Last week we announced a new round of funding for ScyllaDB. In this blog post I will explain where we are now and what to expect next. I invite you to read the funding announcement and news coverage to discover the list of VCs and partners who are putting their trust into Scylla and hear from our customer and users themselves. This round of funding will go towards expanding our database offering, growing the Scylla open source community, and adding features and functionality in our upcoming Scylla Enterprise release.
We at ScyllaDB are becoming increasingly bullish about our database, Scylla. Everything works exactly as we designed it. Scylla easily scales to dozens of nodes across multiple data centers while providing the best throughput and latency. That is why vendors like AppNexus run a Scylla cluster of 47 nodes across five data centers. Learn more by visiting our Users page. If you already use Scylla and are part of our growing community, join companies like Musical.y, Ola Cabs and many others and share your story about the momentum Scylla is generating.
One of the exciting outcomes of having “the power of Cassandra with the speed of Redis” is that users are gravitating to Scylla from other platforms, not just from Cassandra or SQL. The use cases published on our Users page include migrations from Redis, Hbase and Aerospike. This isn’t surprising because the sheer number of niche databases in the field is overwhelming and many want to consolidate their multiple databases onto a single platform. Scylla is designed for this job. Derek Schoettle, general manager at IBM Watson Data Platform, says “ScyllaDB’s NoSQL database offers a powerful combination of low latency and high availability, making it an attractive option for customers of our Watson Data Platform offering.”
Make sure you do not miss the recent i2.8xlarge and c3.2xlarge AWS benchmarks, along with a third-party benchmark done by Samsung. Scylla provides the performance our customers need without any compromise on cost, availability, or complexity. The benchmarks show that Scylla produces gains on almost any given hardware. Unlike Cassandra and other JVM-based projects, we recommend using larger machines and a smaller cluster to get the maximum value out of Scylla. Doing so increases utilization and decreases the human labor and complexity tied to the management of a larger, 100-node cluster. In the coming weeks, we will publish an official AWS i3 image which has the best bang for the buck.
We deliberately and intentionally chose to retain the best traits of Cassandra, including its high availability, polyglot of languages, and friendly protocol, because they raised the bar in the database industry and revolutionized how developers build applications. The time has come to take this great work and raise it to an even higher level.
- Year 1: Creation to beta
During mid-2014 towards Cassandra Summit 2015, we developed our initial beta version. Our focus was to deliver the same API at 10x the speed and allow 1 million operations per server.
- Year 2: Beta to general availability (GA)
From September 2015 until our very first Scylla Summit in 2016 we hardened our technology and transformed it to a product. We added functionality such as multi-datacenters, security, legacy APIs (Thrift, which won’t get deprecated like its Cassandra’s counterpart), basic manageability options, and started to support the first production installations.
Present Scylla projects
- Year 3: Feature parity with Cassandra plus Workload Conditioning
Performance is important, but friendliness and ease-of-use are even more important. #WorkloadConditioning allows Scylla to provide consistent results out of the box. We tune the system automatically and dynamically on your behalf. Forget about tweaking a gazillion knobs. We adjust Scylla for the selected hardware and constantly tune it for the varying workloads.
Scylla is designed to sustain multiple failures. One of the coolest upcoming features is a hot-cache load balancing, where Scylla will read from nodes with hot cache while newly added nodes get automatically warmed up. In Scylla 1.7 (March 2017), we release our counter support. In 1.8 (April 2017) we will have preliminary materialized views (MV) and a new row cache for extremely large partitions.
In 1.9 we will present our secondary indexes which are based on the MV implementation and thus scalable. Later, we’ll finalize the gap with lightweight transactions and Cassandra 3.x storage format.
Along with the open source 1.7 release, we are pleased to announce the Scylla Enterprise release. Designed to support our enterprise customers, Scylla Enterprise has a long lifetime length, adds functionality to migration tooling, enhances security, offers 24/7 support, escalates bug fixes, and more.
Scylla in the future
- Years 4-9 “It is difficult to make predictions, especially about the future”.
We will be laser focused in the coming year on the core database, dividing the future into three categories:
- Core features. There are plenty of improvements to deepen the Scylla advantage. New storage format, additional columnar format, hook Seastar userspace TCP stack to Scylla, K/V optimizations, true multi-tenancy to guarantee SLA among tenants, DBaaS, management console, and more.
- Vertical enhancements. Spark and Scylla are a combination made in heaven. We will enhance the interface with MapReduce aggregations on the Scylla side and prevent expensive full-table-scans. Spark is one example—and Presto, KairosDB, and JanusGraph are additional viable examples. Vertical enhancement isn’t just upwards but also downwards towards the hardware. A userspace filesystem and 3D-xpoint are on the horizon too.
- Horizontal growth. When you have an engine such as Seastar and clustering technology derived from Cassandra, why settle for CQL? We’ve shown that Seastar can create blazing-fast httpd. Alibaba uses Seastar to rewrite Redis in a better way (called Parallel Redis – Pedis). There are more big data projects and we hope to get there one day. This is where you all can chime in and extend Seastar for your use-case. If you search the web well, you’ll find that one of the Seastar adopters was acquired by a giant.
Join our mission
We are humbled to take the path paved by the Apache Cassandra community and we invite the community to join us in our mission to be the number one NoSQL database in performance and availability. Now it’s time for me to get back to work—dive into our Prometheus dashboard which needs more love, play with our Docker image, and suggest ideas on how our log messages need to be simplified. We are very fortunate to have attracted international top talent (we’re hiring!); they are not only great engineers but they also aren’t shy about jumping on a customer issue and traversing all paths until it is resolved. The support crew and enterprise solution engineers await your workload. Go ahead and send questions to the mailing list, chat on Slack and report issues on GitHub.