September 2021

Breakout Session Application Architecture

Summit

Data Stack Requirements in a Microservices World

Wen’s Bio Paras’s Bio
Extensive experience in cloud- Oversees DataStax Wens Gerdyman
based enterprise software, Enterprise data platform built Product Marketing
including microservices on Cassandra Leader

Strong expertise with data, Nearly 10 years in NoSQL

cloud, B2B integration space Paras Mehra
Product Manager
Prior to DataStax, worked at Expertise in distributed
Tibco analytics

© 2021 Enterprise Integration News, Inc.

Data Stack Requirements
in the Microservices World

Wens Gerdyman, Product Marketing Leader at DataStax

Paras Mehra, Product Management Leader at DataStax
Digital transformation
A journey in progress

Act 1 Act 2 Act 3

Connectivity Cloud Data
Scale out networking to Scale out computing to Scale out data to build
connect applications and build cloud and mobile modern applications.
devices. applications
- Data explosion
- Highly interactive
- Real-time decisions

AWS/Microsoft/Google Cloud Apigee | Pivotal DataStax | Snowflake

3
Data today: still an afterthought

The application-first traditional approach locks

application data in many technologies relying on
different infrastructures.
Application

Negative outcomes:
Application Data
● Scalability and availability challenges interfaces pipelines
● Slow time to market
● High complexity and costs

Application Analytical
data systems

Fast lane Deep lane

4
Turning data into an asset
Fast lane Deep lane

Availability and scalability

Data Data
Security, privacy and residency.

Data Data
Standard data APIs service pipelines
By switching to a data-first strategy s
and cloud architectures, enterprises
modernize data:
Application
Cloud-native
● Simple to use in applications
architecture
● Available to scale
● Open for innovation

5
 Application-first approach

µSvc µSvc µSvc

1 2 3 OUTCOMES

CONSTRAINTS Compromised
Performance
Impedance mismatch
between objects and Big Teams
Procedures Joins Queries
Procedures Joins Queries restrictive schema...
No Agility to
Requiring Business Innovate
Logic tightly coupled
Normalized Tables
with Database

Storage optimized
RDBMS
 Data-First Approach

OUTCOMES

µSvc µSvc µSvc WHAT IT ENABLES Super fast writes-

1 2 3 reads
No mismatch between
objects and NoSQL Small
schema (no or relaxed) development
teams
µSvcs inform each other
API API API
to maintain consistency Business Agility

Logic not tightly coupled Frequent

with database Innovations

Optimized for Queries

Open data stack
requirements
Requirement: Distributed for linear scalability
Need more throughput ?
Add more nodes

Horizontal scale

Linear Scalability

100,000 200,000 400,000

transactions/second transactions/second transactions/second
Zero Downtime

Need more capacity ?

Add more nodes. Then,
9 remove them when done
Requirement: Distributed for Availability

Leader: Read/Write Peer: Read/Write

Peer: Read/Write

Peer: Read/Write

Read-only Read-only Read-only

Replica Replica Replica
Peer: Read/Write

● Single point of failure

● Limited availability during Leader
● 100% Availability
election
○ Every node accepts reads & writes
● Doesn’t scale linearly
● Data evenly distributed, no data silos
○ Read availability exceeds write
10
availability
Benefits of a distributed data platform

• Geographic Distribution • Hybrid-Cloud and Multi-Cloud

On-premise
Benefits of a distributed data platform

Always-on Linear
Low Latency
Availability scalability

Performance = Revenue Always be “Open for Business” Agility to keep up with demand

• Every 100 milliseconds of Every second represents Scale out as needed, then scale in.
latency causes a 1% decrease millions in lost revenue during GigaOm says this can result in
in sale (Amazon.com) iPhone Launch (USA-based 75% TCO reduction.
• 2 sec slowdown reduce Telco)
revenue per user by 4.3%
(Bing)
12
Requirement: Run on Kubernetes
Pod 1 Pod 2 Pod n

RC / Deployment

Scale

100,000 TPS 200,000 TPS 400,000 TPS

Horizontal
Pod
Autoscaler

Modern microservices run on Kubernetes Cassandra is a distributed system that enables

companies to deploy new data nodes to deal with
Kubernetes is a distributed system that increase in throughput demand
enables companies to scale horizontally by
deploying new application pods to deal with
increase in throughput demand

Running 2 distributed systems with 2 control planes is complex!

Milestone: Run Cassandra and application pods on k8s

Kubernetes Control Plane

k8s instructions
k8s
instructions

Cass Operator

Cassandra commands

App App App

Pod Pod Pod
Ultimate: Cloud native data pods running in k8s
Self-Managed Managed Service
Your choice of Cloud

Centralized Control Plane: Both self-managed and managed DBaaS

On-Prem DC
Requirement: Data API

Why do you care?

1. Developers want to work with languages and APIs they already know
2. Developers do not care about normalized data modeling. They just
want to start storing data and retrieving them, fast.
3. Front-end developers want to prototype and work directly with the
backend
Data API Layer by DataStax
Drivers Open API

Cassandra GraphQL REST Document

Query Language
● Hierarchy of types ● Row based ● JSON Documents
● SQL like Table
and fields ● Structured Data ● Semi-Structured
Model
● Structured Data ● Key-Value Data Data
● Structured Data
● Key-Value Data ● Weaker Types ● Weaker Types
● Key-Value Data
● Strong Types
● Strong Types

More Structured
Less Structured

17
Benefits of Data API Layer

Data-driven Client Applications APP {REST API}

{Document}
APP
(Schemaless)API}

APP {GraphQL API}

APP CQL API for drivers

• No more barrier to data

• Jump starts developer productivity
• Separation of data and application allow changes to happen
without business disruption
18
Requirement: separation of compute from storage

Both need to scale

out (not optimal) compute
compute
data
data

If Compute and Storage are tightly coupled:

● As compute demand increases, compute-intensive applications need to pay for

increased storage too, e.g., catalogs during promotional periods
● As storage demand increases, storage-intensive applications need to pay for
increased compute capacity too, e.g., CRM moving from 8x5 to 24x7
Requirement: scale in as easily as scale out

?
?

Scaling in is VERY DIFFICULT for a distributed data store.

Shrinking data capacity willy nilly can wreak major havoc.

DataStax solved this problem.

Benefits of Separation of Compute and Storage

75%
TCO Reduction with serverless, pay-as-you-go DBaaS

GigaOm study, 2021

Modernizing Data requires an Open Data Stack

Developers ● Modern Data APIs allow developers to access

data via standard interfaces and languages.
Modern Data APIs ● Open Source technologies to avoid vendor lock-
in.

Database ● Database supporting the availability and

scalability needs of digital applications.
● Management via Kubernetes to manage the data
Kubernetes
like the rest of the applications..
● Hybrid and multi-cloud. Deployment on any
Public | Private | Hybrid | Multi-Cloud cloud platform or on-premises.

22
Example of an open data stack to enable modern apps

Stargate
Data APIs ● Choice of Standard data APIs to fit
the needs of applications.
● Database as a service that does not
require any operations.
Column- Document Key-value Storage Attached ● Open Source, multi-model database
Dynamic Data Family Index

Key-value, Wide Column,

built on Apache Cassandra.
Document. ● Serverless, cloud-native architecture
managed via Kubernetes.
● Multi-cloud. Deployment on any cloud
Multi-Cloud AND platform.
Serverless
Multi-region,Inter-cloud

23
Benefits of an Open Data Stack for Modern Applications

The Open Data Stack DataStax

Simple to build Simple to operate Simple to scale Own your destiny Partners in
● Data APIs ● No operations ● Low latency ● Multi-cloud success
● Multi-model ● Security ● Zero downtime ● Open Source ● Methodology
● Up to 75% TCO ● Global scalability ● Best practices, tools
reduction

24
Modernizing data and beyond
The Open Data Stack powers three strategic initiatives

Modernize data and Empower developers to Decide and act in real-

applications build applications time
Unlock and reuse data to Build and scale modern Enrich applications with instant
modernize applications. applications faster without data actions.
constraints.

25
Free signup astra.datastax.com
linkedin.com/company/datastax
youtube.com/c/DataStax
community.datastax.com
datastax.com/workshops

26
Thank You
 September 2021
Q&A Application Architecture
Summit

Q&A
© 2021 Enterprise Integration News, Inc.
Free signup astra.datastax.com
linkedin.com/company/datastax
youtube.com/c/DataStax
community.datastax.com
datastax.com/workshops

29
 September 2021
For More Information Application Architecture
Summit

The Open, Multi-Cloud Stack for Modern Data Apps

https://www.datastax.com/

Popular Resources Top Downloads Free Trial & More

Resource Center Analyst Report: Gartner: The Future of Sign up for Astra Today
https://www.datastax.com/resources https://auth.cloud.datastax.com/auth/realms/CloudUsers/
Cloud Data Management is Multicloud protocol/openid-connect/registrations?client_id=auth-
https://www.datastax.com/resources/report/gartner-
proxy&response_type=code&scope=openid+profile+emai
• Analyst Reports report-future-cloud-data-management-multicloud
l&redirect_uri=https://astra.datastax.com&
• Customer Stories
• Datasheets Datasheet: DataStax Astra DB 3 Tips: Working & Winning with Kubernetes
https://www.datastax.com/sites/default/files/content/whit
• Podcasts and Cassandra
epaper/files/2021-07/Whitepaper_AstraDB-Designing-
• Tech Guides Serverless-Cloud-Native-DBaaS_6141_07.22.21.pdf https://www.datastax.com/working-and-winning-with-
kubernetes-and-cassandra-3-key-
• Whitepapers
Whitepaper: Apache Cassandra™ Use Cases
insights?utm_medium=sponsor_event&utm_source=IDev
News&utm_campaign=dgds-
Astra – Cassandra as a Service Architecture webinar&utm_content=Winning-w-KandC
https://www.datastax.com/products/datastax-astra https://www.datastax.com/resources/whitepaper/apache
-cassandratm-architecture

Data Modernization - A Future State How Instagram Uses Cassandra to

https://www.datastax.com/resources/whitepaper/data- Busting Old Myths about Apache Operate on a Global Scale
modernization-a-future-state?bypass Cassandra ™ https://www.datastax.com/blog/accelerate-rewind-how-
https://www.datastax.com/resources/whitepaper/data- instagram-uses-cassandra-operate-global-scale-2
modernization-a-future-state?bypass

© 2021 Enterprise Integration News, Inc.