Serverless Architecture - A Gentle Overview

Serverless Architecture: Overview
Ganesh Samarthyam
ganesh@codeops.tech
Organizer: India Serverless Summit
www.inserverless.com

Are you dealing with web-servers, app-
servers, session and entity beans, …: in
short, are you a stone-age programmer?

Greek characters
are scary!
The term “lambda” originates
from the “lambda
calculus” (yes, yes, that same
scary math stuff!)

He he, but lambdas
are fun, not scary

def main(args):
name = args.get("name", "stranger")
greeting = "Hello " + name + "!"
print(greeting)
return {"greeting": greeting}
Lambda
functions!
import com.amazonaws.services.lambda.runtime.Context;
public class helloworld {
public String handleRequest(String name, Context context) {
String greetings = "Hello " + name;
context.getLogger().log("Name: " + name);
System.out.println(greetings);
return greetings;
}
}
function main() {
return {payload: 'Hello world'};
}
Python
Java
JavaScript (Node.JS)

Lambdas is just a fancy
name for stateless
functions without side-
effects!

Evolution: Servers to serverless
Code
Runtime
OS
Hardware
Code
Runtime
OS
Hardware
Code
Runtime
OS
Hardware
Code
Runtime
OS
Hardware
Physical
Servers
Virtual
Machines
Containers Serverless

Evolution: Servers to serverless
Slow-iteration and
deployment
Faster-iteration and
deployment
Fastest-iteration
and deployment
Rapid iteration and
deployment
Single tenency Multi-tenency
Super multi-
tenancy
Extreme multi-
tenancy
Unfriendly for
polyglots
Somewhat friendly
for polyglots
Friendly for
polyglots
Very friendly for
polyglots
Deploy in weeks Deploy in minutes Deploy in seconds
Deploy
independently
Typically alive for
years
Typically alive for
weeks
Typically alive for
hours
Typically alive for
seconds
Physical
Servers
Virtual
Machines
Containers Serverless
Source: https://www.slideshare.net/AmazonWebServices/getting-started-with-serverless-architectures-63429092

What is
server-“lesssss…"?

What is serverless?
“Serverless architectures refer to applications that
signiﬁcantly depend on third-party services (knows
as Backend as a Service or "BaaS") or on custom code
that's run in ephemeral containers (Function as a
Service or “FaaS”)”
- Martin Fowler

Serverless Architecture - A Gentle Overview

– Serverless for Dummies
“Serverless: just put your code in cloud and run it”

What is serverless?
❖ Serverless is often referred to as FaaS - Function as a
Service. Gartner refers to it as fPaaS - function Platform
as a Service.
❖ The model in serverless architecture is this: a distributed
system that reacts to events or process workloads
dynamically based on demand by spinning up
ephemeral (short-lived) containers or computational
resources in the cloud.

On the term “lambda”
❖ The term “lambda” originates from the
“lambda calculus”. In 1936, Alonzo Church
developed a logic system that was later
adopted for computation.
❖ Most languages (including Java, C# and C++)
support lambda functions today.
❖ In languages supporting lambda functions,
it is an unnamed function that takes input
coming variables and returns a value.
❖ A salient characteristic of lambda functions
is the lack of side-effects.

– Doug McIlroy
“Write programs that do one thing and do it well.
Write programs to work together. Write programs
to handle text streams, because that is a universal
interface.”

But, but, “Serverless != No servers”
Lambdas are executed in servers in cloud
- so we still have servers!
What serverless means is you don't have to
care about servers: no provisioning,
maintenance, etc.

Why serverless?
– Google Cloud Functions
https://cloud.google.com/functions/
“Serve users from zero to
planet-scale, all without
managing any
infrastructure.”

– Gartner
"By 2022, most platform as a service (PaaS)
offerings will evolve to a fundamentally serverless
model, rendering the cloud platform architectures
dominating in 2017 as legacy architectures"

Serverless - characteristics
granular payment event-driven
self-managed scale on demand

Serverless - some use cases
mobile back-ends Real-time analytics
data processing static-websites
live-video stream
processing
chatbots

Data processing
source: https://www.slideshare.net/AmazonWebServices/aws-reinvent-2016-getting-started-with-serverless-architectures-cmp211

Static websites

Mobile backends

Real-time analytics

Live video-steam processing

Use case #1: web applications
source: https://www.slideshare.net/AmazonWebServices/aws-reinvent-2016-serverless-architectural-patterns-and-best-practices-arc402

Use case #2: batch processing

Use case #3: stream processing

Lambda architecture style

Towards LessOps model?
Serverless frameworks move us towards LessOps model - meaning no or
few operators are required for monitoring and maintaining the servers and
infrastructure.

– Dr Werner Vogels (Amazon CTO)
"What we’ve seen is a
revolution where complete
applications are being stripped
of all their servers, and only
code is being run. Quite a few
companies are ripping out big
pieces of their applications and
replacing their servers, their
VMs and their containers with
just code… Perhaps we no longer
have to think about servers!”

Popular technologies
AWS Lambda
Google Cloud
Functions
IBM/Adobe/
Apache OpenWhisk
Azure Functions

Five principles for serverless applications
Use a compute service to execute code on demand (no
servers)
Write single-purpose stateless functions
Design push-based, event-driven pipelines
Create thicker, more powerful front ends
Embrace third-party services
Source: Peter Sbarski, “Serverless Architectures on AWS”, Manning, 2017

AWS Lambda
From https://aws.amazon.com/

AWS Lambda
❖ Lambda is one of the earliest technologies for serverless
(13 Nov 2014). It is also the most mature and stable
platform.
❖ http://aws.amazon.com/lambda
❖ Currently supported (as on June 2017) - .NET Core 1.01.
(C#), Java 8, Node.js 4.3, Node.js 6.10.2, Python 2.7 and
3.6.

Three ways to create a function
❖ There are three ways to create the same lambda
function:
❖ AWS Lambda console (web UI)
❖ AWS CLI (from your commandline)
❖ AWS SDK

Execution Environment
❖ Public Amazon Linux AMI version (AMI name: amzn-ami-
hvm-2016.03.3.x86_64-gp2)
❖ Linux kernel version – 4.4.51-40.60.amzn1.x86_64
❖ When creating native (64-bit) binaries for executing in Lambda, this
information is important - so that code can be compiled to this target
❖ The following libraries are available in execution environment
❖ AWS SDK – AWS SDK for JavaScript version 2.45.0
❖ AWS SDK for Python (Boto 3) version 1.4.4, Botocore version 1.5.43
❖ Amazon Linux build of java-1.8.0-openjdk for Java
Source: http://docs.aws.amazon.com/lambda/latest/dg/current-supported-versions.html

A simple lambda
$ cat hello.py
def lambda_handler(event, context):
print "hello srushit"
return 'Hello from srushit'
$ zip -r hello-world hello.py
$ aws lambda create-function --region us-east-1 --function-name
FirstPythonLambda --zip-ﬁle ﬁleb:///Users/gsamarthyam/hello-world.zip --
role arn:aws:iam::431635030606:role/LambdaExecuteRole --runtime
python2.7 --handler hello::lambda_handler
Notes:
(a) Once you upload the function from command-line, it takes time to see the
updated source code from the AWS console online.
(b) In Python, print statements and Logger functions in the logging module are
logged - you can check them in CloudWatch logs.

A hello world lambda in Java
$ cat Hello.java
public class Hello {
public String handleRequest(String name, Context context) {
return "Hello " + name;
}
}
http://docs.aws.amazon.com/lambda/latest/dg/java-programming-model.html

A simple lambda
hello::handleRequest
Input string
“Ganesh”
Context object
Result string
“Hello Ganesh”
AWS Lambda
Execution on a
Linux container

Step #1
$ javac -cp ./lib/aws-lambda-java-core-1.0.0.jar:./lib/aws-lambda-java-events-1.0.0.jar
Hello.java
$
Need to provide the dependencies (Context
class in this case) to compile it; dependency
details available from: https://github.com/
aws/aws-lambda-java-libs
(“Ofﬁcial mirror for interface deﬁnitions and
helper classes for Java code running on the AWS
Lambda platform.”)

Step #2
$ zip -r Hello.zip Hello.class lib/aws-lambda-java-core-1.0.0.jar lib/aws-lambda-java-
events-1.0.0.jar
adding: Hello.class (deflated 35%)
adding: lib/aws-lambda-java-core-1.0.0.jar (deflated 39%)
adding: lib/aws-lambda-java-events-1.0.0.jar (deflated 27%)
$ ls -lh Hello.zip
-rw-r--r-- 1 gsamarthyam staff 14K Jun 19 15:01 Hello.zip
$ ls -lh lib
-rw-r--r-- 1 gsamarthyam staff 29K Jun 23 16:12 aws-lambda-java-core-1.0.0.jar
-rw-r--r--@ 1 gsamarthyam staff 11K Jun 18 16:44 aws-lambda-java-events-1.0.0.jar
Create a zip file with the dependencies - in this
case, it is just a couple of standard dependencies

Step #3
$ aws lambda create-function --region us-east-1 --function-name FirstJavaLambda --zip-file fileb:///Users/
gsamarthyam/serverless/Hello.zip --role arn:aws:iam::431635030606:role/LambdaExecuteRole --handler
Hello::handleRequest --runtime java8
{
"TracingConfig": {
"Mode": "PassThrough"
},
"CodeSha256": "DpRB5zFF1ORVw6tTSIMprA1BgGNmpf+oGBFBy8/APWo=",
"FunctionName": "FirstJavaLambda",
"CodeSize": 13988,
"MemorySize": 128,
"FunctionArn": "arn:aws:lambda:us-east-1:431635030606:function:FirstJavaLambda",
"Version": "$LATEST",
"Role": "arn:aws:iam::431635030606:role/LambdaExecuteRole",
"Timeout": 3,
"LastModified": "2017-06-23T10:50:02.582+0000",
"Handler": "Hello::handleRequest",
"Runtime": "java8",
"Description": ""
}

Step #4
$ aws lambda invoke --invocation-type RequestResponse --function-name FirstJavaLambda
--region us-west-2 --log-type Tail --payload '"Ganesh"' outputﬁle.txt
{
"LogResult":
"U1RBUlQgUmVxdWVzdElkOiBmYzM2N2I2NC01ODAxLTExZTctYjllOS05OTQzNmU4ODEyZmY
gVmVyc2lvbjogJExBVEVTVApsb2cgZGF0YSBmcm9tIHN0ZG91dCAKIHRoaXMgaXMgY29udGl
udWF0aW9uIG9mIHN5c3RlbS5vdXQKbG9nIGRhdGEgZnJvbSBzdGRlcnIgCiB0aGlzIGlzIGNvbn
RpbnVhdGlvbiBvZiBzeXN0ZW0uZXJyCmxvZyBkYXRhIGZyb20gTGFtYmRhTG9nZ2VyIAogdGh
pcyBpcyBjb250aW51YXRpb24gb2YgbG9nZ2VyLmxvZ0VORCBSZXF1ZXN0SWQ6IGZjMzY3YjY
0LTU4MDEtMTFlNy1iOWU5LTk5NDM2ZTg4MTJmZgpSRVBPUlQgUmVxdWVzdElkOiBmYzM2
N2I2NC01ODAxLTExZTctYjllOS05OTQzNmU4ODEyZmYJRHVyYXRpb246IDI2MC45MyBtcwlC
aWxsZWQgRHVyYXRpb246IDMwMCBtcyAJTWVtb3J5IFNpemU6IDEyOCBNQglNYXggTWVtb3
J5IFVzZWQ6IDQwIE1CCQo=",
"StatusCode": 200
}
$ cat outputﬁle.txt
"Hello Ganesh. log stream = 2017/06/23/[$LATEST]680db95163b346579ce89f51535db346"

Step #4
$ echo
"U1RBUlQgUmVxdWVzdElkOiBmYzM2N2I2NC01ODAxLTExZTctYjllOS05OTQzNmU4ODEyZm
YgVmVyc2lvbjogJExBVEVTVApsb2cgZGF0YSBmcm9tIHN0ZG91dCAKIHRoaXMgaXMgY29ud
GludWF0aW9uIG9mIHN5c3RlbS5vdXQKbG9nIGRhdGEgZnJvbSBzdGRlcnIgCiB0aGlzIGlzIGNv
bnRpbnVhdGlvbiBvZiBzeXN0ZW0uZXJyCmxvZyBkYXRhIGZyb20gTGFtYmRhTG9nZ2VyIAogd
GhpcyBpcyBjb250aW51YXRpb24gb2YgbG9nZ2VyLmxvZ0VORCBSZXF1ZXN0SWQ6IGZjMzY3Y
jY0LTU4MDEtMTFlNy1iOWU5LTk5NDM2ZTg4MTJmZgpSRVBPUlQgUmVxdWVzdElkOiBmY
zM2N2I2NC01ODAxLTExZTctYjllOS05OTQzNmU4ODEyZmYJRHVyYXRpb246IDI2MC45MyBt
cwlCaWxsZWQgRHVyYXRpb246IDMwMCBtcyAJTWVtb3J5IFNpemU6IDEyOCBNQglNYXggT
WVtb3J5IFVzZWQ6IDQwIE1CCQo=" | base64 --decode
START RequestId: fc367b64-5801-11e7-b9e9-99436e8812ff Version: $LATEST
log data from stdout
this is continuation of system.out
log data from stderr
this is continuation of system.err
log data from LambdaLogger
this is continuation of logger.logEND RequestId: fc367b64-5801-11e7-b9e9-99436e8812ff
REPORT RequestId: fc367b64-5801-11e7-b9e9-99436e8812ff Duration: 260.93 ms Billed
Duration: 300 ms Memory Size: 128 MB Max Memory Used: 40 MB

Step #4
$ aws lambda get-function --function-name FirstJavaLambda
{
"Code": {
"RepositoryType": "S3",
"Location": "https://prod-04-2014-tasks.s3.amazonaws.com/snapshots/431635030606/FirstJavaLambda-961856c8-8808-4d77-aa26-
bde4c62745d9?X-Amz-Security-
Token=FQoDYXdzENP%2F%2F%2F%2F%2F%2F%2F%2F%2F%2FwEaDJIA1ZghlrarR%2FS8sCK3A9ic4dl6r2QeYmybaLJ%2FR4VxX3e%2BN8413pI
vMLfpND96fpnK5g55zaaI4nQkYcW70zwZu9mZ3%2FXrQqObP1PKpY1BBr0uP7BrDn8wF062XuBQG4AMaoxBQ44cSvUC8cr%2FGVU2MpAzFEK0
0kgOtI9fax%2FB6eAt5xzKN2Zt4SpCtFktfaNTq15Vxm2RIk8s30Yw53jte7fPPtaUZPJZT4%2F872zNrylxSfKAfRdOvrv4%2FvGur21evlliheJ0mriQZ
NBcRkRsNWIC6MkG48inkiRL4AjWN9BNWXl5JhpPxgHfZTEnb2st1SJuuJdx3SyWs2u8Wh8J%2BV90eyuZSqDeXXRQCmk4QNhSalvLboN6M9oTC
VBh7uMdsLDkxje3FOlRdDYMJW9kcyMc%2BcqUOeeeO46m7ky0t%2BKtvDrY887jjziG3PTQjZrFPwmiEHgVImzhfa9RN8eAXb3xcnPqC97rzOM1GZ
kAD%2BuJbECu%2B1t7H93rhG5ObQZVBT4llxBq%2BQXpI0FHjqpVE7wOs7hEuqws65gXxVOltBVHv7TwsTIxibH88uai%2BpyRV%2BC6NUK%2FpL
QZDF8wCqozh%2B%2FHvmAo2smzygU%3D&X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Date=20170623T110025Z&X-Amz-
SignedHeaders=host&X-Amz-Expires=600&X-Amz-Credential=ASIAIHFSB5YQLFLCC34Q%2F20170623%2Fus-east-1%2Fs3%2Faws4_request&X-
Amz-Signature=af988a347e62ebf6fe89d3593b68ada63be6bcd74f6b12222b96b762c7204924"
},
"Configuration": {
"TracingConfig": {
},
"MemorySize": 128,
"CodeSize": 13988,
"Timeout": 3,
"LastModified": "2017-06-23T10:50:02.582+0000",
"Runtime": "java8",
"Description": ""
}
}

How to update the function?
$ aws lambda update-function-code --function-name arn:aws:lambda:us-
east-1:431635030606:function:FirstJavaLambda --zip-file fileb://Hello.zip
{
"TracingConfig": {
},
"CodeSize": 13988,
"MemorySize": 128,
"Timeout": 3,
"LastModified": "2017-06-23T13:34:05.408+0000",
"Runtime": "java8",
"Description": ""
}

How to delete the function?
$ aws lambda delete-function --function-name FirstJavaLambda
$ aws lambda list-functions | grep FirstJavaLambda
$
You can check that once the lambda is deleted,
the list-functions doesn’t list it anymore

Available commands in “aws lambda”
• add-permission
• create-alias
• create-event-source-mapping
• create-function
• delete-alias
• delete-event-source-mapping
• delete-function
• get-account-settings
• get-alias
• get-event-source-mapping
• get-function
• get-function-conﬁguration
• get-policy
• help
• invoke
• invoke-async
• list-aliases
• list-event-source-mappings
• list-functions
• list-tags
• list-versions-by-function
• publish-version
• remove-permission
• tag-resource
• untag-resource
• update-alias
• update-event-source-mapping
• update-function-code
• update-function-conﬁguration

Using AWS SDK to create lambdas
$ cat testing.py
def lambda_handler(event, context):
return 'Hello from Lambda'
$ cat invokelambda.py
import boto3
import json
client = boto3.client('lambda')
response = client.invoke(
FunctionName='arn:aws:lambda:us-east-1:431635030606:function:testing',
InvocationType='RequestResponse',
LogType='Tail',
Payload=json.dumps({})
)
result = response['Payload'].read()
print(result)
$ python invokelambda.py
"Hello from Lambda"
$

Kinds of lambda invocations
• Event based lambdas
are invoked based on
triggers; example: when
a ﬁle is dropped in a S3
bucket, a lambda may
be invoked
• Async in nature
• No response sent back
to the invoker
Event RequestResponse
• Synchronous - when
invoked, it executes and
waits till completion
and returns something
to the invoker
• Happens when invoked
through API gateway,
CLI or AWS console

Specifying invocation types in CLI
Help description for —invocation-type option
for “aws lambda invoke” from AWS CLI

Using base64
❖ Sometimes it is convenient to use base64 whenever you
want to send data to the lambda or return from the
lambda, or get that data again as context. this avoids
adding unicode characters etc.
❖ Logs in lambdas often use base64. You can use the
“base64 --decode” command (mac/linux) to read the
log ﬁle

Using base64 - example
$ cat logresult.txt
{
"LogResult":
"U1RBUlQgUmVxdWVzdElkOiAyZGRmMjQ3Ni01ODE3LTExZTctYWI4YS02ZGQwZWMxYTFlYmMgVmVyc2lvbjogJExBVEVTVApsb2c
gZGF0YSBmcm9tIHN0ZG91dCAKIHRoaXMgaXMgY29udGludWF0aW9uIG9mIHN5c3RlbS5vdXQKbG9nIGRhdGEgZnJvbSBzdGRlcnI
gCiB0aGlzIGlzIGNvbnRpbnVhdGlvbiBvZiBzeXN0ZW0uZXJyCmxvZyBkYXRhIGZyb20gTGFtYmRhTG9nZ2VyIAogdGhpcyBpcyB
jb250aW51YXRpb24gb2YgbG9nZ2VyLmxvZ0VORCBSZXF1ZXN0SWQ6IDJkZGYyNDc2LTU4MTctMTFlNy1hYjhhLTZkZDBlYzFhMWV
iYwpSRVBPUlQgUmVxdWVzdElkOiAyZGRmMjQ3Ni01ODE3LTExZTctYWI4YS02ZGQwZWMxYTFlYmMJRHVyYXRpb246IDUuMDIgbXM
JQmlsbGVkIER1cmF0aW9uOiAxMDAgbXMgCU1lbW9yeSBTaXplOiAxMjggTUIJTWF4IE1lbW9yeSBVc2VkOiA0NiBNQgkK",
"StatusCode": 200
}
$ echo
"U1RBUlQgUmVxdWVzdElkOiAyZGRmMjQ3Ni01ODE3LTExZTctYWI4YS02ZGQwZWMxYTFlYmMgVmVyc2lvbjogJExBVEVTVApsb2c
gZGF0YSBmcm9tIHN0ZG91dCAKIHRoaXMgaXMgY29udGludWF0aW9uIG9mIHN5c3RlbS5vdXQKbG9nIGRhdGEgZnJvbSBzdGRlcnI
gCiB0aGlzIGlzIGNvbnRpbnVhdGlvbiBvZiBzeXN0ZW0uZXJyCmxvZyBkYXRhIGZyb20gTGFtYmRhTG9nZ2VyIAogdGhpcyBpcyB
jb250aW51YXRpb24gb2YgbG9nZ2VyLmxvZ0VORCBSZXF1ZXN0SWQ6IDJkZGYyNDc2LTU4MTctMTFlNy1hYjhhLTZkZDBlYzFhMWV
iYwpSRVBPUlQgUmVxdWVzdElkOiAyZGRmMjQ3Ni01ODE3LTExZTctYWI4YS02ZGQwZWMxYTFlYmMJRHVyYXRpb246IDUuMDIgbXM
JQmlsbGVkIER1cmF0aW9uOiAxMDAgbXMgCU1lbW9yeSBTaXplOiAxMjggTUIJTWF4IE1lbW9yeSBVc2VkOiA0NiBNQgkK" |
base64 --decode
START RequestId: 2ddf2476-5817-11e7-ab8a-6dd0ec1a1ebc Version: $LATEST
log data from stdout
this is continuation of system.out
log data from stderr
this is continuation of system.err
log data from LambdaLogger
this is continuation of logger.logEND RequestId: 2ddf2476-5817-11e7-ab8a-6dd0ec1a1ebc
REPORT RequestId: 2ddf2476-5817-11e7-ab8a-6dd0ec1a1ebcDuration: 5.02 msBilled Duration: 100 ms
Memory Size: 128 MB Max Memory Used: 46 MB
$

Understanding the “Context” object
$ cat ContextInfo.java
public class ContextInfo {
public Context details(Object input, Context context) {
System.out.println("Memory limit in MBs: " + context.getMemoryLimitInMB());
System.out.println("Function name: " + context.getFunctionName());
System.out.println("AWS Request ID: " + context.getAwsRequestId());
System.out.println("Log Stream Name: " + context.getLogStreamName());
System.out.println("Log Group Name: " + context.getLogGroupName());
System.out.println("Client Context: " + context.getClientContext());
System.out.println("Identity: " + context.getIdentity());
System.out.println("Remaining time (in milli seconds): "+
context.getRemainingTimeInMillis());
return context;
}
}

Compiling and running the lambda
$ cat lamdafy.sh
#!/bin/bash
set -x #echo on
rm logresult.txt
rm outputfile.txt
javac -cp ./lib/aws-lambda-java-core-1.0.0.jar:./lib/aws-lambda-java-events-1.0.0.jar ContextInfo.java
zip -r ContextInfo.zip lib/aws-lambda-java-core-1.0.0.jar lib/aws-lambda-java-events-1.0.0.jar ContextInfo.class
aws s3 cp ContextInfo.zip s3://myplasticbucketforlambda
aws lambda update-function-code --function-name ContextInfoLambda --s3-bucket myplasticbucketforlambda --s3-key
ContextInfo.zip
aws lambda invoke --invocation-type RequestResponse --function-name ContextInfoLambda --region us-east-1 --log-type
Tail --payload file://inputfile.txt outputfile.txt >> logresult.txt
jq -r .LogResult logresult.txt | base64 --decode
cat outputfile.txt
echo ""
jq . outputfile.txt

Memory limit in MBs: 128
Function name: ContextInfoLambda
AWS Request ID: 32320904-5805-11e7-8904-ebe1abfa66b8
Log Stream Name: 2017/06/23/[$LATEST]acf60f48492b4df59ef8ebe534fffe00
Log Group Name: /aws/lambda/ContextInfoLambda
Client Context: null
Identity: lambdainternal.api.LambdaCognitoIdentity@589838eb
Remaining time (in milli seconds): 2778

getFunctionName()
Name of the lambda function being executed (e.g.,
FirstJavaLambda)
getFunctionVersion() Version of the function (or its alias if it points to an alias)
getInvokedFunctionArn()
Get the ARN (Amazon Resource Name), for example,
arn:aws:lambda:us-west-2:431635030606:function:HelloWorld
getLogStreamName() Returns the CloudWatch log stream name
getLogger() Returns the logger associated with the given context object
getMemoryLimitInMB()
Amount of memory set as limit when creating the lambda
function (value returned in MBs)
getRemainingTimeInMillis()
Returns the amount of time remaining for execution (based on the
time limit set during creating the function)

Watching CloudWatch logs
To watch with CloudWatch, create a new role with following accesses:
arn:aws:iam::aws:policy/AWSLambdaExecute
arn:aws:iam::aws:policy/AWSLambdaFullAccess
Streams stdout and stderr are redirected to
CloudWatch logs; can be viewed from AWS console

Using the /tmp directory
❖ Applications can use “/tmp” directory
❖ If the underlying container gets recycled, the ﬁles in the
“/tmp” directory will be gone!
❖ Also, /tmp is limited to 500 MB size - that is a limiting
factor for many kinds of applications.
❖ For example, if we want to convert a video from one
format to another, then this limit matters.

Dealing with code dependencies
❖ External libraries and dependencies must be bundled with the
lambda function as zip ﬁles (or jar ﬁles depending on the context)
❖ There is limitation to the size of the dependencies that we can
upload, currently it is 30MB.
❖ That is quite small for real-world requirement where the
dependent libraries can be large.
❖ A workaround for this limitation is to upload it into S3 bucket,
put it in to /tmp directory for the lambda and then use it.

Using lambdas with AWS services
❖ For real-world uses, its practically not feasible to use lambda without
the AWS eco-system:
❖ S3 can be used for processing images, converting document formats,
etc from the lambdas. Large dependencies for the lambda can be
stored in S3 and processed from the lambdas
❖ SNS (Simple Notiﬁcation Service) can be used for event processing
❖ DynamoDB can be used for persistence and data processing
❖ Alexa can be used for automated voice responses with Amazon Echo
for example
❖ Lex can be used for chatbots with lambda as the back-end code

How does a lambda terminate?
❖ How does a lambda terminate?
(A) Timeout - once the given maximum duration is
reached, the function will terminate
(B) Normal termination - returning from the lambda -
when the function returns, the lambda terminates
(C)Abnormal termination - forcefully terminating the
lambda with calling functions (e.g. process.exit() in
nodejs)

Lambda resource limits
Source: http://docs.aws.amazon.com/lambda/latest/dg/limits.html

Lambda deployment size limits
Source: http://docs.aws.amazon.com/lambda/latest/dg/limits.html

AWS Serverless Stack
❖ API Gateway
❖ Lambda
❖ Greengrass
❖ Step Functions
❖ SQS
❖ DynamoDB
❖ …

Costing aspects
❖ There are two factors for pricing calculation
1. Request costs: refers to number of executions - it is
calculated at 100ms slices
2. Execution costs: these are variable costs; change the memory
size and the execution cost will be calculated accordingly
❖ There is no cost for idle time
❖ You can use the online pricing calculator to estimate how much it
may cost you: https://s3.amazonaws.com/lambda-tools/
pricing-calculator.html

Costing aspects
❖ In AWS lambda, the increase in the memory is associated
with increased CPU and network I/O speeds.
❖ In free tier, it makes sense to increase it to the max MBs.
❖ Note that the cost is not based on actual consumption,
but on the max. cap.
❖ Lots of calls execute only for short duration, like 30 ms.
However billing is done for 100ms minimum. Hence, one
way is to design applications that make fewer but longer
calls.

Serverless != Always Lower Costs!
❖ It is a myth that adopting serverless will always reduce
your costs. It is NOT true. The key in cost savings is in
how you utilise serverless. For example, serverless is
suitable if you have spiky workloads (or idle scenarios).
It is NOT suitable in scenarios where your demand is
stable.

Internally, it uses containers!
❖ Internally, AWS Lambda uses Linux Containers
❖ When a lambda function is invoked, it launches a
container (with the provided conﬁg settings)
❖ When a function is called for the ﬁrst time (or after it is
updated), spawning the container adds some latency
❖ Container is terminated and is “frozen” for reuse
❖ If called again “soon”, it is “thawed” and invoked
again
Based on: http://docs.aws.amazon.com/lambda/latest/dg/lambda-introduction.html

Freeze/Thaw cycle
freeze/thaw cycle - how the execution gets frozen up and is warmed up again

Freeze/Thaw cycle
freeze/thaw cycle - how the execution gets frozen up and is warmed up again
Execution
Terminate
Freeze
Thaw
Cold start

How AWS Lambda works
❖ Handlers are executed on workers. Time for starting-up of
a worker is around 1 to 2 seconds.
❖ In case of a “load burst”, the Lambda load balancer sends
them to new workers - that will incur latencies.
❖ For a container to execute the handler immediately, it must
be in running state.
❖ However, if handlers are called only once in a while,
then the container goes to paused or stopped state and it
takes time to get back to running state again.

Cold vs. warm invocation
❖ Cold invocation - requires warm-up time (e.g., when
starting for the ﬁrst time) - the latency may be
perceivable to the users
❖ Actions like code redeployment or making
conﬁguration changes can force redeployment of the
lambda that make it cold invocation
❖ Warm invocation - lambda already deployed and ready
to serve as soon as it is triggered or called

How to keep your container “warm”?
❖ You can use a “warming trigger” to
ensure that the lambda function
doesn’t go cold. Trigger it within
every few minutes.
❖ Schedule CloudWatch events to
ﬁre every few minutes to keep the
container “warm”
❖ In serverless framework, there is a
plugin that keeps the lambda
warm by sending an event every
5 minutes or so: https://
github.com/FidelLimited/
serverless-plugin-warmup

Best Practice
Don’t depend on container reuse, but make best use of it
Example #1: copy ﬁles from s3 and use it from /tmp, but check if the ﬁles
in /tmp exist before trying to access them!
Example #2: Instantiate AWS client, establish database connects, etc outside
the handler - they will not be initialised again if container is reused

How about other languages?
❖ Lambda currently supports limited set of languages
(JavaScript - with Node.js, Python, Java and C#)
❖ What if you want to use some other language, say Go?
❖ You can still run Go by following a few steps:
❖ http://blog.0x82.com/2014/11/24/aws-lambda-
functions-in-go/
❖ https://serifandsemaphore.io/go-amazon-
lambda-7e95a147cec8

Security
❖ IAM (Identity & Access Management) provides the security for your
lambdas
❖ Speciﬁc accesses are often required; for example, to watch lambda
logs with CloudWatch, create a new role with following accesses:
❖ arn:aws:iam::aws:policy/AWSLambdaExecute
❖ arn:aws:iam::aws:policy/AWSLambdaFullAccess
❖ Due to security and legal aspects, serverless functions may need to
run privately. In this case, lambda functions can be executed on AWS
VPC (Virtual Private Cloud) inside a private network.

Using “aws iam” - example
$ aws iam list-roles | jq -r .Roles[].Arn
arn:aws:iam::431635030606:role/helloworld-dev
arn:aws:iam::431635030606:role/service-role/javatesting
arn:aws:iam::431635030606:role/LambdaExecuteRole
$ aws iam list-roles | jq -r .Roles[].Arn | grep ExecuteRole
arn:aws:iam::431635030606:role/LambdaExecuteRole

What’s not suitable
❖ Extremely long running jobs
❖ Jobs that require extensive communication between
each other
❖ Constant loads and well-known costs
❖ Responses with very low-latency

Rearchitecting to serverless functions
❖ Moving to serverless is not as
easy as ﬂipping a switch: for
existing applications, it may
even require re-architecting/re-
writing.
❖ This is because serverless
computation architecture
model is quite different from
the way traditionally
software is written and
deployed.

Costing in Google Cloud Functions
* Includes both Background and HTTP Functions.
Source: https://cloud.google.com/functions/

So, what’s the catch?
❖ Runtime and resource limitations
(memory, dependency sizes)
❖ Vendor lock-in (because non-
trivial lambdas will use services
from providers)
❖ Lambdas are still low-level
“primitives”
❖ Tooling yet to mature (for testing,
deploying, monitoring, …)
❖ …

Serverless looks awesome - let’s explore!

Links for serverless technologies
❖ AWS Lambda: https://aws.amazon.com/lambda/
❖ Azure Functions: https://functions.azure.com/
❖ Google Cloud Functions: https://cloud.google.com/functions/
❖ Apache OpenWhisk: https://github.com/openwhisk
❖ Serverless framework: https://github.com/serverless/serverless
❖ Fission: https://github.com/ﬁssion/ﬁssion
❖ Hyper.sh: https://github.com/hyperhq /
❖ Funktion: https://funktion.fabric8.io/
❖ Kubeless: http://kubeless.io/

Tools to simplify using AWS Lambda
❖ Most frameworks available today in serverless domain try to simplify
AWS! Here is a partial list (all of the for AWS):
❖ Chalice (Python)
❖ ClaudiaJS (NodeJS)
❖ Dawson (Node)
❖ Lambada Framework (Java)
❖ Shep (Node)
❖ Sparta (Go)
❖ Zappa (Python)

Node Lambda (for Node.js)
❖ This is a “command line tool to locally run and deploy
your node.js application to Amazon Lambda”
❖ Why? Instead of testing your Lambda in AWS, you can
write it locally and run it and test it; when everything is
good, then deploy it in AWS Lambda
❖ URL: https://github.com/motdotla/node-lambda

❖ “The Standard Library for Functions as a Service.
Discover pre-built APIs, compose your own, build apps,
and move your business faster than ever with new
"server-less" technology.”
❖ Aims to be the “standard library for the web”!
❖ Check out https://stdlib.com/ and https://
github.com/stdlib/lib

Book Recommendations
❖ Has many interesting examples
❖ The supporting website has
demos that you can try out:
https://eventdrivenapps.com/
❖ URL: https://
www.manning.com/books/
aws-lambda-in-action

❖ Focus on & examples using AWS
Lambda
❖ Covers serverless patterns and
architectures
❖ Written by Dr. Peter Sbarski, head
of Serverlessconf
❖ Serverless Architectures on AWS,
Peter Sbarski, Manning
Publications, 376 pages, April 2017.
URL: https://www.amazon.com/
Serverless-Architectures-AWS-
examples-Lambda/dp/1617293822

❖ Serverless: Patterns of Modern
Application Design Using
Microservices (Amazon Web
Services Edition), Obie
Fernandez, LeanPub, 2016.
https://leanpub.com/
serverless

❖ Comprehensive - free - guide
on AWS Lambda is available
here: http://
docs.aws.amazon.com/
lambda/latest/dg/lambda-
dg.pdf

Links
❖ M. Fowler’s early article on serverless http://
martinfowler.com/articles/serverless.html
❖ AWS Lambda local runner: https://github.com/
cagataygurturk/aws-lambda-local-runner
❖ Java on AWS Lambda: https://www.infoq.com/news/
2017/06/fearless-aws-lambda
❖ The twelve-factor app: https://12factor.net/

Video Resources
❖ Serverless Architectures: What, why, why not, and
where next? - Mike Roberts
❖ Serverless: the Future of Software Architecture - Peter
Sbarski
❖ Serverless computing options with Google Cloud
Platform - Bret McGowen

Serverless Summit
India’s first conference on
serverless technologies!

Image Credits
❖ http://www.nag.co.za/wp-content/uploads/2013/04/Stone-Age-PC.jpg
❖ http://cdn2.hubspot.net/hub/148740/file-1853309085-jpg/RAM_-_stoneage.jpg
❖ https://upload.wikimedia.org/wikipedia/commons/thumb/9/97/WernerVogels.JPG/220px-WernerVogels.JPG
❖ https://image.slidesharecdn.com/devopssummit-openkeynote-noops-141105113330-conversion-gate01/95/noops-no-
operations-2-638.jpg?cb=1420567266
❖ http://f4.bcbits.com/img/a2377654591_10.jpg
❖ https://cdn.meme.am/cache/instances/folder981/500x/75497981/roll-safee-servers-cant-crash-if-there-are-no-servers.jpg
❖ https://www.commitstrip.com/wp-content/uploads/2017/04/Strip-Severless-650-finalenglishV2.jpg
❖ https://www.trovaweb.net/images/pizzeria-tavola-calda-barbecue-self-service-free-e-mangia-palermo/pizzeria-tavola-calda-barbecue-
self-service-free-e-mangia-palermo-11.JPG
❖ http://2ndwatch.com/wp-content/uploads/2016/05/AWS-Lambda.png
❖ http://wendyonline.nl/wp-content/uploads/2015/11/eekhoorn-nootjes-1024x656.jpg
❖ https://innovationcentre.com.au/wp-content/uploads/2017/03/michievent.jpg
❖ http://orig00.deviantart.net/da2f/f/2013/034/f/4/alonzo_and_lambda_by_kinow-d5tqvau.png
❖ https://travistation.files.wordpress.com/2012/04/david-hasselhoff-as-michael-knight-in-knightrider-thumbs-up-263w_350h.jpg
❖ http://www.otona-beauty.com/wp-content/uploads/2015/08/150803_06.jpg
❖ https://s-media-cache-ak0.pinimg.com/736x/d7/7c/f7/d77cf70f4e8b56a834b9f8475c7ba5e2.jpg

Image Credits
❖ https://www.incimages.com/uploaded_ﬁles/image/970x450/getty_dv778060_45540.jpg
❖ https://s3.amazonaws.com/s3.timetoast.com/public/uploads/photos/8221934/download_%285.jpg?
1478393447
❖ http://cliparts.co/cliparts/ATb/rkr/ATbrkrMXc.jpg
❖ http://bloggingwizard.adamconnell1.netdna-cdn.com/wp-content/uploads/2015/07/How-To-Grow-Your-
Blogs-Audience-S.png
❖ http://www.esosolutions.com/images/home_analytics.svg
❖ https://images-na.ssl-images-amazon.com/images/I/51SPHnoK%2BrL.jpg
❖ http://cdn.wonderfulengineering.com/wp-content/uploads/2016/01/Here%E2%80%99s-How-You-Can-
Combat-The-Exhaustion-You-Feel-All-The-Time-4.png
❖ http://awscomputeblogmedia.s3.amazonaws.com/
zombie_high_level_architecture_of_survivor_serverless_chat_app.png
❖ https://s-media-cache-ak0.pinimg.com/236x/99/00/51/990051c22c8abba5fbd6dc21786e91f3--ﬁre-and-ice-
crps.jpg
❖ http://pngimg.com/uploads/snake/snake_PNG4075.png

reachus@codeops.tech
www.codeops.tech
@CodeOpsTech
+91 98801 64463

Serverless Architecture - A Gentle Overview

More Related Content

What's hot (20)

Similar to Serverless Architecture - A Gentle Overview (18)

More from CodeOps Technologies LLP (20)

Recently uploaded (20)

Serverless Architecture - A Gentle Overview