Google Docs Design Deep Dive with Google SWE! | Systems Design Interview Question 13

[@Snehilw]

Not true dude, I listen to all your videos sometimes multiple times end to end and know a few friends who do too. Keep up the good work!

[@Ms452123]

Do you think you making PowerPoints of the concepts that you covered in this entire series helped you with your long term understanding of the fundamentals. Essentially can creating a PowerPoint on a topic be a high yield method of learning.

[@jordanhasnolife5163]

I believe that writing things down that you're trying to learn has proven to be one of the most effective methods of studying

[@theghostwhowalk]

Great explanation!!

[@utkarshgupta2909]

Jordon can you please make a video series on databases... We guys really have lot of problem choosing the right database sometimes.

[@jordanhasnolife5163]

Perhaps I shall

[@shubhankar915]

Can we use cassandra to store the document changes in Operational transformation as well? What advantages does HBase give over cassandra! Thanks!

[@jordanhasnolife5163]

HBase is column oriented storage which is nice if you want to get lower latency just getting the characters in the doc as opposed to a bunch of metadata which will probably be in the db

[@mann8737]

Hey Jordan! First, Your videos are really helping me out for interview prep. I have a question tho about HBase. How would the HBase schema look like for storing documents? What is going to be the row key, sort key and what about the columns?

[@jordanhasnolife5163]

Yeah so think of it like this: Partition key: documentId Sort key: documentIndex Columns: characterAtIndex

[@yusufsipahi3916]

hello @@jordanhasnolife5163 , I want to clarify the db schema for an event of an insertion in OT. Let's say our server gets an insert request for a document at index "i". In that case I believe this request does not only effect documentIndex at "i" but we also need to propagate this change to rest of the indexes by incrementing them by 1 in our db. Don't you think this has negative impact on db writes?

[@jordanhasnolife5163]

@@yusufsipahi3916 In theory sure, in practice you could use like a linked list or something

[@hemanthaugust7217]

So, based on the schema given below, what is documentIndex? I understand documentId is unique for entire document. Is docIndex similar to pageNumber? So, each docId has multiple docIndexes? Does each character become a separate Row in the DB? Isn't it a lot? Can't we store entire line as a Row?

[@yusufsipahi3916]

how would you change your design for video/image insertion into the documents?

[@jordanhasnolife5163]

Barely to be honest, you're just adding a link to s3 instead of text

[@2tce]

So what would you handle character deletions as opposed to insertions? Just delete that index?

[@jordanhasnolife5163]

Yes but deletions are tricky in the sense that they require FIFO broadcast - you can't delete a character unless it already exists, so you need to make sure that at least from a single client, all of the operations are being broadcast in the correct order

[@Awaytake]

Credits: Martin Kleppmann (kzbin.info/www/bejne/r3TGiKOwjp2Xf6M)

[@jordanhasnolife5163]

I'm a dirty plagiarizer!

[@HustlerMS]

Nah, your content is superb. I posted the link for someone who might be interested in that talk. Thanks for uploading.

[@jordanhasnolife5163]

@@HustlerMS haha thanks man, was just messing around I appreciate you attaching it

[@mhp647]

it would be way more easy to follow if you would have some images or slides on screen to focus on them, not your face

[@jordanhasnolife5163]

Remaking all of these starting in a month, that's the plan