(
gcp_project: 'cleanfox-XXXX',
devices: [
{
ios_model_id: 'ipadmini4',
ios_version_id: '12.0',
locale: 'fr_FR',
orientation: 'portrait'
},
{
ios_model_id: 'iphone7plus',
ios_version_id: '12.0',
locale: 'fr_FR',
orientation: 'portrait'
},
{
ios_model_id: 'iphone8',
ios_version_id: '12.0',
locale: 'fr_FR',
orientation: 'portrait'
},
{
ios_model_id: 'iphonexsmax',
ios_version_id: '12.0',
locale: 'fr_FR',
orientation: 'portrait'
}
]
)
end

lane :increment_build do
version = get_version_number
latestBuildNumber = latest_testflight_build_number(version: version)
increment_build_number(
build_number: latestBuildNumber + 1,
xcodeproj: "Cleanfox.xcodeproj"
)
end

lane :develop do
increment_build
build_app(scheme: "Sandbox",
workspace: "Cleanfox.xcworkspace",
include_bitcode: true)
end

lane :beta do
increment_build
build_app(scheme: "Staging",
workspace: "Cleanfox.xcworkspace",
include_bitcode: true)
upload_to_testflight
end

lane :prod_testflight do
increment_build_number(
build_number: latest_testflight_build_number + 1,
xcodeproj: "Cleanfox.xcodeproj"
)
build_app(scheme: "Production",
workspace: "Cleanfox.xcworkspace",
include_bitcode: true)
upload_to_testflight(skip_waiting_for_build_processing: true)
end
end

Also, we created a Jenkinsfile, which defines a set of steps (each step calls a certain actions — lane — defined in the above Fastfile) that will be executed on Jenkins based on the branch name (GitFlow model):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
def bucket = 'mobile-artifacts-foxintelligence'

node('mac') {
try {
stage('Checkout') {
checkout scm
notifySlack('STARTED')
}

stage('Install Dependencies') {
sh "pod install"
}

stage('Build') {
if (env.BRANCH_NAME == 'master'){
sh "bundle exec fastlane prod_testflight"
}
if (env.BRANCH_NAME == 'preprod'){
sh "bundle exec fastlane staging"
}
if (env.BRANCH_NAME == 'develop'){
sh "bundle exec fastlane develop"
}
}

stage('Push') {
sh "aws s3 cp Cleanfox.ipa s3://${bucket}/ios/cleanfox-${commitID()}.ipa"

if (env.BRANCH_NAME == 'master'){
sh "aws s3 cp Cleanfox.ipa s3://${bucket}/ios/cleanfox-latest.ipa"
}
if (env.BRANCH_NAME == 'preprod'){
sh "aws s3 cp Cleanfox.ipa s3://${bucket}/ios/cleanfox-preprod.ipa"
}
if (env.BRANCH_NAME == 'develop'){
sh "aws s3 cp Cleanfox.ipa s3://${bucket}/ios/cleanfox-develop.ipa"
}
}

stage('Test') {
if (env.BRANCH_NAME == 'master'){
sh 'bundle fastlane tests --scheme "Production"'
}
if (env.BRANCH_NAME == 'preprod'){
sh 'bundle fastlane tests --scheme "Staging"'
}
if (env.BRANCH_NAME == 'develop'){
sh 'bundle fastlane tests --scheme "Sandbox"'
}
}
}catch(e){
currentBuild.result = 'FAILED'
throw e
}finally{
notifySlack(currentBuild.result)
}
}

def notifySlack(String buildStatus){
buildStatus = buildStatus ?: 'SUCCESSFUL'
def colorCode = '#FF0000'
def subject = "Name: '${env.JOB_NAME}'\nStatus: ${buildStatus}\nBuild ID: ${env.BUILD_NUMBER}"
def summary = "${subject}\nMessage: ${commitMessage()}\nAuthor: ${commitAuthor()}\nURL: ${env.BUILD_URL}"

if (buildStatus == 'STARTED') {
colorCode = '#546e7a'
} else if (buildStatus == 'SUCCESSFUL') {
colorCode = '#2e7d32'
} else {
colorCode = '#c62828c'
}

slackSend (color: colorCode, message: summary)
}

def commitAuthor(){
sh 'git show -s --pretty=%an > .git/commitAuthor'
def commitAuthor = readFile('.git/commitAuthor').trim()
sh 'rm .git/commitAuthor'
commitAuthor
}

def commitID() {
sh 'git rev-parse HEAD > .git/commitID'
def commitID = readFile('.git/commitID').trim()
sh 'rm .git/commitID'
commitID
}

def commitMessage() {
sh 'git log --format=%B -n 1 HEAD > .git/commitMessage'
def commitMessage = readFile('.git/commitMessage').trim()
sh 'rm .git/commitMessage'
commitMessage
}

The pipeline is divided into 5 stages:

  • Checkout: clone the GitHub repository.
  • Quality & Unit Tests: check whether our code is well formatted and follows Swift best practices and run unit tests.
  • Build: build and sign the app.
  • Push: store the deployment package (.ipa file) to an S3 bucket.
  • UI Test: launch UI tests on Firebase Test Lab across a wide variety of devices and device configurations.

If a build on the CI passes, a Slack notification will be sent (broken build will notify developers to investigate immediately).

Note the usage of the git commit ID as a name for the deployment package to give a meaningful and significant name for each release and be able to roll back to a specific commit if things go wrong.

Once the pipeline is triggered, a new build should be created as follows:



At the end, Jenkins will launch UI Tests based on XCTest framework on Firebase Test Lab across multiple virtual and physical devices and different screen sizes.



We gave a try to AWS Device Farm, but we needed to get over 2 problems at the same time. We sought waiting for a very short time, to receive tests result, without paying too much.

Test Lab exercises your app on devices installed and running in a Google data center. After your tests finish, you can see the results including logs, videos and screenshots in the Firebase console.



You can enhance the workflow to automate taking screenshots through fastlane snapshot command and saves hours of valuable time you’ll burn taking screenshots. To upload the screenshots, metadata and the IPA file to iTunes Connect, you can use deliver command, which is already installed and initialized as part of fastlane.

The Android CI/CD workflow is quite straightforward, as it needs only the JDK environment with Android SDK preinstalled, we are running the CI on a Jenkins slave deployed into an EC2 Spot instance. The pipeline contains the following stages:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
def bucket = 'mobile-artifacts-foxintelligence'

node('android') {
try {
stage('Checkout') {
checkout scm
notifySlack('STARTED')
}

stage('Clean & Prepare') {
sh "./gradlew clean"
}

stage('Quality Tests') {
sh "./gradlew lintDebug"
androidLint pattern: 'app/build/reports/lint-results-debug.xml'
}

stage('Unit Tests') {
sh "./gradlew testDebug --stacktrace"

if (env.BRANCH_NAME == 'master'){
sh "./gradlew testReleaseUnitTest"
}
if (env.BRANCH_NAME == 'preprod'){
sh "./gradlew testStagingUnitTest"
}
if (env.BRANCH_NAME == 'develop'){
sh "./gradlew testSandboxUnitTest"
}

publishHTML([reportDir: 'app/build/reports/tests/testDebugUnitTest', reportFiles: 'index.html', reportName: 'Unit Tests Report'])

junit 'app/build/test-results/testDebugUnitTest/*.xml'
}

stage('Build'){
sh "./gradlew assembleDebug"

if (env.BRANCH_NAME == 'master'){
sh "./gradlew compileReleaseKotlin"
}
if (env.BRANCH_NAME == 'preprod'){
sh "./gradlew compileStagingKotlin"
}
if (env.BRANCH_NAME == 'develop'){
sh "./gradlew compileSandboxKotlin"
}
}

stage('Push'){
sh "aws s3 cp app/build/outputs/apk/debug/app-debug.apk s3://${bucket}/android/${commitID()}.apk"
}

stage('UI Tests'){
sh "./gradlew assembleDebugAndroidTest"
sh "gcloud firebase test android run --app app/build/outputs/apk/debug/app-debug.apk --test app/build/outputs/apk/androidTest/debug/app-debug-androidTest.apk"
}
}catch(e){
currentBuild.result = 'FAILED'
throw e
}finally{
notifySlack(currentBuild.result)
}
}

The pipeline could be drawn up as the following steps:

  • Check out the working branch from a remote repository.
  • Run the code through lint to find poorly structured code that might impact the reliability, efficiency and make the code harder to maintain. The linter will produces XML files which will be parsed by the Android Lint Plugin.
  • Launch Unit Tests. The JUnit plugin provides a publisher that consumes XML test reports generated and provides some graphical visualization of the historical test results as well as a web UI for viewing test reports, tracking failures, and so on.


  • Build debug or release APK based on the current Git branch name.
  • Upload the artifact to an S3 bucket.
  • Similarly, after the instrumentation tests have finished running, the Firebase web UI will then display the results of each test — in addition to information such as a video recording of the test run, the full Logcat, and screenshots taken:


To bring down testing time (and reduce the cost), we are testing Flank to split the test suite into multiple parts and execute them in parallel across multiple devices.

Our Continuous Integration workflow is sailing now. So far we’ve found that this process strikes the right balance. It automates the repetitive aspects, provides protection but is still lightweight and flexible. The last thing we want is the ability to ship at any time. We have an additional stage to upload the iOS artifact to Test Flight for distribution to our awesome beta tests.

Like what you’re read­ing? Check out my book and learn how to build, secure, deploy and manage production-ready Serverless applications in Golang with AWS Lambda.

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Why you should join Foxintelligence at the AWS Summit Paris 2019



Episode 5:Build a Docker Swarm Cluster on AWS

This is the first video in a 10 part series by Mohamed Labouardy, showing how to build a simple DevOps pipeline, including built-in security at each stage. To follow the series, please join the community to receive notification as each episode is released.



Episode 4:Manage a Secure Private Docker Registry with Sonatype Nexus and ACM

This is the first video in a 10 part series by Mohamed Labouardy, showing how to build a simple DevOps pipeline, including built-in security at each stage. To follow the series, please join the community to receive notification as each episode is released.



How to play PUBG on AWS

AWS GPU instances are known for deep learning purposes but they can also be used for running video games. This tutorial goes through how to set up your own EC2 GPU optimised instance to run the top-selling and most played game “PlayerUnknown’s Battlegrounds (PUBG)”.

To get started, make sure you are in the AWS region closest to you, select Microsoft Windows Server to be the AMI and set the instance type to be g2.2xlarge. The instance is backed by Nvidia Grid GPU (Kepler GK104), 8x hardware hyper-threads from an Intel Xeon E5–2670 and 15GB of RAM.



For games with resource-intensive, you should use the next generation of GPU instances: P2, P3 and G3 (have up to 4 NVIDIA Tesla M60 GPUs).

After this is done, click on “Launch Instances”, and you should see a screen showing that your instance is been created:



To connect to your Windows instance, you must retrieve the initial administrator password and specify this password when you connect to your instance using Remote Desktop:

Before you attempt to log in using Remote Desktop Connection, you must open port 3389 on the security group attached to your instance



After you connect, install Microsoft Direct X11 after installing Chrome (it saves a lot of time):



Next, install the graphic driver for maximum gaming performance:



Once installed, make sure to reboot the instance for changes to take effect:



Then, install Steam, login using your account and install PUBG from the “Library” section:



You can take advantage of AWS high network performance (up to 10 Gbps of bandwidth):



Once the game is installed, you can play PUBG on your virtualized GPU instance:



You can take this further, and use Steam In-Home Streaming feature to stream your game from your EC2 instance to your Mac:



Enjoy the game ! you can now play your games on any device connected to the same network:



You might want to bake an AMI based on your instance to avoid set it up all again the next time you want to play and use spot instances to reduce the instance cost. Also, make sure to stop your instances when you’re done for the day to avoid incurring charges. GPU instances are costly (disk storage also costs something, and can be significant if you have a large disk footprint).

Drop your comments, feedback, or suggestions below — or connect with me directly on Twitter @mlabouardy.

Episode 3:Deploy a Highly Available Jenkins Cluster on AWS

This is the first video in a 10 part series by Mohamed Labouardy, showing how to build a simple DevOps pipeline, including built-in security at each stage. To follow the series, please join the community to receive notification as each episode is released.



Episode 2:Build an AWS VPC using Infrastructure as Code

This is the first video in a 10 part series by Mohamed Labouardy, showing how to build a simple DevOps pipeline, including built-in security at each stage. To follow the series, please join the community to receive notification as each episode is released.



Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×