How does Uber scale to millions of concurrent requests?
Vložit
- čas přidán 27. 07. 2024
- Fulfillment Platform Rearchitecture → goo.gle/35dJrMr
Building Uber’s Fulfillment Platform → goo.gle/3fJ6rEX
Cloud Spanner → goo.gle/3qLIovs
How does Uber scale millions of concurrent requests for rides and food deliveries while keeping costs under control? In this episode of Architecting with Google Cloud, Priyanka Vergadia interviews Ankit Srivastava, Senior Staff Engineer at Uber, on how to get a ride or order food with the “simple” press of a button. Watch to learn how Ankit Srivastava and the Uber team led the ground-up redesign of Uber’s fulfillment platform leveraging Cloud Spanner.
Chapters:
0:00 - Intro
1:00 - What powers the simple user experience at Uber?
2:40 - How does the Uber request flow work?
3:35 - Uber’s challenges with on-premises database & why they switched to Cloud Spanner
5:50 - How did Uber transition from NoSQL to NewSQL - Spanner
7:24 - In a hybrid environment, how does Uber reduce latency?
9:00 - What were the benefits of Spanner to Uber?
9:35 - How did Uber migrate to Spanner live - while we were still ordering rides and food?
11:53 - How Google team helped Uber migrate their on-premise database to Spanner.
14:30 - What were the performance optimizations done by Uber to deliver the user experience?
16:20 - How do you optimize for cost at this scale?
18:30 - Summary
Ankit Srivastava’s LinkedIn → goo.gle/3qOclek
Priyanka Vergadia’s LinkedIn → goo.gle/3rHcFKW
Architecting with Google Cloud playlist → goo.gle/3g7xAC9
Subscribe to Google Cloud Tech → goo.gle/GoogleCloudTech
#ArchitectingWithGoogleCloud
product: Cloud - General; fullname: Priyanka Vergadia; - Věda a technologie
Demo of the new text to speech seems impressive!
😂😂😂
😂
Ha ha
😂
😁😁😁😁😁😁😁😁😁😁😁
The team have been doing absolutely amazing to keep the service running with no downtime. And really amazed to know Uber using all these tools during migration.
It's totally incredible how some of the scaling considerations were handled and seamlessly migrated between the two underlying architectures. Kudos to the teams that enabled this!!
I love these use cases. They are great in many ways. We can gain an understanding of how a system as big as uber works, understand its challenges and and understand how different Google Cloud Platform technologies can help on this. I vote for more videos like this. o/
Loved the fact that they were tackling with the exact same problems like we do day to day… saga pattern, eventual consistency.. and also stumbling upon similar solutions. So relatable and glad they chose spanner. Next logical step is to pipeline that data into BQ and run your ML models 😊
Loved the use case. Need more videos like this
I love the host’s exuberance, great video!
Thank you very much 🙂
Amazing . This is truly planet scale!
Love this video, especially hearing the reality of maintaining operations through the continuous improvement and rollout processes. I am very curious about the Geospatial Index and the matching engine components.
Interesting watching this as an ex Enterprise Architect and now a Lawyer! Good to see these new solution architectures.
one of the best videos in youtube right now.
Really great questions!
awesome use case! love it :)
honestly, just the concept of being able to handle millions of traffic is so exciting...but at the same time with all the networking and distributed architecture involved it's so complex to understand it. I was able to grasp only 30 - 35 percentage of it. I wonder about the Spanner's USPs compared to AWS Aurora
Great use case introduction !
Demo of the new text to speech seems impressive! Good work Google
An interesting topic on spanner
Very useful and nicely explained video.Can you do a video on any Social media like Facebook , Instagram using GCP products.
Great explanation!!
Glad you liked it
it did!!! thank you for this!
amazing , got goose bumps while watching. well done to Ankit. , Preetam, and Priyanka and team.
fact that spanner gives ACID guarantees on planet scale is testament to pros that build this tech
Wow 😳
How the new google cloud Alloy db different from Cloud spanner ? Now google have 3 different relational DB's- Cloud SQL, Cloud spanner and Alloy DB. What is the intent behind all these DB's ?
Can I get the PPT from where the engineers are reading this.
When we say moving the completed orders to new system powered by spanner and new orders will also be handled by new system powered by spanner, how do we handle auto increment keys conflict ? ( Just asking if in case we have auto increment keys ), one work around is to have a gap of a estimated numbers in New database i.e. say old db current order no is 1000 in New db we start with 2000 so that when this db is up and old data from previous db is copied over there will be no key conflict.
You avoid that by not using auto increment keys, you use randomized UUIDs for primary keys. They already came from distributed nosql so that didn't change.
Does Uber has any third party agents who can charge only 50 - 60% of original fee from customers and book orders on their behalf… any thoughts here pls ?
Is there any use case in Africa companies
I really like the part "...You literally changing the car tyres when the car is moving..." 🙂
In terms of human resources, how many people have worked to achieve these results? Thank you
Infrastructure or development?
@@chandrakanthotkar7262 Both ... in total ... thank you
Uber could have done a better job at explaining business needs to Uber Request flow and what/why inconsistencies were created that required fixing with Saga or why serialization is needed.
is he reading ppt presentation?
UBER’s new flawed NON-Anonymous rating system.
With the new UBER rating system. Drivers are required to rate passengers immediately upon the end of a trip. Rate the passenger that just got out of your car before you can advance. The passenger has up to a month or more to rate the drivers that took them on trips.
So, now with the new system. The passenger is able to see the number of each star that composes their total rating. Therefore. If passenger had zero 3 star ratings before taking a trip. Then after the trip the passenger has a 3 star rating. The passenger knows exactly which driver gave the 3 start rating.
This will lead to an automatic retaliatory rating. The retaliatory rating will now doubt be a 1 star rating for the driver.
This means each driver will need to issue nothing other than 5 star ratings from this point forward in fear of retaliatory ratings.
This new system with into effect 2/17/22 and has already become a huge problem for the drivers.
Uber succeeded by ditching responsibility for drivers and consumers. Taxi drivers and passengers in old business model had employee insurance, companies cared about workers and clients. Uber makes billions while not providing any of that. Everyone loses
🤔
Just an observation, Indians are present everywhere on the top(IT, Pharma, Scientists, Banks, Industrialists etc) in the middle and in bottom as well, what happens if all NRI comes in India and focus on building and using their expertise in making India great. Would love to hear your opinions, suggestions or consequences.
Uber is amazon . aren't they using AWS?
why google is excited for Uber's/ AWS's success story?
For transactional Database they are using Spanner, One of the best Hybrid Cloud User Story