Synchronous and Asynchronous Communication between Microservices

System Design for SDE-2 and above: arpitbhayani.me/masterclass
System Design for Beginners: arpitbhayani.me/sys-design
Redis Internals: arpitbhayani.me/redis
Build Your Own Redis / DNS / BitTorrent / SQLite - with CodeCrafters.
Sign up and get 40% off - app.codecrafters.io/join?via=...
In this video, I discussed the importance of communication patterns in microservices. Synchronous communication, where services interact in real-time via API calls, ensures completeness but can lead to blocking and scalability issues. Advantages include real-time updates, while disadvantages involve blocking calls and server provisioning challenges. Cascading failures and strong coupling between services are also risks. Synchronous communication is crucial for tasks like database queries and real-time interactions, ensuring immediate responses in scenarios like chat or e-commerce checkouts.
Recommended videos and playlists
If you liked this video, you will find the following videos and playlists helpful
System Design: • PostgreSQL connection ...
Designing Microservices: • Advantages of adopting...
Database Engineering: • How nested loop, hash,...
Concurrency In-depth: • How to write efficient...
Research paper dissections: • The Google File System...
Outage Dissections: • Dissecting GitHub Outa...
Hash Table Internals: • Internal Structure of ...
Bittorrent Internals: • Introduction to BitTor...
Things you will find amusing
Knowledge Base: arpitbhayani.me/knowledge-base
Bookshelf: arpitbhayani.me/bookshelf
Papershelf: arpitbhayani.me/papershelf
Other socials
I keep writing and sharing my practical experience and learnings every day, so if you resonate then follow along. I keep it no fluff.
LinkedIn: / arpitbhayani
Twitter: / arpit_bhayani
Weekly Newsletter: arpit.substack.com
Thank you for watching and supporting! it means a ton.
I am on a mission to bring out the best engineering stories from around the world and make you all fall in
love with engineering. If you resonate with this then follow along, I always keep it no-fluff.

[@theSDE2]

Your content is a goldmine for backend engineers. Keep posting more detailed videos. Thanks

[@mayurikonde9723]

very well explained....thank you

[@shriramparamesh]

fantastic explanations

[@sahilpuri9545]

Hi Arpit , Amazing content and explanation :). I have a query regarding one of your synchronous calls example - Payment services. I think i have seen asynchronous calls being used for Payments now a days (Example - Amazon) since PGs have higher latencies so its better to process the payment/order initially for user experience and then use a webhook for updated status later. Thats why Amazon gives a message after order completion incase payment failed from PG and we need to re-do the payment.

[@AsliEngineering]

Yes. The payment processing can also be done asynchronously. Great observation.

[@yognirog]

Best explanation, thanks Arpit

[@safwanmansuri3203]

Bhai what an explanation kudos 👏

[@d4devotion]

How I did not come to know about your channel before :) . Nice bro.

[@SaketAnandPage]

What do you mean by , because of network partitioning the service is not available?

[@jayeshdalal7]

Aript , let's suppose if any service consume the message with in time even after retry , and we pushed those message in back in queue , can we use cron job so that if let's suppose those all transaction which status is not chanage with in 1 hr they can again process ?

[@gunahawk6893]

Hi arpit iam a py dev, for backend which is the best framework for learning lld Or hld

[@srawat1212]

I feel this info is widely known by software engineers. Can you please make a video on blocking vs non-blocking call ? I think many people confuse it with sync/async.

[@sainihith]

@Arpit Bhayani For Asynchronous communication if multiple services are dependent on one microservice is it better to use multiple message queues for each microservice or use pub/sub mechanism for communication

[@AsliEngineering]

Pubsub/message stream would be a better choice.

[@jayeshdalal7]

@@AsliEngineering pubsub or different topic consume should also works cmiiw here.

[@ankitjhunjhunwala9895]

We return response to user only after getting acknowledgment from the queue, how to handle if there is some issue with the message broker and it's crashing, in this case should we return 200 to user or 500? Can't we avoid waiting for an acknowledgement from queue before returning response to user, and manage error in queue(while putting message) by monitoring logs, since the reaction is already registered in db?

[@AsliEngineering]

No you should not. If you are unable to persist in queue then return an error to user

[@shervilgupta92]

Thanks for the amazing content as always, just one query regarding async communication: Do we always need a queue? Event loop in js and reactive programming in spring, do they also fall under the category of async communication ?

[@AsliEngineering]

Yes. We always need a queue for async communication between services. Event loop is async intra-process although similar concept but different usecases.

[@shervilgupta92]

@@AsliEngineering Got it, thanks !

[@sartajsingh9949]

Amazing video but one doubt ..if we have multiple reaction-service nodes then how the request forwarding would happen without a load-balancer?

[@Avinashkk360]

Reaction service will be under load balancer. What was mentioned is, if we use aysnc communication, we do not need load balancer for notification service.

[@uditsharma2536]

Aprit, I am curious to know how Async communication can make the loose coupling between the services. Because in that scenario, if anything changes, you need to inform the caller, Whether it is schema changes, API changes, or event definition changes.

[@AsliEngineering]

But so long as communication interface does not change each service can independently be developed. This is loose coupling. Go through my Microservices playlist to understand it in depth

[@uditsharma2536]

thanks for the quick reply. Your video was really good and informative. I understand the loose coupling concept. But I am just curious to know how async communication makes things loosely coupled.

[@lifeisgood7935]

Can we use redis cache in reaction service to store messages in case notification service is down? This will avoid the need to have queue in between. Could you talk about the pros and cons of this approach?

[@charan775]

but then the implementation logic for retrying falls under each service. everyone will needs to implement it and make sure it will definitely retry.

[@iamnoob7593]

dude , redis cahce is no use here , if notif service is down , then the message will be there in queue until it is consumed and processed successfully . if there is issue in processing in notif , then as per design if it will retry maybe 3 times , still no success , then the message would be in queue , and not evicted. This depends on which message broker , then developer/infra will have to fix the defect in the notification service if any causing it or due to infra issues.

[@vaibhavmehta36]

Arpit, can we have a video on components generally used in system design with pros & cons of each component so that we know when to use Kafka and when to use RabbitMQ

[@AsliEngineering]

Something I cover in my course. Would not be possible to cover it on YT. Sorry. But you can definitely find a ton of resources on this topic.

[@prerakhere]

correction at 18:32 synchronous instead of asynchronous.

[@aditijalaj5036]

how is the communication over HTTP between two services Synchronous if we make a call and Await the response using Promises?

[@AsliEngineering]

You are blending two very different concepts here. Unfortunately both are named Asynchronous. Async processing mentioned in the video is about delegating actual task processing. In case of async http, it is about not letting your process get blocked on getting the response.

[@Aahnik]

sir, i had a doubt... in real life how this is implemented ? multiple processes consuming from a single queue / message broker ? suppose only one service is consuming from the queue, then after consuming, the queue can probably discard that... in case of multiple services, what pattern or technique is used ?