Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Scylla release: version 1.0

scylladb.com

Post reply on HN
Print view View original

11–16 of 16 posts

Re: Scylla release: version 1.0

#11

Post by glommer »

continuations wrote: post #5

What's the status of secondary index? It was described as "half-ready" here[1]. Is that still the case? Also does Scylla support some sort of data locality as far as secondary index is concerned? For example if I want to store comments, I'd want to store all comments that belong to the same forum thread in a single node. All these comments have the same thread_id. Then I can have a secondary index on thread_id. When…

We are starting developing our solution for secondary index right now. We haven't yet made a decision regarding how we'll implement it.

Re: Scylla release: version 1.0

#12

Post by glommer »

lobster_johnson wrote: post #8
continuations wrote: post #6

How does Scylla compare to Aerospike? They seem pretty similar.

I don't know Scylla, but Aerospike is intentionally designed for low-latency networking [1] (i.e., typically colocated within a single rack), which can be a challenge in the cloud. [1] https://aphyr.com/posts/324-jepsen-aerospike

There is nothing Cloud-specific in Scylla, though. Scylla will also target low latencies, and we have been very successful at that. But we also commit to supporting the Cassandra data model, so we won't really chase extra latencies if it means breaking that.

Re: Scylla release: version 1.0

#13

Post by continuations »

glommer wrote: post #11
continuations wrote: post #5

What's the status of secondary index? It was described as "half-ready" here[1]. Is that still the case? Also does Scylla support some sort of data locality as far as secondary index is concerned? For example if I want to store comments, I'd want to store all comments that belong to the same forum thread in a single node. All these comments have the same thread_id. Then I can have a secondary index on thread_id. When…

We are starting developing our solution for secondary index right now. We haven't yet made a decision regarding how we'll implement it.

So a feature that was described as half-ready 7 months ago is actually still undecided design-wise 7 months later? That's surprising. When do you think secondary index will be available?

Re: Scylla release: version 1.0

#14

Post by glommer »

continuations wrote: post #13
glommer wrote: post #11

Earlier quoted context omitted.

We are starting developing our solution for secondary index right now. We haven't yet made a decision regarding how we'll implement it.

So a feature that was described as half-ready 7 months ago is actually still undecided design-wise 7 months later? That's surprising. When do you think secondary index will be available?

This feature was never described as half-ready. The ticket said we had half ready code, but it's part of the design process to choose whether or not we'll use that code. I can see, however, how this may give a reader the wrong impression, and I apologize for that.

As can be seen in the follow ups for that ticket, we haven't yet decided if we'll support standard secondary index, SASI, some form of materialized views, or all of them. Until we do, it's hard to commit to a timeline. Basic secondary index (for which we have code half ready) is pretty simple, and if we do implement it, it should land in our main version in a couple of weeks. The others are a bit more complicated.

Re: Scylla release: version 1.0

#15

Post by IvanVergiliev »

henrikschroder wrote: post #7
glommer wrote: post #3

Earlier quoted context omitted.

We don't support mixed clusters, though, and are unlikely to do so. Most people testing Scylla at this point are doing A/B testing, using some proxy to replicate the writes to both clusters and then comparing the two.

Okay, I understand that the burden of correctly supporting the internal gossip protocol is very high. It's a shame though, because it makes it harder for medium-sized installations to convert. If you're running a small cluster, then getting a few more machines to test it out isn't that much of a burden. If you're so big that you have multiple clusters, you can probably roll through and upgrade cluster by cluster or k…

On the other hand, if the performance benefits prove to be true, you might be able to test Scylla with much less machines than you have for Cassandra.

Re: Scylla release: version 1.0

#16

Post by tzach »

continuations wrote: post #6

How does Scylla compare to Aerospike? They seem pretty similar.

They are similar as both built in C/C++ with performance in mind. They are very different as Aerospike is a Key Value store, while Scylla, like Cassandra is Columnar (Key, Key, Value). This gives Scylla a much richer semantic.

There are more differences, like Aerospike holding all keys in memory, while Scylla does not have such limitation. I'm sure there are more differences in tunable consistency, HA and multi DC, but I'm not an Aerospike expert.

Post reply on HN