great design, thanks! have been bingeing your videos before my Meta interview, appreciate that they go into more depth than even a lot of paid content out there.
Is the reason for a hexagon grid by Uber the lower number of neighbours? Max 7 hexagons to check (1 target and 6 neighbours). Where with the rectangles you have 9 to check (1 target and 8 neighbours).
Could be that, I think they're also just more similar to a circle so finding all points within your close hexagon/pentagon better approximates doing so with a circle without doing as much server side filtering afterwards.
As of now, no. In the future, I'll perhaps create more spaces to talk about this stuff, but I'm generally low on time atm and checking that would spread me a bit too thin. Thanks for the suggestion though!
Hi Jordan, I think I am misunderstanding the concept of bytes, but I have a question about the capacity estimates - in this video, you said that you might store 4 bytes for geohash index for example. In previous videos, you've mentioned that a character can be 2 bytes long. As a result, are you basically saying that 4 bytes for a geo-hash will result in max 2 characters? Or in this case, are you assuming without the extended unicode and you are assuming 1 byte / character, meaning our geohash would be at max 4 characters?
One of the best system design series out there! Keep up the good work, man!
You deserve way more subscribers man!
1 milion Uber drivers in US. 15 millionn Uber rides per day. Awesome talk -- Thanks!!
Of course thanks for watching!
great design, thanks! have been bingeing your videos before my Meta interview, appreciate that they go into more depth than even a lot of paid content out there.
Thanks man! That's the goal
Is the reason for a hexagon grid by Uber the lower number of neighbours? Max 7 hexagons to check (1 target and 6 neighbours). Where with the rectangles you have 9 to check (1 target and 8 neighbours).
Could be that, I think they're also just more similar to a circle so finding all points within your close hexagon/pentagon better approximates doing so with a circle without doing as much server side filtering afterwards.
didnt know about the compaction, we just use a mapping service, great info as usual. nicely explained
Yeah it's a cool one. I think that the next go around I'll devote a full video to google maps!
6:42 Geospatial indices
7:31 in-memory storage, partitioning
10:03 updating locations
10:57 Ride request
12:33 Routing: Djikstra's, A*, Compaction (contraction?) hierarchies
20:10 Post-ride route calc
23:01 Architecture
Thanks for that!
Do you still have the references you talked about in the video?.. Not seeing them in the description. HMMs are very interesting for sure!
highscalability.com/blog/2022/1/25/designing-uber.html
Can you do a video on how the payment system would work?
Sure
Another great overview. I really appreciate the effort you put into these.
Thanks for watching!
based and meat-bang-out-pilled
They call me the butcher cause of how I treat my meat
Helping a lot. Thank You.
By the way, is there any discord channel or other platform to get help?
As of now, no. In the future, I'll perhaps create more spaces to talk about this stuff, but I'm generally low on time atm and checking that would spread me a bit too thin. Thanks for the suggestion though!
@@jordanhasnolife5163 I can create a Slack space for this. For those who want to ask/collaborate on these distributed systems challenges. 🤔
Hmmm, what was that cheat code, again ? 🤣
😎
Hi Jordan, I think I am misunderstanding the concept of bytes, but I have a question about the capacity estimates - in this video, you said that you might store 4 bytes for geohash index for example. In previous videos, you've mentioned that a character can be 2 bytes long.
As a result, are you basically saying that 4 bytes for a geo-hash will result in max 2 characters? Or in this case, are you assuming without the extended unicode and you are assuming 1 byte / character, meaning our geohash would be at max 4 characters?
Hey Stephen! I'm saying that we can use a single integer as our geohash index (more so like a quadtree), hence 4 bytes (32 bits) :)