ScyllDB's configuration parameters automatic changes.

[ashraf.youssef.mahgoub@gmail.com]

Hi, 

While bench-marking ScyllaDB (V 1.5), we noticed that it performs significantly better than Cassandra (V3.9) in terms of throughput and 99th percentile latency.  However, C* seems to reach saturation in performance faster than ScyllaDB. I'm aware of ScyllaDB's property that it is self-managing and it performs measurements under the hood to adopt its performance to the applied workload (as stated in https://www.compose.com/articles/talking-scylladb-with-cto-avi-kivity/).

1- Is there a way to view the configurations parameters while they change over-time ?? 

2- How does this feature make use of the user-specified parameter values int the Scyall.yaml file ? 

3- Can the user change the configurations in the Yaml file while the node is running, or a shutdown/restart is needed to apply the new configurations?

Best

Ashraf 

[Glauber Costa]

On Thu, Feb 9, 2017 at 2:46 PM, <ashraf.yous...@gmail.com> wrote:
> Hi,
>
> While bench-marking ScyllaDB (V 1.5), we noticed that it performs
> significantly better than Cassandra (V3.9) in terms of throughput and 99th
> percentile latency. However, C* seems to reach saturation in performance
> faster than ScyllaDB. I'm aware of ScyllaDB's property that it is
> self-managing and it performs measurements under the hood to adopt its
> performance to the applied workload (as stated in
> https://www.compose.com/articles/talking-scylladb-with-cto-avi-kivity/).
>
> 1- Is there a way to view the configurations parameters while they change
> over-time ??

Best way is to deploy our monitoring stack alongside scylla
(https://github.com/scylladb/scylla-grafana-monitoring). It's not
exclusive for changes of this kind, but it will show various internal
metrics that can give insight into the system. Over time we're adding
more and more information there.

> 2- How does this feature make use of the user-specified parameter values int
> the Scyall.yaml file ?

It depends on the parameter. The goal is to have less and less
parameters, but some of them can be used to provide
operational bounds until they are addressed.

One example is the memtable_total_space_in_mb parameter, that was used
to inform Scylla how much memory to use for memtables, and Scylla
would respect it. Starting in v1.5 that is done automatically so that
parameter is ignored.

> 3- Can the user change the configurations in the Yaml file while the node is
> running, or a shutdown/restart is needed to apply the new configurations?

Unfortunately it's not possible to change those while the node is
running at the moment. This is a nice idea for improvement.


>
> Best
> Ashraf
>
> --
> You received this message because you are subscribed to the Google Groups
> "ScyllaDB users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to scylladb-user...@googlegroups.com.
> To post to this group, send email to scyllad...@googlegroups.com.
> Visit this group at https://groups.google.com/group/scylladb-users.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/scylladb-users/57d73bc5-6737-4a49-83f8-69fffffaa5e9%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[ashraf mahgoub]

OK great. When you mentioned the "memtable_total_space_in_mb" parameter  being overwritten by ScyllaDB, is there any other parameters being overwritten as well? Also I was comparing the "Leveled_Compaction" strategy to the default "Size_Tiered" strategy and for Cassandra I'm getting better performance. However, for ScyllaDB it doesn't seem to make much of a difference. Also with "Size_Tiered" I see SSTables levels when I use "Nodetool cfstats" command, which means some "Leveled" properties are adopted to the default "Size_Tiered" compaction in ScyllaDB ? 

Best

Ashraf

On Thu, Feb 9, 2017 at 2:57 PM, Glauber Costa <gla...@scylladb.com> wrote:

> email to scylladb-users+unsubscribe@googlegroups.com.
> To post to this group, send email to scylladb-users@googlegroups.com.

> Visit this group at https://groups.google.com/group/scylladb-users.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/scylladb-users/57d73bc5-6737-4a49-83f8-69fffffaa5e9%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "ScyllaDB users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/scylladb-users/WLR0FsFBf64/unsubscribe.
To unsubscribe from this group and all its topics, send an email to scylladb-users+unsubscribe@googlegroups.com.
To post to this group, send email to scylladb-users@googlegroups.com.

Visit this group at https://groups.google.com/group/scylladb-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/scylladb-users/CAD-J%3DzasPSOmaKh9P7CzKOC%2BEzm-r96emzbY%2B55t-yjy5GWr2Q%40mail.gmail.com.

[Glauber Costa]

On Sun, Feb 12, 2017 at 1:37 PM, ashraf mahgoub
<ashraf.yous...@gmail.com> wrote:
> OK great. When you mentioned the "memtable_total_space_in_mb" parameter
> being overwritten by ScyllaDB, is there any other parameters being
> overwritten as well?

Yes. In the memtable arena alone, there are more that we are ignoring.
But we'll get to more of them in the future.

> Also I was comparing the "Leveled_Compaction" strategy
> to the default "Size_Tiered" strategy and for Cassandra I'm getting better
> performance. However, for ScyllaDB it doesn't seem to make much of a
> difference.

No compaction strategy is better or worse per se. They are extremely
workload dependent. If you tell us more about your workload we can try
to help you find one that fits.

> Also with "Size_Tiered" I see SSTables levels when I use
> "Nodetool cfstats" command, which means some "Leveled" properties are
> adopted to the default "Size_Tiered" compaction in ScyllaDB ?
>
> Best
> Ashraf
>
> On Thu, Feb 9, 2017 at 2:57 PM, Glauber Costa <gla...@scylladb.com> wrote:
>>

>> > email to scylladb-user...@googlegroups.com.
>> > To post to this group, send email to scyllad...@googlegroups.com.

>> > Visit this group at https://groups.google.com/group/scylladb-users.
>> > To view this discussion on the web visit
>> >
>> > https://groups.google.com/d/msgid/scylladb-users/57d73bc5-6737-4a49-83f8-69fffffaa5e9%40googlegroups.com.
>> > For more options, visit https://groups.google.com/d/optout.
>>
>> --
>> You received this message because you are subscribed to a topic in the
>> Google Groups "ScyllaDB users" group.
>> To unsubscribe from this topic, visit
>> https://groups.google.com/d/topic/scylladb-users/WLR0FsFBf64/unsubscribe.
>> To unsubscribe from this group and all its topics, send an email to

>> scylladb-user...@googlegroups.com.
>> To post to this group, send email to scyllad...@googlegroups.com.

>> Visit this group at https://groups.google.com/group/scylladb-users.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/scylladb-users/CAD-J%3DzasPSOmaKh9P7CzKOC%2BEzm-r96emzbY%2B55t-yjy5GWr2Q%40mail.gmail.com.
>> For more options, visit https://groups.google.com/d/optout.
>
>

> --
> You received this message because you are subscribed to the Google Groups
> "ScyllaDB users" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to scylladb-user...@googlegroups.com.
> To post to this group, send email to scyllad...@googlegroups.com.

> Visit this group at https://groups.google.com/group/scylladb-users.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/scylladb-users/CANcbxpgDoZ5KpUAAw-uwniGWFqyT-T11T4f_ZLRrykBW%3DjLD6Q%40mail.gmail.com.