AWS re:Invent 2021 - DynamoDB deep dive: Advanced design patterns

[@iNemoden]

This is the only lecture I don’t need a x1.5/x2 speed for 😂

[@badscrew4023]

Agreed, even at 1x I needed to pause the video many times to process the information :D

[@christopher5731]

had to turn this to 0.75x

[@Pokepasa]

@@christopher5731 same here!!

[@chrismiller3591]

This is one hell of a stream of consciousness. Rick knows his stuff.

[@chanyt2364]

Most productive 50 minutes of learning I have had in a long time. Great breakdown.

[@amazonwebservices]

AMAZING to hear that! 🤩 Keep Learning! 🙌

[@shammyh]

Excellent fast paced summary of DynDB and NoSQL broadly! Thank you!

[@mikeyinger4204]

I watched an earlier single table design video, but the resolution of your screen sessions was so low that I couldn't see what you were typing. I learned a lot, because I had to watch it quite a few time in a way I wouldn't have, had I been able to see the chars you typed. I designed a single table. This talk - a fun ride through Info/AWS/DDB arch to table design and coding, all the time explaining what to use and how you map from the relational model... in less than an hour. Wow!

[@bhomiktakhar8226]

Excellent outline on why Nsql came into the world and how dynamodb is so elegant in handling traffic like a baller.

[@CyprienSafi]

At 16:19 the operations that can be performed on a key are listed. 'contains', 'in', and 'count' are not correct except in filter (which is the case for all attributes)

[@christiangrundemann9843]

i feel stupid now

[@SkaChild]

Great talk, one of my biggest takeaways from this and from other dynamodb talks is that it's OK to over select, I.e your pk,sk-condition might not hit exactly what you want, but its OK to throw away some percentage of items after that due to the fact that the rcu is rounded up AFTER returning items

[@HanyunGong]

Maybe it's just me, but for the booking system example, it seems much simipler using SQL in general on any scale V.S. DynamoDB+Stream+Lambda ?

[@serhii.almazov]

For in-house booking with 100 customers - you're definitely right. But imagine building something at a scale of the google calendar

[@realpillboxer]

@@serhii.almazov Very few products need to exist at Google scale. Don't get me wrong -- having options of using traditional RDBMS or a more complex cloud-native stack is great. However, knowing your use cases, access patterns, extensibility, budgetary concerns, enterprise or department strategy, and developer skill sets will get you a lot further than cramming your product into the wrong ecosystem.

[@serhii.almazov]

​@@realpillboxer Completely agree. E.g. I would not recommend going microservices to the small startup when they are in a search of their customers. But once that startup has launched, got traction, and needs to scale up the development and the infrastructure - that's when the microservices and scaling can be important. Same for other technologies - start as simple as possible, and replace it with more scalable/low-latency/highly available/etc but also more complex technology when needed

[@BatterRuslan]

And it's not about example only. If you'll take any other real apps, you'll see that writes and mutations will make your project full of distributed locks or need of sharding. Which will make development extremely expensive.

[@GamingCentralTV1]

I like this guy lmao. He is probably great to work for. He is clearly pro databases not pro aws. If there was a better solution out there best believe he will he honest about it. Thank you for the insightful lecture.

[@ianokay]

Is it not odd to discuss SQL in terms of JOIN algorithm time complexity?

[@DomGarzaSR]

And people were talking down on me for choosing dynamo db! I have 4 tables and I know my data!!!!

[@Rickety3263]

BEST. DATABASE. EXPLANATION. EVER. Now I’m going to re-watch this at 1/2 speed

[@seymabiskin8728]

Hi, can I trigger the same lambda from different Dynamodb streams? Then what would be the implications about sequencing as I want to keep ordering? I could not find any docs around this specific case.

[@awssupport]

Hello! This doc has more info on your ask: go.aws/48vak9G. If you need additional assistance, check out these great resources: go.aws/get-help. ^LG

[@abduljabbarazam943]

Fantastic information dense video. I had to listen at x1 speed😁

[@drasticfred]

Can users create partial or range based indexes with Dynamodb?

[@recover89]

40:30 Dog ate my badge.

[@jcwrightson]

Wow... I just got schooled! 😎

[@TheMaddybaddy]

07:38 SQL - good for OLAP NoSQL - built for OLTP at scale Well, I never thought that I would say that to AWS Tech Lead, but I don't agree. NoSQL databases are ridiculously bad with transactions: DynamoDB/MongoDB started to support transaction processing only a few years ago and they totally weren't "built for OLTP at scale" at all. You wouldn't even want to compare RDBMS with NoSQL from the transaction point of view. I remember how I was shocked when one of my engineering colleagues talked about MongoDB and how good it is a few years ago but then told me that the default level of isolation there is "read uncommitted". It's just a complicated key/value storage with good scaling and additional features like TTL (that looks like a joke in DynamoDB as well).

[@brothermalcolm]

@26:10 sql vs nosql

[@hkaushik]

I am wondering how and when he catch a breath😮

[@nickvanhoof5212]

Good stuff. Did someone implement this by any chance? I'd love to play with it hands on to get it in my fingers.

[@AMS_at_home]

Found a lot of interesting stuff, thanks for material

[@amazonwebservices]

It's our pleasure! 😀 ☁️ 💻

[@jy.coffee5961]

This is fantastic!

[@blunderfoxbeta]

Can you tell how to fetch records without comparing partition key and without using scan. I know the answer is no.

[@andreadiotallevi5780]

Excellent and inspiring!

[@zdenektison3934]

The reality with DynamoDb is not so nice at it looks. It has many limitations which you will need to count with. The same with DynamoDb streams and Lambda. E.g. Lambda is executed every 250ms, it doesn’t sound for me like a real-time stream. GSI are eventually consistent, LSI limits the max data size you can have in partition and many, many more. Not even speaking about custom filters, almost impossible to implement it efficiently without expensive read all and be charged for all data you have in partition. So you can exceed a max parttion capacity (3K RCUs)by query that return empty result set.

[@baumannalexj]

kzbin.info/www/bejne/rpfbc5usfNWil5Isi=tXRFRjxr9xS7w-HN&t=1605 "moore's law was a friend to relational data" I never really connected this fully. I kept asking "why is this coming up now? RDB has been used for decades." Absolutely makes sense that our data collection is outpacing the hardware.

[@lambar0]

Good Stuff

[@AWSEventsChannel]

Thank you so much! 🙌

[@superchargingdynamodb]

Does this guy ever breath? He's seriously knowledgable about NoSQL and Dynamo.

[@hp1612]

Drink every time he says Moore’s law

[@justcreate1387]

No.

[@rickharold7884]

Love it

[@rubalsohi1988]

hve to play it @0.75 lol

[@EmanueleDiSaverio]

The guy talks super fast and very monotone ... it's too bad because the information would be so golden

[@chanyt2364]

I have him on playback speed 0.75, much better for my 'processor' 😂🧠👀

[@kumara1045]

NoSQL does not given consistency, it is eventually consistent. If I say in interview NoSQL is the best for OLTP I would be fail miserably

[@studiousguy8138]

There is context to everything. All reporting / business intelligence / OLAP generally run on RedShift/MySQL-type DBs. Most app development doesn't need jack-of-all-trades type non-horizontally scalable SQL DBs, highly specialized NoSQL DB's work. For Payments etc, you can use strongly consistent DB's like SQL or even DDB comes with strong consistency option, you will have to lose some throughput or latency goes high, but it gets the work done.

[@codevev]

Dang this guy talks in 2x speed

[@JamesB-ds9]

Almost nothing this man said made any sense to me. I'm sure he knows what he is talking about, but its completely inaccessible to someone who knows very little about how DynamoDB works.

[@JamesDavis-g2m]

Taylor Deborah Walker Carol Anderson Laura

[@mackosajt86]

to me as not a native speaker he talks to fast. like someone who needs to go to the toilet.

[@vitalytablet5899]

It's impossible to listen to him.

[@ianokay]

This guy talks like a salesman more than anything (speak so fast that it seems to only make sense, and don't give time for consideration). This felt like watching an infomercial

[@prateeksingh2909]

What is the purpose of this video ...to show you are a parrot 🦜 or convey knowledge