Saved 55% cost, 75% CPU, at 33x speed - Why and how Jira moved from JSON to Protobuf

[@Nischal.shetty02]

The title looks like my resume after ATS told me to add stats on my project

[@rbansal42]

It works though😂

[@jiraiyasensei1914]

Amazing way to start my Sunday

[@vishalreddy6834]

Can you please share the notes which you are showing in the video ?

[@sharma01ketan]

The title shows that Arpit sir now works on AdTech :)

[@pratipbhattacharyya552]

Great Explanation!

[@chansecode821]

Weren't they using GZipped JSON earlier? Did the payload size reduced to 80% from GZipped JSON to Protobuf?

[@explain-tvv]

nice explaination. which app you use to teach?

[@theashishmaurya]

One dumb question, why they keep writting to Memcached everytime a request is not found? Instead of migrating all the data one time directly? What purpose this incremental migration serves and what are the consequences if not done in this way?

[@Genji_hehe]

To keep the cache warm and updated ig? You dont have infinite cache to move all the data at once, you either you hit or miss and ask the db. Any cold data will be flushed out at some point.

[@indianengineer5802]

Arpit according to you when generally Avro is preferred over Protobuf ? I rarely see getting them used except in Kafka

[@AsliEngineering]

Avro is typically preferred when schemas need to be determined at runtime (kind of schemaless workloads) because Avro includes the complete schema with the data. Protobuf requires you to define schema beforehand and then highly compacts the data with field tags. Hence Avro is used in Big Data (esp Data Lakes - some usecases) where the schema is stored along with data. Hope this helps.

[@indianengineer5802]

@@AsliEngineering Thanks Arpit, got it !

[@projectmayhem72]

Bro how do you record with ipad? I would highly want to know as I want to build a streaming channel with ipad notes

[@tanmayaeron8230]

how are updates on an issue handled? there has to be a process maybe async sending writes to issue service and not waiting for a read request for db to become consistent.

[@thehungrybird]

I think they are overwriting the updates in issues db and later syncing them in main db

[@AashutoshRathi]

"Jira has issues", lol.

[@EdwinManual]

this title is ATS optimised

[@aniketjain3656]

Hey😊

[@AbhishekSingh-nm4gw]

Very unlikely. The video title is a click bait. I just talked to folks at Jira, and they said there is no credible evidence to back this claim.

[@AsliEngineering]

Blog hai bhai unka. Check the link below the description.

[@AndrewMcClenaghan4182]

I can’t see the link to an Atlassian blog or similar about this. Do you have one? Curious to read that as well

[@amit_go]

Even I switched from JSON to MessagePack for internal messaging in Portsicle. MessagePack reduced the compute time by 70% compared to json.