Great overview Keerti! You are supplementing the discussion with the necessary depth and direction. Frankly speaking Gaurav needs to up his game - he clearly lacks some depth and breadth.
I suggest that instead of simply stating the calculations of capacity estimates, we should perform calculations to determine them accurately. Additionally, the worldwide bandwidth calculation appears to be incorrect, it should be 300 GBPS instead of 4 GBPS. By the way video provided valuable insights into various network protocols, enriching my understanding significantly. Keep up the commendable efforts in sharing informative content.
18:10, instead of using RTMP over TCP, i feel for low latency streaming, we should use DTLS(datagram transport layer security). it is built on datagram protocol. To provide encryption, message authentication, and integrity for datagram-based applications that use unreliable transport protocols such as UDP . DTLS is effectively the datagram-equivalent of TLS , the protocol commonly used to secure communications over TCP. Here reliability challenge is still in DTLS(because of underline UDP) but RTMP is reliable(underline TCP). For Low latency but reliability doesn't matter: DTLS for High reliable : RTMP @gkcs and @keerti, can you please provide feedback on this.
A and B could be machines on local network (behind a NAT) so they don’t have an external ip address. How can A and B initiate a peer to peer communication.
sorry cant stop myself from commenting , you didnt even scratch the surface. May work out for Associate SDE position OR could work out for a PM role interview. What happens in case of ~500 people conference? Any tree algorithms useful here like minimum spanning tree? What happens insode RTMP server? Any Queue or Stream processing, any notes around adaptive bit rate?
About the part where we are using media server for webinars for broadcasting to all connected users let's say 100, for each user, how do we do that? How do we manage the load on the server?
how about publishing IP address periodically to Zoom server.. There may be more challenges if we think it through. Like establishing connection to 1000's of connection with people on the call.
@@RajnishKumar-hr5om There is a call server ( which solves all these problems like dynamic IPs address or lets say if in that particular port many users are accessing then symmetic NAT block these and hence peer to peer connections goes in vain ) which we called as TURN in webRtc protocol
Perfect! Super interesting. Love this way of teaching system design.
good details shared. I liked.
Great overview Keerti! You are supplementing the discussion with the necessary depth and direction. Frankly speaking Gaurav needs to up his game - he clearly lacks some depth and breadth.
I suggest that instead of simply stating the calculations of capacity estimates, we should perform calculations to determine them accurately. Additionally, the worldwide bandwidth calculation appears to be incorrect, it should be 300 GBPS instead of 4 GBPS.
By the way video provided valuable insights into various network protocols, enriching my understanding significantly. Keep up the commendable efforts in sharing informative content.
18:10, instead of using RTMP over TCP, i feel for low latency streaming, we should use DTLS(datagram transport layer security). it is built on datagram protocol. To provide encryption, message authentication, and integrity for datagram-based
applications that use unreliable transport protocols such as UDP . DTLS is effectively the datagram-equivalent of TLS , the protocol commonly used to secure communications over TCP.
Here reliability challenge is still in DTLS(because of underline UDP) but RTMP is reliable(underline TCP).
For Low latency but reliability doesn't matter: DTLS
for High reliable : RTMP
@gkcs and @keerti, can you please provide feedback on this.
Distributed Locking, its use case & implementation, trade off, bottle neck etc.
A and B could be machines on local network (behind a NAT) so they don’t have an external ip address. How can A and B initiate a peer to peer communication.
amazing stuffs mam
Nice video.... thanks
What do they store in DB?. Like is it only the meeting metadata and userDetails. or there are any specific strategy to store the DB design?
sorry cant stop myself from commenting , you didnt even scratch the surface. May work out for Associate SDE position OR could work out for a PM role interview. What happens in case of ~500 people conference? Any tree algorithms useful here like minimum spanning tree? What happens insode RTMP server? Any Queue or Stream processing, any notes around adaptive bit rate?
About the part where we are using media server for webinars for broadcasting to all connected users let's say 100, for each user, how do we do that? How do we manage the load on the server?
Address of A or Address B can change any time, in that case what will happen. ISP did not provide static address for user it will change in few hours.
how about publishing IP address periodically to Zoom server.. There may be more challenges if we think it through. Like establishing connection to 1000's of connection with people on the call.
@@RajnishKumar-hr5om There is a call server ( which solves all these problems like dynamic IPs address or lets say if in that particular port many users are accessing then symmetic NAT block these and hence peer to peer connections goes in vain ) which we called as TURN in webRtc protocol
Hi Keerti, i want to join the HLD course, but 5th april date is passed and course is already started, when will new batch will start for HLD ?
👍👍👍
Isn't that 30GB per day?
There is no use of backof the estimate at staff level. Its useless calculation and gives you nothing...
Hello,
May i know what's your opinion on BS from IITM. Please reply
it's great...
LinkedIn system design
Your editing is good but TH-cam Thumbnail is very low quality.
I will create a Thumbnail for you.
Boring Thumbnail=Less Views