Scylla Blog

AWS announced their new generation of Graviton2 System on a Chip (SoC), based on the Arm Neoverse N1 core. AWS claims they are much faster than their predecessors, a claim that we put to the test in this article.

For you to get the most out of your big data applications, let’s explore the effects of concurrency in distributed databases and provide you with tools to correctly configure your infrastructure for maximum performance, including client-side parallelism and timeout settings.

In the first part of this blog we’ve learned a bit about compression theory and how some of the compression algorithms work. In this part we focus on practice, testing how the different algorithms supported in Scylla perform in terms of compression ratios and speeds.

In this two-part blog we’ll focus on the problem of storing as much information as we can in the least amount of space as possible. This first part will deal with the basics of compression theory and implementations in Scylla.

In this article we will explore one IoT/time-series classical scenario in which knowledge of how the cache operates can mean the difference between a fully cached workload that will be fast, and a fully storage-bound workload that will of course perform much worse.

Indexing is a useful tool that provides more types of queries on your tables. In principle, columns we wish to be queryable should be declared when the table is created, as part of a table’s primary key. Secondary Indexing is a neat way of making other columns queryable, but it comes with a cost of additional storage space and processing power to maintain the secondary index data coherent with the primary index information.

In practical terms, how do you begin to plan your ultra-efficient Scylla topology? The shift can be intimidating for those coming from a background in Cassandra, or for those coming from a primary-replica architecture, or from a plain old monolithic database implementation. In this post we’ll layout a checklist for designing a Scylla cluster that incorporates both horizontal and vertical scaling.

The I3en family is a class of instances clearly targeted at storage-intensive workloads, but CPUs are still needed to process that data. The largest of the I3en, i3en.24xlarge, ships with 48 cores of the Xeon Platinum 8175M, clocking at 2.50GHz.

In this post we introduce the new Scylla workload prioritization mechanism, explaining the vision behind developing this feature and how it is implemented, and most importantly, we show you test results of how it performs in a real-world setting.

In this article, we will compare Scylla Cloud and Google Cloud Bigtable. We show that Scylla Cloud is one-fifth the cost of Cloud Bigtable under optimal conditions (perfect uniform distribution) and that when applied to Zipfian distribution, the difference grows to 25x.