April 2016 Summaries

In the conclusion of a two-part series, the article examines the Seastar I/O Scheduler used by ScyllaDB to manage disk I/O, focusing on its ability to provide predictable latencies, fairness, and balance among various priority classes. The Seastar scheduler employs a shared-nothing architecture, dividing global resources like I/O depth among shards, and utilizes I/O Queues to facilitate local parallelism and efficient resource allocation. By classifying requests into priority classes with assigned shares, ScyllaDB ensures balanced resource distribution, even under heavy workloads. The scheduler's design allows for dynamic adjustment of shares, maintaining low latencies and efficient disk usage. Future versions may include dynamic share adjustments and support for multiple disk arrays with separate schedulers, enhancing performance and scalability as core numbers increase. The article highlights that relocating the disk I/O scheduler to userspace within a thread-per-core model provides more stable and faster results, showcasing a significant advancement in scalable database performance.

ScyllaDB 1.0.2, released by the ScyllaDB team on April 27, 2016, is a bug-fixing update to the stable 1.0 branch, ensuring backward compatibility and supporting rolling upgrades. This release addresses several issues, such as improving error handling for accessing non-existing user-defined types, accepting SSTables without summary files, enhancing read performance by optimizing the use of summary entries, and fixing a regression where collections with only deleted items appeared as empty rather than null. Additionally, it prevents the automatic start of the Scylla server during installation on Ubuntu 14.04. Contributions from developers include improvements in SSTable handling, user-defined type implementations, and collection query tests.

In datastores like ScyllaDB, a userspace Disk I/O Scheduler is crucial for managing competing disk I/O requests, ensuring that prioritization and fair allocation of resources are maintained, unlike traditional methods that rely on kernel-level scheduling. The Seastar framework, used by ScyllaDB, employs a thread-per-core design to enhance scalability and avoid iowait times, which can occur when requests exceed the processing capacity of lower layers in the stack, such as the filesystem or disk array. By managing I/O requests in userspace, ScyllaDB can provide rich processing capabilities, including request prioritization and cancellation, ultimately reducing latency and improving performance. Experiments demonstrate that using an I/O Scheduler can significantly lower latency at the 99th and 99.9th percentiles, even in noisy environments like Amazon EC2. The upcoming second part of the article will delve into the design of the I/O Scheduler and its role in ensuring fairness among multiple users.

Avi Kivity, the CTO of ScyllaDB, is set to present a talk at the Percona Live conference in Santa Clara, California, focusing on ScyllaDB, a distributed NoSQL database compatible with Cassandra, designed to handle over a million requests per second per node. His presentation will highlight how ScyllaDB utilizes systems programming techniques in a horizontally scalable NoSQL architecture to achieve significant performance enhancements, including a tenfold improvement over traditional Cassandra through its thread-per-core design. The conference, taking place from April 18-21 at the Santa Clara Convention Center, will feature Avi's talk on April 19th at 2:20 PM. Attendees interested in ScyllaDB's design, usage, or future developments are encouraged to participate and can find more information and register on the Percona Live website. To stay updated on future events and ScyllaDB news, individuals can subscribe to the newsletter, follow the blog's RSS feed, or follow @ScyllaDB on Twitter.

ScyllaDB 1.0.1 is a bug-fixing release from the ScyllaDB team, ensuring backward compatibility and supporting rolling upgrades from version 1.0.0. This update addresses various issues, including the improper formatting of error messages when reading malformed SSTables, compatibility problems reading SSTables from older Apache Cassandra versions, and mismatched digests of data queries that could increase latency, particularly in multi-data center setups. Enhancements include a more robust memory allocator, an auto-detecting network interface configuration script, and support for bonded interfaces, along with the removal of misleading log messages from the REST HTTP server. The release also corrects an incorrect URL provided by the scylla_io_setup utility and includes contributions from various developers to improve the handling of SSTables, mutation queries, and test cases for query and data handling.

ScyllaDB has announced the release of its production-ready version 1.0, highlighting its high availability, throughput, and low latency as key strengths, which have been demonstrated in recent benchmarks showing significant performance advantages over Apache Cassandra. This is achieved through a complete rewrite in C++14, using fully asynchronous reactive themes and kernel bypass techniques, allowing ScyllaDB to deliver up to 10 times the throughput and latency benefits without downsides. With a scale-out design similar to Cassandra, ScyllaDB has already been adopted by some early users and evaluated by Fortune 500 companies across various industries. The database's ease of migration, operation, and lower mean time between failures make it a compelling choice, offering better performance and reduced cost of ownership. ScyllaDB plans to enhance its compatibility with existing and newer Cassandra versions in future releases while continuing to expand its database core and feature set. The company encourages users to engage with their community through GitHub feature requests and the scylladb-users mailing list to help shape its development.