Apache Cassandra Use Case Interviews from Cassandra Summit Europe 2014, 27 Dec 2014

Previous Meeting

⏯

youtube image

►

From YouTube: Apache Cassandra at ING with Gary Stewart, Advisory IT Specialist

Description

Speakers: Patrick McFadin, Chief Apache Cassandra Evangelist at DataStax & Gary Stewart, Advisory IT Specialist at ING

A

A

All right we're at cassandra summit, 2014 europe and welcome gary hi. So why don't you tell us about yourself where you work.

B

My name is gary stuart, I'm working at ing in the netherlands and back-end developer and database experience. I tend to have a love-hate relationship with cassandra, but it's starting to become more of a love relationship in the beginning with hate. So, oh.

A

Look at you okay, so we got a little hate. So what tell me about your experience with cassandra like early on what? What did you not like about cassandra? Not.

B

Being able to join tables, I mean it's really sort of uh you know you start with a database and you think you're used to it. You know database and cassandra has cql3, so it looks and smells like sql right, but once you start getting past the underlying sql limitations and you're getting availability and scalability in return, but you need to give up some other stuff and when you give up the other stuff- and you start understanding that, then you know why you're doing it.

B

So you have to spend a bit more time developing and a bit less time and operations. That's uh what.

A

What kind of databases did you use before using sql server? Okay, so relational databases, relation um did you ever use a mainframe? Yes, because we talked to chris earlier and he said he used the mainframe. Okay, so see those.

B

Are still out there right, that's one of the uh potential use cases for the tender we need to uh have scalability at the fingertips, so that and also availability, because we are experiencing outages and with a single instance. uh Customers are affected right and we take that very seriously. So we're really now exploring many different ways of solving this problem and cassandra being masterless is designed to tolerate failure, which is uh why we think it's a good use case right. It's a good product for our use.

A

Case so your talk today here at the summit you're going to be talking about consistency, challenges, so maybe talk about one of the ways you you solve some of those issues for what your use cases are. So what what kind of what? What do you use? What techniques do you use.

B

Well, there's a couple of challenges with distributed systems, the first one that we mentioned is about time differences, because when you have different machines, you run the risk of having other sync times, and if you have to make decisions based on time, then, if it's out of sync, then you run into some weird scenarios.

B

So we've managed to solve that by using a finite state machine and essentially ignoring time. There's. Another challenge is because it's distributed, you don't have transactions.

B

So when you throw a request out and you get a timeout- you don't know if it's out there or if it's not so essentially the client and the nodes can be out of sync. So we've solved that again with the finite state machine item, potential, modeling and knowing our critical path.

B

So there's a lot of you know little attention to details that you need to pay attention to, and we talk about that in the presentation. Well,.

A

I I tell people to look at your presentation, often because it is very it's a different way of thinking. Yeah and it's nice. You don't have to give up those things like uh availability, yeah, yeah and so there's other ways to fix it. So let's switch gears a little bit. um You do a lot of talking in community. I remember we.

B

Did a meetup together, yeah.

A

So tell me about what what what your involvement in the community and what do you think about the cassandra community.

B

Chris and I uh in the netherlands, we took over the cassandra community. I think it was done by ebuddy before then um and ing wanted to also start promoting uh engineering culture and basically making a sort of contribution to the community right. So me and chris took this opportunity to host a couple meetups right and it's been fantastic, we're getting international speakers. The turnout is great. The community is now, I think, 346 members when we started it was 120.

B

So now, since the last year, it's uh growing uh more than two and a half times yeah, there's.

A

A lot of casinos in there isn't it.

B

And that's one of the maybe the downside of cassandra is there's not a lot of skills on the market, so this is the reason for the community if we're really trying to make an awareness and get other people interested, so the more people using the product, uh the better it is for us and the better it's for calendar, and then we can all proceed in the right direction.

A

So um at ing does, is there a lot of people or is there centralized? Are the people just? Is there a central group that does cassandra work or is it pretty distributed of people who use casino.

B

There are many teams uh we're into devops, so we have a, I think, over 150 scrum teams, so that can be quite challenging.

B

What we're trying to do now is gather all the knowledge in one place, so we can get the best practices and try to keep in a community way so that we also within the company, have a nice way of working together and then take it forward so that we share the knowledges and.

A

There's a quite a big project right now I remember seeing one of your cios that talked about uh some integrations and things like that. So a lot of devops work being done there and docker as well. Right yeah, that's really interesting! So what was the project? I can't remember the name of it.

B

The project is a darwin project and um is.

A

It an evolution, sorry, is it an evolution? Well, it's.

B

The first step to the new bank uh with the lint bank: there we want to essentially break the front facing services from the back end so that we can reduce the load and ensure availability. But the challenge we have is the consistency and that's what we talked about in our right. Exactly.

A

And it goes to prove you don't have to be a startup to use cassandra when when was ing a startup. What year was that.

B

Was a startup yeah about 100 years ago yeah? I think it was longer.

A

Than that was it, I think I remember we were talking once in the uh part of the u.s purchase was financed by ing, and that was in the 1700s.

B

It's been around for a long.

A

Time, that's a really long time. It's amazing, um you don't have to be a startup right been around for hundreds of years great well, thank you very much uh for coming down here and talking to me and thank you thank you for coming to the summit and doing a talk, you're welcome. Thank you.