Upgrade to Pro — share decks privately, control downloads, hide ads and more …

joyful-journey-spring-boot.pdf

 joyful-journey-spring-boot.pdf

Quarkus applications are quick to code, start fast, and require much less hardware than their Spring Boot equivalents. They use less memory, less electricity, and handle much higher load. And, most importantly, they feel great to work with. Fast feedback, great readability, brilliant tools – it’s all about the developer joy.

Can we quantify the financial benefits of switching from Spring Boot to Quarkus? Yes. (Spoiler – it’s good!) In this talk, Holly will explain the architectural differences which make Quarkus fast. Finally, she’ll demonstrate options for getting from Spring to Quarkus quickly and cheaply, for a near-immediate return on investment.

Holly Cummins

November 19, 2026

Transcript

211 slides
  1. The Joyful Journey from Spring Boot to Quarkus ff Holly

    Cummins Quarkus Team | IBM Senior Technical Sta Member | Java Champion
  2. “Pleasure in the job puts perfection in the work.” still

    a misquote Aristotle @holly_cummins #IBM
  3. hey boss, I want work to have more joy. but

    Aristotle? pleasure in the job and perfection in the work? @hollycummins.com #RedHat
  4. hey boss, I want work to have more joy. but

    Aristotle? pleasure in the job and perfection in the work? @hollycummins.com #RedHat
  5. hey boss, I want work to have more joy. but

    Aristotle? pleasure in the job and perfection in the work? ἡδονή? @hollycummins.com #RedHat
  6. hey boss, I want work to have more joy. χρήματα!

    but Aristotle? pleasure in the job and perfection in the work? ἡδονή? @hollycummins.com #RedHat
  7. “30 to 40% better developer productivity vis-a-vis Spring Boot, and

    this is for an ex-Spring Boot developer” – Christos Sotiriou, DXL technical lead at Vodafone Greece
  8. how much hardware does an app need? low load pack

    many instances onto a node memory is the limiting factor high load horizontally scale to support load max throughput is the limiting factor
  9. 300 MiB RAM 300 MiB RAM 300 MiB RAM 300

    MiB RAM machine (1GiB RAM)
  10. low load scenario: cost per instance per year assumptions: EC2

    c6i.xlarge instance, us-east-1, $0.17 per hour
  11. low load scenario: cost per instance per year pp rr

    ii quarkus 1489.2 0.17 spring assumptions: EC2 c6i.xlarge instance, us-east-1, $0.17 per hour
  12. low load scenario: cost per instance per year pp rr

    ii quarkus 1489.2 0.17 US$64.75 spring US$135.38 jvm assumptions: EC2 c6i.xlarge instance, us-east-1, $0.17 per hour
  13. low load scenario: cost per instance per year pp rr

    ii quarkus 1489.2 0.17 jvm native US$64.75 US$19.86 spring US$135.38 US$64.75 assumptions: EC2 c6i.xlarge instance, us-east-1, $0.17 per hour
  14. low load scenario: cost per instance per year pp rr

    ii quarkus 1489.2 0.17 spring jvm native US$64.75 US$19.86 2.1x US$135.38 US$64.75 assumptions: EC2 c6i.xlarge instance, us-east-1, $0.17 per hour
  15. low load scenario: cost per instance per year pp rr

    ii quarkus 1489.2 0.17 spring jvm native US$64.75 US$19.86 2.1x 3.3x US$135.38 US$64.75 assumptions: EC2 c6i.xlarge instance, us-east-1, $0.17 per hour
  16. 15,000 tps machine 10,000 tps 15,000 tps 15,000 tps machine

    machine 20,000 tps 15,000 tps 15,000 tps machine machine 30,000 tps
  17. 15,000 tps machine 10,000 tps 15,000 tps 15,000 tps machine

    machine 20,000 tps 15,000 tps 15,000 tps machine machine 30,000 tps 40,000 tps
  18. 15,000 tps machine 10,000 tps 15,000 tps 15,000 tps machine

    machine 20,000 tps 15,000 tps 15,000 tps 15,000 tps 15,000 tps 15,000 tps machine machine machine machine machine 30,000 tps 40,000 tps
  19. how much can quarkus save you? low load quarkus native

    uses ⅓ the memory of spring native high load quarkus on jvm can handle 1.7x more requests than spring boot
  20. we also see better response times users care (a lot)

    about response time #Quarkus @holly_cummins
  21. https://github.com/quarkusio/quarkus/issues/58302#issuecomment-6791692367 “On a side note: we have just moved this

    application from Spring Boot + Next.js to Quarkus REST + Qute in a native image, and it has been a breath of fresh air. The pages got faster, and the pod went from 512Mi/768Mi (and that was only the Spring backend, Next.js ran separately) to 64Mi/ 128Mi. Thank you all for Quarkus!” @holly_cummins #IBM
  22. https://github.com/quarkusio/quarkus/issues/58302#issuecomment-6791692367 8x less memory for pods “On a side note:

    we have just moved this application from Spring Boot + Next.js to Quarkus REST + Qute in a native image, and it has been a breath of fresh air. The pages got faster, and the pod went from 512Mi/768Mi (and that was only the Spring backend, Next.js ran separately) to 64Mi/ 128Mi. Thank you all for Quarkus!” @holly_cummins #IBM
  23. https://github.com/quarkusio/quarkus/issues/58302#issuecomment-6791692367 8x less memory for pods “On a side note:

    we have just moved this … but also application from Spring Boot + Next.js to Quarkus REST + Qute in a native image, and it has been a comparing native vs breath of fresh air. The pages got faster, and the jvm pod went from 512Mi/768Mi (and that was only the Spring backend, Next.js ran separately) to 64Mi/ 128Mi. Thank you all for Quarkus!” @holly_cummins #IBM
  24. https://github.com/quarkusio/quarkus/issues/58302#issuecomment-6791692367 8x less memory for pods “On a side note:

    we have just moved this … but also application from Spring Boot + Next.js to Quarkus REST + Qute in a native image, and it has been a comparing native vs breath of fresh air. The pages got faster, and the jvm pod went from 512Mi/768Mi (and that was only the Spring backend, Next.js ran separately) to 64Mi/ 128Mi. Thank you all for Quarkus!” @holly_cummins #IBM
  25. airline maintenance scheduling application: “[With Quarkus], we can run 3

    times denser deployments without sacrificing availability and response times of services. ” – Thorsten Pohl @holly_cummins #Quarkus
  26. airline maintenance scheduling application: “[With Quarkus], we can run 3

    times denser deployments without sacrificing availability and response times of services. ” 3x – Thorsten Pohl less hardware @holly_cummins #Quarkus
  27. 2019: moved digital experience to quarkus (from spring boot) Christos

    Sotiriou, DXL technical lead at Vodafone Greece
  28. 2019: moved digital experience to quarkus (from spring boot) •

    developer productivity ⬆ 30-40% Christos Sotiriou, DXL technical lead at Vodafone Greece
  29. 2019: moved digital experience to quarkus (from spring boot) •

    developer productivity ⬆ 30-40% • startup times 40s ➡ 12s Christos Sotiriou, DXL technical lead at Vodafone Greece
  30. 2019: moved digital experience to quarkus (from spring boot) •

    developer productivity ⬆ 30-40% • startup times 40s ➡ 12s • RAM 800 MB ➡ 360 MB Christos Sotiriou, DXL technical lead at Vodafone Greece
  31. 2019: moved digital experience to quarkus (from spring boot) •

    developer productivity ⬆ 30-40% • startup times 40s ➡ 12s • RAM 800 MB ➡ 360 MB Christos Sotiriou, DXL technical lead at Vodafone Greece
  32. 2x less memory 2019: moved digital experience to quarkus (from

    spring boot) • developer productivity ⬆ 30-40% • startup times 40s ➡ 12s • RAM 800 MB ➡ 360 MB Christos Sotiriou, DXL technical lead at Vodafone Greece
  33. 2x less memory 2019: moved digital experience to quarkus (from

    spring boot) • developer productivity ⬆ 30-40% • startup times 40s ➡ 12s • RAM 800 MB ➡ 360 MB 30-40% more productive Christos Sotiriou, DXL technical lead at Vodafone Greece
  34. how does a java framework start? build time runtime packaging

    (maven, gradle…) @holly_cummins #Quarkus
  35. how does a java framework start? build time runtime </>

    load and parse • config files • properties • yaml • xml • etc. @holly_cummins #Quarkus
  36. how does a java framework start? build time runtime </>

    @ @ • classpath scanning and annotation discovery • attempt to load class to enable/disable features @holly_cummins #Quarkus
  37. how does a java framework start? build time runtime </>

    @ @ build a metamodel of the world @holly_cummins #Quarkus
  38. how does a java framework start? build time runtime </>

    @ @ start • thread pools • I/O • etc. @holly_cummins #Quarkus
  39. how does a java framework start? build time runtime </>

    @ @ ready to do work! @holly_cummins #Quarkus
  40. </> @ @ what if we start the application more

    than once? @holly_cummins #Quarkus
  41. </> @ @ </> @ @ what if we start

    the application more than once? @holly_cummins #Quarkus
  42. </> @ @ </> @ @ </> @ @ what

    if we start the application more than once? @holly_cummins #Quarkus
  43. </> @ @ </> @ @ </> @ @ </>

    @ @ what if we start the application more than once? @holly_cummins #Quarkus
  44. </> @ @ </> @ @ </> @ @ </>

    @ @ what if we start the application more than once? @holly_cummins #Quarkus
  45. </> @ @ </> @ @ so much work gets

    redone every time </> @ @ </> @ @ what if we start the application more than once? @holly_cummins #Quarkus
  46. doing more up-front - speeds up start - shrinks memory

    footprint - improves throughput (!) @holly_cummins
  47. do the work at build time reactive core ffi ffi

    e ciency developer experience (also e ciency)
  48. do the work at build time reactive core live reload

    ffi ffi e ciency developer experience (also e ciency)
  49. do the work at build time reactive core live reload

    e ciency developer experience (also e ciency) ffi ffi continuous testing
  50. do the work at build time reactive core e ciency

    minimising boilerplate live reload developer experience (also e ciency) ffi ffi continuous testing
  51. do the work at build time reactive core e ciency

    minimising boilerplate developer experience (also e ciency) ffi dev ui ffi live reload continuous testing
  52. do the work at build time reactive core e ciency

    minimising boilerplate developer experience (also e ciency) ffi dev ui ffi live reload continuous testing
  53. do the work at build time reactive core e ciency

    minimising boilerplate developer experience (also e ciency) ffi dev ui ffi live reload continuous testing
  54. do the work at build time reactive core minimising boilerplate

    e ciency developer experience (also e ciency) dev ui ffi security ffi live reload continuous testing
  55. do the work at build time reactive core minimising boilerplate

    e ciency integrated platform with rapid release cadence ffi ffi security live reload developer experience (also e ciency) dev ui continuous testing
  56. do the work at build time reactive core minimising boilerplate

    e ciency integrated platform with rapid release cadence security ffi ffi tree-shaking live reload developer experience (also e ciency) dev ui continuous testing
  57. do the work at build time reactive core minimising boilerplate

    e ciency integrated platform with rapid release cadence security ffi ffi some library code isolated to buildtime tree-shaking live reload developer experience (also e ciency) dev ui continuous testing
  58. do the work at build time reactive core ffi ffi

    e ciency developer experience (also e ciency)
  59. do the work at build time live reload reactive core

    ffi ffi e ciency developer experience (also e ciency)
  60. do the work at build time live reload reactive core

    e ciency developer experience (also e ciency) ffi ffi continuous testing
  61. do the work at build time reactive core e ciency

    minimising boilerplate live reload developer experience (also e ciency) ffi ffi continuous testing
  62. do the work at build time reactive core e ciency

    minimising boilerplate developer experience (also e ciency) ffi dev ui ffi live reload continuous testing
  63. do the work at build time reactive core e ciency

    minimising boilerplate developer experience (also e ciency) ffi dev ui ffi live reload continuous testing
  64. do the work at build time reactive core e ciency

    minimising boilerplate developer experience (also e ciency) ffi dev ui ffi live reload continuous testing
  65. “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
  66. claude -p "Read skills/migrate-spring-to-quarkus/ SKILL.md \ and execute it. \

    Source: ~/tmp/spring-petclinic \ Target: ~/tmp/spring-petclinic-quarkus \ Strategy: full-quarkus, Java 21, \ persistence: panache-active-record. \ Proceed autonomously."
  67. The boundary Deterministic rules AI • Reproducible • Flexible •

    Cheap (zero tokens) • Expensive (tokens) • Auditable • Variable • Same input = same output • Needs context to shine
  68. How we built the Migration skill Prerequisite Planning Build Code

    Testing Cleanup Messaging Frontend Module Gate condition Compile check Prerequisite Always — abort if JDK < 17 — Planning Always — writes migration-spec.yaml — Build Spring Boot markers in build file ✓ Code Spring annotations in source ✓ Messaging @KafkaListener, @RabbitListener, etc. ✓ Frontend Thymeleaf, JSP, FreeMarker templates ✓ Testing @SpringBootTest in test sources ✓ Cleanup Always ✓
  69. hybrid: deterministic + AI DURING BEFORE AFTER AI Skill Spring

    Boot Project Validators Context-dependent work Semantic verification • • • • • • • Project setup • Persistence (entities, relations, ID strategy, tables) • REST endpoints • Config • Messaging Controller logic Template conversion Config reasoning Strategy decisions Test migration Edge cases 0 tokens · auditable tokens · flexible · variable Analysers feedback loop (if validation fails) MCP: deterministic tools exposed as discoverable tools for the AI agent Verified Quarkus Project
  70. tl;dpa (too long didn’t pay attention) cheaper at low load

    cheaper at high load greener more secure
  71. tl;dpa (too long didn’t pay attention) cheaper at low load

    cheaper at high load greener more secure more productive
  72. tl;dpa (too long didn’t pay attention) cheaper at low load

    cheaper at high load greener more secure more productive happier
  73. stl;dpa (still too long didn’t pay attention) e ciency quarkus

    can be packed ~2x more densely than spring boot quarkus can handle ~2x more tra c on the same hardware ffi ffi @holly_cummins developer experience quarkus creates joy quarkus boosts productivity by #Quarkus
  74. stl;dpa (still too long didn’t pay attention) e ciency 2x

    quarkus can be packed ~2x more densely than spring boot less infra quarkus can handle ~2x more tra c on the same hardware ffi ffi @holly_cummins developer experience quarkus creates joy quarkus boosts productivity by #Quarkus
  75. stl;dpa (still too long didn’t pay attention) e ciency 2x

    quarkus can be packed ~2x more densely than spring boot less infra quarkus can handle ~2x more tra c on the same hardware ffi ffi @holly_cummins developer experience 40% quarkus creates joy quarkusmore boosts productivity by productive #Quarkus
  76. tl;dpa happy, productive people (too long didn’t pay attention) frictionless

    development experience deployment density auto-provision services zero-con g live coding continuous testing developer UI lower cloud bill Medium Nano greener fi @holly_cummins #Quarkus

More Decks by Holly Cummins

Other Decks in Programming