Im still trying to understand the partition part of CAP, is that the distributed phase? If a DB is distributed, each version becomes a partition, and then the need for some sort of synchronization occurs to provide consistency so everyone sees the same data?
Yeah, I think you got it. In the context of CAP, we're already talking about a system that is partition-tolerant; i.e. distributed. CAP is really: in a partition-tolerant system (distributed), you need to make a tradeoff between consistency (everyone sees the same data) and availability (everyone can use the system). But you can't have both. An example of partition-tolerant are two databases in physically different locations (even different countries) that need to have the same data (synchronised).
How does a master-slave architecture then imply that data is consistent? It depends on the specific consensus protocol. Obviously if you read from a slave before it has time to copy over the txn logs, the data will be stale.
I suspect, in master-slave cases, the client will only read/write from the master node. The slaves will replicate the data and be ready to replace the master in case of partition failure.
It certainly would @petchpaitoon . The guidelines for this Theorem that should be a consideration for all things distributed which especially applies for the cloud native world. For me, I would say that it accomplishes A-P, where the Consistency happens eventually as data is synchronized across partitions or replicated to individual instances. There is also a special config in Kafka to tell it to acknowledge receipt once replication has occurred between all the brokers ( a basic Kafka Cluster consists of 3 brokers ). Let me know if you have any further questions or have any thoughts on any new videos I could do to further go deeper in this topic.
Explanation is great. Also I like the background combined with this black t-shirt and the invisible whiteboard.
Im still trying to understand the partition part of CAP, is that the distributed phase? If a DB is distributed, each version becomes a partition, and then the need for some sort of synchronization occurs to provide consistency so everyone sees the same data?
Yeah, I think you got it. In the context of CAP, we're already talking about a system that is partition-tolerant; i.e. distributed.
CAP is really: in a partition-tolerant system (distributed), you need to make a tradeoff between consistency (everyone sees the same data) and availability (everyone can use the system). But you can't have both.
An example of partition-tolerant are two databases in physically different locations (even different countries) that need to have the same data (synchronised).
I want to commend you on your ability to draw sdrawkcab so well.
Even if you dellepssim availability.
very clever :-) Yes it happens sometimes the sgnillepssim when writing and mixing colors, etc.
best video on this topic, no cap
this guy is the best! thanks so much! learning so much from you
The content was great, but that's an 11/10 for writing backwards with such neat hand writing!
Mirrored Video?
This man is so cool !
I really want to meet him
Damn man...
example DB of CA?
Probably it is not possible... but I am not sure...
How does a master-slave architecture then imply that data is consistent? It depends on the specific consensus protocol. Obviously if you read from a slave before it has time to copy over the txn logs, the data will be stale.
I suspect, in master-slave cases, the client will only read/write from the master node. The slaves will replicate the data and be ready to replace the master in case of partition failure.
Superb video!
First person to learn
What about CA?
Great explanation, thx!
CA is not possible at all right?
mans looks like a floating head with arms
What about the Confluent Kafka that work as data store ? What kind of CAP that it will archeive ?
It certainly would @petchpaitoon . The guidelines for this Theorem that should be a consideration for all things distributed which especially applies for the cloud native world. For me, I would say that it accomplishes A-P, where the Consistency happens eventually as data is synchronized across partitions or replicated to individual instances. There is also a special config in Kafka to tell it to acknowledge receipt once replication has occurred between all the brokers ( a basic Kafka Cluster consists of 3 brokers ). Let me know if you have any further questions or have any thoughts on any new videos I could do to further go deeper in this topic.
Availability **
Thanks Abhinav, yes looks like I missed that one in the heat of the moment :-)
Still trying to understand how this guy writes in reverse
Set a camera behind a glass wall and film it, and then flip the video horizontally
@@yasu9493 Oooh sneaky sneaky
its availAbility
nice
Notice he is writing in the opposite direction xD, It would be hard for me.
We flip it in post-production. Search on "lightboard videos" for more details
Epic .
Not clear ....not a good video sorry