How Quarkus makes your apps cheaper, greener, and happier

how quarkus makes development faster greener happier Holly Cummins Senior
Principal Software Engineer, Quarkus London Java Community November 11, 2025 https://bsky.app/profile/hollycummins.com

how quarkus makes development faster greener happier cheaper Holly Cummins
Senior Principal Software Engineer, Quarkus London Java Community November 11, 2025 https://bsky.app/profile/hollycummins.com

@holly_cummins #Quarkus #IBM #RedHat

@holly_cummins #Quarkus #IBM #RedHat why are you smiling?

@holly_cummins #Quarkus #IBM #RedHat why are you smiling? work is
not a place to be happy!

#RedHat @holly_cummins ΑΡΙΣΤΟΤΈΛΗΣ

#RedHat @holly_cummins «τελειοῖ δὲ τὴν ἐνέργειαν ἡἡδονή» ΑΡΙΣΤΟΤΈΛΗΣ

#RedHat @holly_cummins «Η ευχαρίστηση στην δουλειά βάζει την τελειότητα στην
εργασία» ΑΡΙΣΤΟΤΈΛΗΣ

#RedHat @holly_cummins “Pleasure in the job puts perfection in the
work.” Aristotle

#RedHat @holly_cummins joy has business value.

#RedHat @holly_cummins job satisfaction is the no. 1 predictor of
organizational performance

#RedHat @holly_cummins not all software creates joy

@holly_cummins #RedHat most java web frameworks are verbose there is
a lot of typing

@holly_cummins #RedHat most java web frameworks are resource-hungry they hog
cpu and ram

@holly_cummins #Quarkus #IBM #RedHat

@holly_cummins #Quarkus #IBM #RedHat “I can’t bring up the microservices
in my Java dev stack

@holly_cummins #Quarkus #IBM #RedHat “I can’t bring up the microservices
in my Java dev stack … on my brand new Apple laptop with a M1 chip and 64GB of RAM.” - fintech CTO

@holly_cummins #Quarkus #IBM #RedHat in production, it’s worse

@holly_cummins #Quarkus #IBM #RedHat in production, it’s worse example microservices
maths:

maths: airline maintenance scheduling system

maths: airline maintenance scheduling system single service: ½ core + 1 GB RAM

maths: airline maintenance scheduling system single service: ½ core + 1 GB RAM HA → 3x instances

maths: airline maintenance scheduling system single service: ½ core + 1 GB RAM HA → 3x instances ~100 microservices

maths: airline maintenance scheduling system single service: ½ core + 1 GB RAM HA → 3x instances ~100 microservices = 150 cores + 300 GB RAM

@holly_cummins #Quarkus #IBM #RedHat all this resource usage is expensive

@holly_cummins #Quarkus #IBM #RedHat slow development is expensive

@holly_cummins #Quarkus #IBM #RedHat why is this happening? mismatch between
what we need and what the platform is optimised for

@holly_cummins #Quarkus #IBM #RedHat java was not designed for the
cloud

@holly_cummins #RedHat old Java frameworks were optimised for …

@holly_cummins #RedHat long-lived processes old Java frameworks were optimised for
…

@holly_cummins #RedHat long-lived processes annual (!) deployments old Java frameworks
were optimised for …

@holly_cummins #RedHat long-lived processes annual (!) deployments late-binding old Java
frameworks were optimised for …

@holly_cummins #RedHat long-lived processes annual (!) deployments late-binding re-configurable without
restart old Java frameworks were optimised for …

@holly_cummins #Quarkus #IBM #RedHat Container platform machine go go go
go go go go go go go go go go go

@holly_cummins #Quarkus #IBM #RedHat Container platform machine node.js node.js node.js
node.js node.js node.js node.js machine go go go go go go go go go go go go go go

@holly_cummins #Quarkus #IBM #RedHat Container platform machine HotSpot Heap HotSpot
Heap HotSpot Heap HotSpot Heap machine node.js node.js node.js node.js node.js node.js node.js machine go go go go go go go go go go go go go go

can we do better than that?

enter … quarkus a Java framework that gets you going
faster, faster.

@holly_cummins #Quarkus #IBM #RedHat traditional cloud-native java stack traditional cloud-native
java stack traditional cloud-native java stack traditional cloud-native java stack node.js node.js node.js node.js node.js node.js node.js go go machine go go go go go go go go go go go go go go go go go go go quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus spoiler: we made stuff better :) container orchestration machine machine machine https:/ /developers.redhat.com/blog/2017/03/14/java-inside-docker/

@holly_cummins #Quarkus #IBM #RedHat machine quarkus quarkus quarkus quarkus quarkus
quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus container orchestration machine traditional cloud-native java stack traditional cloud-native java stack traditional cloud-native java stack traditional cloud-native java stack … a lot better quarkus native

@holly_cummins #Quarkus #IBM #RedHat machine quarkus quarkus quarkus quarkus quarkus
quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus quarkus container orchestration machine traditional cloud-native java stack traditional cloud-native java stack traditional cloud-native java stack traditional cloud-native java stack … a lot better quarkus native (but quarkus on jvm is also way smaller than traditional java)

quarkus applications start fast quarkus + graalvm 0.014 Seconds rest
application quarkus + open jdk 0.75 Seconds traditional cloud-native stack 4.3 Seconds https://quarkus.io/blog/runtime-performance/

@holly_cummins to the code! demo!

@holly_cummins #Quarkus #IBM #RedHat mvn quarkus:build -Pnative target/code-with-quarkus-1.0.0-SNAPSHOT-runner

thanks to markus eisele for the graphics code https://github.com/holly-cummins/meme-generator

ok but does startup time matter?

ok but does startup time matter? fast boot time means
auto-scaling works better more resilience to load spikes

quarkus + GraalVM 13 MB quarkus + OpenJDK 74 MB
traditional cloud-native Stack 140 MB rest application https://quarkus.io/blog/runtime-performance/ quarkus improves memory utilization

#Quarkus @holly_cummins

#Quarkus @holly_cummins hey, wanna see quarkus? hey, wanna see quarkus?
hey, wanna see quarkus? hey, wanna see quarkus? hey, wanna see quarkus?

hey, wanna see quarkus? hey, wanna see quarkus? hey, wanna see quarkus? uhh … are you supposed to shut down applications after using them?

hey, wanna see quarkus? hey, wanna see quarkus? hey, wanna see quarkus? uhh … are you supposed to shut down applications after using them? 120 instances (!)

ok but does memory footprint matter?

ok but does memory footprint matter? in the cloud, memory
footprint is money

@holly_cummins #Quarkus #IBM #RedHat remember the airline scheduling application?

@holly_cummins #Quarkus #IBM #RedHat remember the airline scheduling application? “[With
Quarkus], we can run 3 times denser deployments without sacrificing availability and response times of services. ” – Thorsten Pohl

let’s talk about throughput. https://www.redhat.com/en/resources/mi-quarkus-lab-validation-idc-analyst-paper 48 concurrent connections Traditional cloud
native stack 3555 req/s

let’s talk about throughput. quarkus native 3212 req/s https://www.redhat.com/en/resources/mi-quarkus-lab-validation-idc-analyst-paper 48
concurrent connections Traditional cloud native stack 3555 req/s

let’s talk about throughput. quarkus native 3212 req/s https://www.redhat.com/en/resources/mi-quarkus-lab-validation-idc-analyst-paper a
trade-off of throughput against footprint 48 concurrent connections Traditional cloud native stack 3555 req/s

@holly_cummins native compilation trade-offs throughput startup time + footprint

@holly_cummins native compilation trade-offs throughput startup time + footprint this
is a classic tradeoff

but … traditional cloud native stack 3555 req/s quarkus native
3212 req/s https://www.redhat.com/en/resources/mi-quarkus-lab-validation-idc-analyst-paper a trade-off of throughput against footprint 48 concurrent connections

but … traditional cloud native stack 3555 req/s quarkus on
jvm 6389 req/s quarkus native 3212 req/s https://www.redhat.com/en/resources/mi-quarkus-lab-validation-idc-analyst-paper a trade-off of throughput against footprint 48 concurrent connections

but … traditional cloud native stack 3555 req/s quarkus on
jvm 6389 req/s quarkus native 3212 req/s https://www.redhat.com/en/resources/mi-quarkus-lab-validation-idc-analyst-paper no trade-off, just better :) a trade-off of throughput against footprint 48 concurrent connections

@holly_cummins we beat the trade-off. throughput startup time + footprint

@holly_cummins we beat the trade-off. throughput startup time + footprint
it’s a double-win.

#Quarkus @holly_cummins users care (a lot) about response time

better startup better footprint better throughput wait, isn’t that something
for nothing?

@holly_cummins #Quarkus #IBM #RedHat the two ways of improving performance
•trade off one thing against another •eliminate waste

@holly_cummins #RedHat where the win comes from

@holly_cummins #RedHat application frameworks were optimised for dynamism

@holly_cummins #RedHat application frameworks optimised for dynamism dynamism has a
cost

@holly_cummins #RedHat paying a dynamism tax … even though the
app is not dynamic

@holly_cummins #RedHat a highly dynamic runtime in a container is
pointless

@holly_cummins #RedHat cloud apps are immutable now

@holly_cummins #Quarkus #IBM #RedHat how does a java framework start?

build time

build time runtime

packaging (maven, gradle…) build time runtime

build time runtime

</> build time runtime load and parse • config files • properties • yaml • xml • etc.

</> build time runtime

@ @ </> build time runtime • classpath scanning and annotation discovery • attempt to load class to enable/disable features

@ @ </> build time runtime

@ @ </> build time runtime build a metamodel of the world

@ @ </> build time runtime

@ @ </> build time runtime start • thread pools • I/O • etc.

@ @ </> build time runtime ready to do work!

@holly_cummins #Quarkus #IBM #RedHat what if we start the application
more than once? @ @ </>

more than once? @ @ </> @ @ </>

more than once? @ @ </> @ @ </> @ @ </>

more than once? @ @ </> @ @ </> @ @ </> @ @ </>

more than once? @ @ </> @ @ </> @ @ </> @ @ </> so much work gets redone every time

@holly_cummins Hibernate speed example: JTA auto-wiring

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”);

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”);

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”); Class.forName(“AnotherJTAImplementation”);

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”); Class.forName(“AnotherJTAImplementation”); …

@holly_cummins Hibernate speed example: JTA auto-wiring Class.forName(“LikelyJTAImplementation”); Class.forName(“APossibleJTAImplementation”); Class.forName(“AnotherJTAImplementation”); Class.forName(“NicheJTAImplementation”);
…

Class.forName(“VeryNicheJTAImplementation”); …

Class.forName(“VeryNicheJTAImplementation”); … ~129 auto-wiring attempts

Class.forName(“VeryNicheJTAImplementation”); … ~129 auto-wiring attempts every single start.

@holly_cummins it’s not just JTA this happens for lots of
internal service bindings

@holly_cummins JVM footprint example: Hibernate

@holly_cummins JVM spends time loading classes for specific databases JVM
class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database footprint example: Hibernate

class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database class for unused database turns out they’re never used footprint example: Hibernate

turns out they’re never used JIT spends time unloading classes footprint example: Hibernate

@holly_cummins Hibernate example: ~500 classes which are only useful if
you're running an Oracle database loaded and then unloaded

@holly_cummins Hibernate example: ~500 classes which are only useful if
you're running an Oracle database loaded and then unloaded every single start.

@holly_cummins the true cost of loaded classes isn’t just memory
+ start time

+ start time method dispatching:

@holly_cummins interface the true cost of loaded classes isn’t just
memory + start time method dispatching:

@holly_cummins unused implementation the one we want interface unused implementation
unused implementation the true cost of loaded classes isn’t just memory + start time method dispatching:

@holly_cummins unused implementation the one we want interface megamorphic call
slow dispatching unused implementation unused implementation the true cost of loaded classes isn’t just memory + start time method dispatching:

+ start time the one we want interface

+ start time the one we want monomorphic call fast dispatching interface

@holly_cummins how do we fix all this?

@holly_cummins #Quarkus #IBM #RedHat @ @ </> build time runtime
what if we initialize at build time?

start • thread pools • I/O • etc. what if we initialize at build time?

ready to do work! start • thread pools • I/O • etc. what if we initialize at build time?

@holly_cummins #Quarkus #IBM #RedHat @ @ </> repeated starts are
now efficient

@holly_cummins #Quarkus #IBM #RedHat @ @ </> repeated starts are
now efficient less wasted work

@holly_cummins #Quarkus #IBM #RedHat the Quarkus way enables native compilation
native (graalvm) @ @ </> jvm build time

ok but we don’t need native

ok but we don’t need native quarkus is faster and
smaller than legacy frameworks, even running on the jvm

@holly_cummins

@holly_cummins doing more up-front

@holly_cummins doing more up-front - speeds up start

@holly_cummins doing more up-front - speeds up start - shrinks
memory footprint

@holly_cummins doing more up-front - speeds up start - shrinks
memory footprint - improves throughput (!)

ok but is performance all there is?

ok but is performance all there is? developer joy

@holly_cummins #Quarkus #IBM #RedHat with quarkus, developers have to type
less boilerplate

to the code! @holly_cummins #RedHat

@holly_cummins #Quarkus #IBM #RedHat mvn quarkus:dev zero-config live coding

@holly_cummins #Quarkus #IBM #RedHat tests are run on every code
change “reverse code coverage” means only relevant tests run mvn quarkus:dev continuous testing

@holly_cummins #Quarkus #IBM #RedHat developer UI

@holly_cummins #Quarkus #IBM #RedHat zero-config testcontainers

@holly_cummins #Quarkus #IBM #RedHat testcontainers integration … without quarkus @TestConfiguration(proxyBeanMethods
= false) public class ContainersConfig { @Bean @ServiceConnection public PostgreSQLContainer<?> postgres() { return new PostgreSQLContainer<>(DockerImageName.parse("postgres:14")); } } public class TestApplication { public static void main(String[] args) { SpringApplication .from(MySpringDataApplication::main) .with(ContainersConfig.class) .run(args); } } @Import(ContainersConfig.class)

@holly_cummins #Quarkus #IBM #RedHat testcontainers integration … without quarkus

@holly_cummins #Quarkus #IBM #RedHat zero-config testcontainers integration the only thing
you need to do to make testcontainers work is not configure anything # configure your datasource quarkus.datasource.db-kind = postgresql quarkus.datasource.username = sarah quarkus.datasource.password = connor quarkus.datasource.jdbc.url = jdbc:postgresql://localhost:5432/mydatabase # drop and create the database at startup quarkus.hibernate-orm.database.generation = drop-and-create

you need to do to make testcontainers work is not configure anything # drop and create the database at startup quarkus.hibernate-orm.database.generation = drop-and-create

you need to do to make testcontainers work is not configure anything

you need to do to make testcontainers work is not configure anything quarkus also auto-invokes flyway and liquibase

@holly_cummins #Quarkus #IBM #RedHat more opinions, less boilerplate

@holly_cummins #Quarkus #IBM #RedHat @ApplicationScoped public class GreetingRepository { public
Entity findByName(int name) { return find("name", name).firstResult(); } void persist(Entity entity) {} void delete(Entity entity) {} Entity findById(Id id) {} List<Entity> list(String query, Sort sort, Object... params) { return null; } Stream<Entity> stream(String query, Object... params) { return null; } long count() { return 0; } long count(String query, Object... params) { return 0; } } example: panache + hibernate

@holly_cummins #Quarkus #IBM #RedHat example: panache + hibernate @ApplicationScoped public
class GreetingRepository implements PanacheRepository<Greeting> { public Entity findByName(int name) { return find("name", name).firstResult(); } }

@holly_cummins #Quarkus #IBM #RedHat DAO example: panache + hibernate @ApplicationScoped
public class GreetingRepository implements PanacheRepository<Greeting> { public Entity findByName(int name) { return find("name", name).firstResult(); } } repository pattern

@holly_cummins #Quarkus #IBM #RedHat example: panache + hibernate

@holly_cummins #Quarkus #IBM #RedHat example: panache + hibernate active record
pattern @Entity public class Greeting extends PanacheEntity { public String name; public LocalDate issued; @Version public int version; public static List<Greeting> getTodaysGreetings() { return list("date", LocalDate.now()); } }

@holly_cummins “but isn’t that dynamism expensive?”

@holly_cummins no. “but isn’t that dynamism expensive?”

@holly_cummins no. it’s done at build time “but isn’t that
dynamism expensive?”

@holly_cummins #Quarkus #IBM #RedHat ok but i don’t like magic

enrichment happens at build time

enrichment happens at build time no performance drag

doing more up-front enables better devex runtime build time @
@ </>

doing more up-front enables better devex runtime build time we
can do cool code introspections here that would be too expensive and annoying to do at runtime @ @ </>

@holly_cummins #Quarkus #IBM #RedHat ok but i still don’t like
magic

@holly_cummins #Quarkus #IBM #RedHat the old ways all still work
ok but i still don’t like magic

@holly_cummins #Quarkus #IBM #RedHat the old ways all still work
but you don’t have to type all the stuff unless you want to ok but i still don’t like magic

@holly_cummins #Quarkus #IBM #RedHat ok but we use spring

@holly_cummins #Quarkus #IBM #RedHat <dependency> <groupId>io.quarkus</groupId> <artifactId>quarkus-spring-web</artifactId> </dependency> <dependency> <groupId>io.quarkus</groupId>
<artifactId>quarkus-spring-data-jpa</artifactId> </dependency> ok but we use spring option 1: compatibility libraries option 2: migration tooling • konveyor • open rewrite

@holly_cummins #Quarkus #IBM #RedHat “After a week of development with
quarkus, I was able to regain the same level of productivity as when I was developing with Spring Boot.” – Fawaz Paraïso, Decathlon

@holly_cummins #Quarkus #IBM #RedHat the big picture …

@holly_cummins #Quarkus #IBM #RedHat does being faster and lighter mean
quarkus is greener?

@holly_cummins #RedHat density Source: Clement Escoffier experiment 1: cloud Setup:
• 800 requests/second, over 20 days • SLA > 99% • AWS instances Assumptions: • Costs are for us-east-1 data centre

@holly_cummins #RedHat Setup: • 800 requests/second, over 20 days •
SLA > 99% Assumptions: Source: Clement Escoffier x Teads cloud carbon impact of framework choice interpolated carbon metrics

@holly_cummins #RedHat Setup: • 800 requests/second, over 20 days •
SLA > 99% Assumptions: Source: Clement Escoffier x Teads cloud carbon impact of framework choice the carbon is lower because the cost is lower interpolated carbon metrics

@holly_cummins #RedHat Setup: • REST + CRUD • large heap
• RAPL energy measurement • multiple instances to support high load  Assumptions: • US energy mix Source: John O’Hara experiment 2: RAPL measurements

@holly_cummins #RedHat Setup: • REST + CRUD • large heap
• RAPL energy measurement • multiple instances to support high load  Assumptions: • US energy mix Source: John O’Hara experiment 2: RAPL measurements quarkus on JVM has the lowest carbon … because it has the highest throughput

• quarkus cuts carbon by ~2-3x* • native consumes more
carbon than JVM (but it’s great for serverless scenarios!) carbon measurements: conclusions

to the code! @holly_cummins #RedHat

https://github.com/quarkusio/spring-quarkus-perf-comparison

@holly_cummins #Quarkus #IBM #RedHat ours is the only profession where
“lazy” is a compliment

@holly_cummins #Quarkus #IBM #RedHat we need to make it a
profession where “happy” is also a compliment

@holly_cummins #Quarkus #IBM #RedHat tl;dpa (too long didn’t pay attention)
deployment density lower cloud bill frictionless development experience Medium Nano auto-provision services zero-config live coding continuous testing developer UI greener happy, productive people

slides Holly Cummins IBM http://hollycummins.com/greener-faster-happier-quarkus-ljc/

How Quarkus makes your apps cheaper, greener, a...

How Quarkus makes your apps cheaper, greener, and happier

More Decks by Holly Cummins

Other Decks in Programming

Featured

Transcript