Sowing the Seeds of Software Development: What Can You Do with Serverless – In 2021!

© 2021, Amazon Web Services, Inc. or its Affiliates.
Danilo Poccia
Chief Evangelist (EMEA)
@danilop
Sowing the Seeds of
Software Development
What Can You Do with Serverless – In 2021

View Slide


View Slide

Photo by Jeremy Bishop on Unsplash

View Slide

We have an idea
for a new
web/mobile app!
How can we build
it quickly and put
it in the hands of
customers?

View Slide

I want to build a mobile app that helps my
customers store and analyze videos taken with
their smartphone.
You can use a serverless architecture to focus
on the features you want to build. It’ll be much
easier to scale up if needed. You’ll also be
driven to think about security as you build it.
Serverless
Com
puting

View Slide

Serverless function
API Call
File uploaded
Database updated
Your code
Your business logic
Its own permissions
Something happens
Event
Serverless function
Serverless function
Serverless function
Serverless
Com
puting

View Slide

Analysis
Function
Videos
Results
Get Results
HTTP API
Upload
Get
Serverless
Com
puting
Event
Write
Read
Event

View Slide

Many functions used by my app return very
quickly, often in less than 20ms.
Billing
&
Cost O
ptim
ization
Billing granularity for Lambda functions has
been reduced from 100ms down to 1ms with
no minimum duration. This lowers the price
for most Lambda functions, more so for short
duration functions.

View Slide

REPORT RequestId: 35a7e0cb-4902-490d-b8d3-eb315dded660
Duration: 27.40 ms Billed Duration: 100 ms
Memory Size: 1024 MB Max Memory Used: 472 MB
Before
With 1ms billing granularity
REPORT RequestId: a24d03b5-429d-4ca3-a490-878a52a0182f
Duration: 27.55 ms Billed Duration: 28 ms
Memory Size: 1024 MB Max Memory Used: 472 MB
1m
s Billing
Granularity

View Slide

Lambda functions now can have up to 10 GB of
memory and 6 vCPUs. This is more than a 3x
increase compared to previous limits.
I want to be able to process
high resolution videos.
Large
Functions

View Slide

https://twitter.com/mooyoul/status/1334205430839472129
FFm
peg
First Case
Study

View Slide

M
L
Inference
w
ith
PyTorch
https://aws.amazon.com/blogs/aws/new-for-aws-lambda-functions-with-up-to-10-gb-of-memory-and-6-vcpus/
Same Costs
5x Faster

View Slide

I want to process videos faster. It’ll give a
better customer experience, and with a
serverless architecture will also be cheaper!
AVX2
support
AWS Lambda now supports Advanced Vector
Extensions 2 (AVX2), an extension to the x86
instruction set architecture that enables
running a set of highly
parallelizable operations simultaneously.

View Slide

© 2021, Amazon Web Services, Inc. or its Affiliates. Source: https://unsplash.com/photos/IMXhx6qhvf0. Photo credit: Daniel Seßler.
Bicubic filter
Bilinear filter Lanczos filter
Im
age
Resizing
U
sing
AVX2
https://aws.amazon.com/blogs/compute/creating-faster-aws-lambda-functions-with-avx2/

View Slide

Im
age
Resizing
U
sing
AVX2
Filter Without AVX2 With AVX2
Performance
improvement
Bilinear 105 ms 71 ms 32%
Bicubic 122 ms 73 ms 40%
Lanczos 136 ms 77 ms 43%
https://aws.amazon.com/blogs/compute/creating-faster-aws-lambda-functions-with-avx2/

View Slide

The Lambda functions analyzing videos need
many large dependencies. Also, I want to use a
custom machine learning model to extract
more information.
Packaging
&
Deploym
ent
AWS Lambda now supports container images
as a packaging format with an image size up to
10 GB. This also helps if you invested in
container tooling for your development
workflows.

View Slide

FROM public.ecr.aws/lambda/nodejs:latest
COPY app.js package*.json ./
RUN npm install
CMD [ "app.lambdaHandler" ]
Dockerfile
Dockerfile
U
sing
ECR
Public
https://aws.amazon.com/blogs/aws/new-for-aws-lambda-container-image-support/
You can use base container images prepared
by AWS and shared via Docker Hub and
Amazon Elastic Container Registry
(ECR) Public.

View Slide

With serverless I can use small functions that
react to events. But now my main function is
becoming too big, mixing different
functionalities such as video resizing and using
a custom machine learning model.
O
rchestration
&
Choreography
You can decompose a large function is small
tasks that you can execute (and retry)
independently from each other.
AWS Step Functions now supports
Synchronous Express Workflows.

View Slide

Synchronous
Express W
orkflow
s
https://aws.amazon.com/blogs/compute/new-synchronous-express-workflows-for-aws-step-functions/

View Slide

I want to understand better what my functions
are doing in production.
Observability is key! To help you with that,
Amazon CloudWatch Lambda Insights is now
generally available. Also, have a look a the
AWS Distro for OpenTelemetry.
O
bservability
&
Governance

View Slide

CloudW
atch
Lam
bda
Insights
https://aws.amazon.com/blogs/aws/using-amazon-cloudwatch-lambda-insights-to-improve-operational-visibility/

View Slide

I want to use S3 as the source of truth while
processing images and extracting metadata.
Isn’t S3 eventually consistent while updating
objects? I have the same issue with my data
lake where I store and analyze my logs and
application data.
Storage
Amazon S3 now delivers strong
read-after-write consistency
automatically for all applications.

View Slide

Storage

View Slide

I want to keep my data at rest encrypted.
Storage
Amazon S3 Bucket Keys reduce the costs of
Server-Side Encryption with AWS Key
Management Service (SSE-KMS)

View Slide

Storage

View Slide

I want to use DynamoDB for its low latency
and on-demand scalability, but I am not
familiar with NoSQL and its syntax.
Database
You now can use a SQL-compatible query
language to query, insert, update, and delete
table data in Amazon DynamoDB.
Also, the NoSQL Workbench can help you
design your data model and build your
operations.

View Slide

N
oSQ
L
W
orkbench

View Slide

I want to create a chatbot to simplify customer
interactions and give access to customer
support. I’d like each customer to use their
native language.
N
atural
Language
Support
Amazon Lex has added language support for
French, Spanish, Italian, Canadian French,
German, and Latin American Spanish.

View Slide

M
ulti-Language
Chatbot

View Slide

I want to use the chatbot
from my contact center.
N
atural
Language
Support
Amazon Connect now supports Amazon Lex
chatbots for French, Spanish, Italian and
Canadian French, Latin American Spanish, and
German.

View Slide

Sentiment Analysis & User Insights
Cloud-native Contact Center Reference Architecture for Multilingual IT Service Desk
This reference architecture provides an omnichannel experience
for end users calling the IT Service Desk (SD) using one or more
support channels of their choice, such as a landline phone using
public switched telephone network (PSTN), a softphone from
their client device, or from their smartphone using an internet
connection.
The SD agent, working from a corporate office, uses a web-based
softphone (CCP) from an internet browser to connect to Amazon
Connect and interact with end users by answering calls,
performing web chats, or setting their status in the CCP UI. SD
agents can also use a traditional telephone service using the
PSTN.
SD agents can also connect to Amazon Connect remotely
through internet connectivity from their home office. All data
exchanged with Amazon Connect is protected in transit between
the user’s web browser and Amazon Connect using TLS 1.2 or
above encryption standards.
Amazon Connect provides inbound/outbound voice channel,
web, and mobile chat capabilities with skills-based routing. As
part of the initial configuration, Amazon Connect enables you to
claim the phone number and define the support queues, contact
flows, and routing profiles, which can notify users through email
or SMS about the status of the ticket. Using Connect Tasks, SD
agents and managers can automate manual tasks such as user
follow-ups, setup reminders, and more.
Amazon Connect integrates with customers’ customer
relationship management (CRM) or any business intelligence (BI)
applications hosted in AWS or an on-premise location, or any of
the existing IT Service Management (ITSM) ticketing tools like
ServiceNow to pull caller information. This is achieved either by
using connectors developed by the AWS Partner ecosystem, or by
creating a custom integration connector using AWS Lambda.
Further integrating Amazon Connect with Amazon Lex, you can
create intelligent conversational chatbots to automate high
volume user contacts without increasing the number of SD
agents, and without compromising end users experience. End
users are able perform self-service tasks such as password
change, procurement of IT accessories, or scheduling meetings
using natural conversation language with Amazon Lex. In
conjunction with the contact flows feature in Amazon Connect,
businesses can turn automated interactions into natural
conversations to improve end-user experience.
To improve the efficiency and quality of the contact handling, IT
SD can gain end-user insights using Amazon Transcribe and
Amazon Translate to convert speech to text from the voice
recording stored in Amazon S3. By leveraging Amazon
Comprehend, SD agents can then analyze the key topics and
relationships in any given text using Amazon Athena. Amazon
Macie can be used to protect sensitive data stored in the call
recording, and focus on the incident or problem context. With this
analysis, the IT SD can identify user sentiments, and address the
common issue or a trend that impacts end users’ productivity or
satisfaction.
Amazon Connect enables the SD agents to see real-time metrics
such as SD agents logged in, call abandon rates, number of calls
handled, SD agent activity status, and so on. This is done by
analyzing contact trace records (CTRs) using Amazon Kinesis
Data Streams and Lambda functions, storing them in Amazon
Redshift, and viewing them on the Amazon Quicksight
dashboard.
© 2021, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Use AWS building blocks for modernizing the IT service desk with an intelligent, omnichannel-based contact center capability using
Amazon Connect. This reference architecture supports voice, web, mobile, and conversational chatbot channels, including end user
sentiment analysis, automation, real-time analytics dashboard, and reporting.
Phone, Speech Assistant, Web and Chat
Voice & Chat Support Channel
Inbound &
Outbound Voice
Analytics & Reporting
Internet
Amazon Kinesis
Data Firehose
Amazon Kinesis
Data Streams
Mobile
client
End Users
Client
Client
Service Desk
Agents
Office Contact
Control Pane
(CCP)
Customer
GW (Primary)
Remote
Service Desk
Agents
Internet
GW
Customer
GW (Backup)
Contact
Control Pane
(CCP)
CRM System
BI System
Internet
Conversational Bot Support Channel
NLP based Bots capability
Amazon Lex
Amazon
DynamoDB
Amazon Lex
Validation
Lambda function
Amazon SMS
Events
AWS Lambda
Amazon Connect
Contact
Flows
Amazon
Redshift
AWS Lambda
Amazon S3
Amazon
QuickSight
Amazon
Transcribe
Amazon
Comprehend
Amazon Macie
Amazon S3
(stores call
recording)
Amazon
Athena
Metrics
Call
Recordings
Amazon
Translate
Amazon
QuickSight
Amazon S3
Machine Learning
capabilities
PSTN
AWS S2S VPN
AWS Direct
Connect
1
2
4
3
5
6
7
Integration with 3rd
party systems (A)
A. Amazon Connect is an open platform, providing out-of-the-box integrations for leading CRM tools such as Salesforce and Zendesk, Workforce Management (WFM) tools, and analytics tools. Using Lambda, customers
can create their own integration to their existing ITSM or CRM products. To learn more about various APN partners who have created a custom integration connector with their product to leverage Amazon Connect, see
Amazon Connect Partners.
B. Amazon Connect Integration offers a Quick Start guide to organizations using ServiceNow as their ticketing tool. Organizations can use Lambda functions to jumpstart their integration of ServiceNow and Amazon
Connect, resulting in one unified ITSM UI for SD agents and end users.
1
2
3
4
5
6
7
AWS Reference Architecture
(B)
E-mail
notification
Text
messages
Amazon
Connect
Tasks
for
automation
AWS
Lambda
Customer Workloads
Customer Location
APN Partners
ServiceNow
Contact Center
Reference
Architecture

View Slide

Photo by veeterzy on Unsplash

View Slide

Thank you!
Please give me your feeback J
Danilo Poccia
Chief Evangelist (EMEA)
@danilop

View Slide

Sowing the Seeds of Software Development: What Can You Do with Serverless – In 2021!

Sowing the Seeds of Software Development: What Can You Do with Serverless – In 2021!

Danilo Poccia

More Decks by Danilo Poccia

Other Decks in Programming

Featured

Transcript