Apache Cassandra Apache Cassandra World Party | April 2021, 4 May 2021

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: Cassandra in Adobe Audience Manager | Apache Cassandra World Party 2021

Description

Presentation from the first Cassandra World Party on April 28, 2021. Learn more about Apache Cassandra at https://cassandra.apache.org/.

Serban Teodorescu of Adobe presents “Cassandra in Adobe Audience Manager” on various topics covering what Adobe has done in more or less different ways than most.

About the Presenter
Serban is an SRE at Adobe, where he’s part of a small team that manages 30+ Cassandra clusters for Adobe Audience Manager. Previously, he was a Python programmer, and he’s still trying to find out how a developer who preferred SQL databases ended up as an SRE for a Cassandra team. Apart from Cassandra and Python, he’s interested in automating his tasks with terraform, AWX, and StackStorm.

Apache Cassandra is an open source NoSQL distributed database trusted by thousands of companies for scalability and high availability without compromising performance. Linear scalability and proven fault-tolerance on commodity hardware or cloud infrastructure make it the perfect platform for mission-critical data. Get started: https://cassandra.apache.org/quickstart/

The Apache Cassandra World Party 4.0 was a virtual, one-day celebration on Wednesday, April 28, 2021 in preparation for the launch of 4.0, the most stable database release ever shipped.

Join the Cassandra community: https://cassandra.apache.org/community/

#ApacheCassandra #database #NoSQL

A

Okay, so our uh next speaker is your band: he's a site, reliability engineer at the dope, uh where he's part of a small team that manages 30, plus cassandra clusters for adobe audience manager.

A

Previously he was a python programmer and he is still trying to find out how a developer, who preferred sql databases, ended up as necessary for cassandra team apart from cassandra and python. He is also interested in automating his tasks with terraform, awx and snakestorm.

A

In this talk, he will go through various topics covering what the dog has done in more or less different ways than most so welcome sheban, and I think we are ready to go patrick. Thank.

B

You it's uh it's not about this relationship, it's not about the catholic zero! It's about how we are using kassandra, adobe, executive, clean audience manager.

B

Can switch to the next slide, so um I think everyone heard about uh adobe, but most would only know about uh photoshop, pdf and other related products and perhaps flash which we don't like to mention.

B

We are doing much more, it's not the time or place to essentially talk about this. Just saying that audience manager is counting clicks and cookies. You can find lots.

B

um We've been using cassandra since 2011 and uh 2.1.

B

The largest cassandra deployment in adobe we have 900 significance, is in production, storing 260 terabytes of data 34 fasteners in aws.

B

um So what are we doing different? We were, we do have a large deployment, but other have much larger, so I'll just go through differences. First, the replication factor. We are using repetition, factor 2 with consistencies level 1, and this is possible because we are using the lambda architectural, protecting process, pushing the data to cassandra which should ultimately fix any missed right.

B

This is done through cassandra streaming, which is much more efficient than regular inserts, and there are major improvements on steering cassandra program, looking forward to use that and we are using uh v nodes with the token allocation algorithm and.

B

This creates the verification factor and the usage of genomes means that there is a system probably into those from different tracks or different available without the aws. So speak, it's a it's a high resolution problem to have them sharing data. So if we lose any two nodes, we're in danger of having availability problems, so we must be very quick of fixing this.

B

We have. We have a fully automated process, help us to replace uh road fully automatically in place to create the instance you can use. It was here that both in the previous screen, um another automation example, is a very resistance. uh Os major version upgrade, which is the electric technique presented at apache 2019 by.

B

It's based on overriding the root device without using or moving the data, kept, not needed from stores. Now the difference where we are not using the oracle for open source java, we are using azure. The java runtime engine supports garbage collection.

B

It helps a lot as you can see in the draft with the united nations, france and uh we switched four or five years ago, and it has been just uh great. You can see more details in their link there, or I know this was an old situation where we had to perform an intentional state break something usually to be avoided. In a knowledge field, databases we had to add a new deployment in a new aws region by sitting on already existing one.

B

We extended the existing catandra clusters by adding the data center to each deployed in the new region after they, after they were synchronized with the data center sourcing to independent clusters, everything in production with no visible impacts for the customers.

B

uh We had some contributions between this time rate and would fix some very small. I think there's some contribution to the impact cluster uh magic exporter from commission and some presentations across our data structure and some local incidents.

B