In most cases, caches are a front-end patch to legacy SQL and NoSQL databases, precipitated by sluggish throughput and unacceptable latencies. External caches such as Memcached or Redis may improve performance in the short term but add cost and complexity, poor horizontal scaling, lack of persistence, and an inability to store everything in DRAM, which results in diminishing returns as you scale.
ScyllaDB’s achieves single-digit millisecond P99 latency, ensuring that your applications are responsive and perform even under heavy loads.
ScyllaDB is designed to handle large-scale deployments with ease, providing linear performance compared to Memcached or Redis even as the workload increases.
ScyllaDB shared-nothing, integrated memory cache, and asynchronous I/O optimize cloud utilization far more than single-threaded Memcached or multi-threaded Redis.
ScyllaDB persists all data from the database and its integrated cache, reducing costs. Users also avoid the complexity of managing an external cache, including failover periods, intervals between snapshots, and data loss avoidance.
In ScyllaDB, consistent hashing is done at the database level, enabling automatic load balancing and efficient use of hardware resources across availability zones and regions, thus eliminating single points of failure.
If your data-intensive apps require hundreds of thousands to millions of operations/second and msec single-digit latency, offload the datasets required by these apps from your legacy database and skip the cost and complexity of adding an in-memory cache.
Disney+ Hotstar, India’s most popular streaming service, experienced high growth rates over six years. They constantly increased their Redis cluster size, struggling to handle their write-heavy workloads, particularly during high-traffic periods. With ScyllaDB, they could handle millions of concurrent user watch lists and multiple devices.
Eniro is the leading search company in the Nordics. High latencies, a lack of data persistence, and timeouts prevented them from meeting their SLAs with Redis. They migrated from Kafka and Redis to ScyllaDB, eliminating their timeouts and meeting their SLAs with P99 latency in single-digit msec reads and sub-msec writes.
This blog by Felipe Cardeneti Mendes takes an in-depth look at database and cache internals and the tradeoffs in each, using a benchmarking exercise between ScyllaDB and Memcached.
Join this webinar for a technical discussion of different approaches to caching (pre-caching vs. caching, side cache vs. transparent cache), reasons why external caching is a bad choice, why Linux’s default caching doesn’t work well for databases, and the advantages and architecture of ScyllaDB’s specialized row-based cache.
Apache® and Apache Cassandra® are either registered trademarks or trademarks of the Apache Software Foundation in the United States and/or other countries. Amazon DynamoDB® and Dynamo Accelerator® are trademarks of Amazon.com, Inc. No endorsements by The Apache Software Foundation or Amazon.com, Inc. are implied by the use of these marks.
