Why use Redise Flash technology in Redise Pack
For Redis use cases such as real-time analytics, time-series data analysis, search, machine learning and more, dataset sizes can become voluminous and the cost of memory can become prohibitive. Using Flash memory as an extension to RAM can deliver cost-efficiency, while Redis Enterprise Flash continues to deliver memory-like latencies and high throughput, even at large volumes.
The unique Redise Flash technology in Redise Pack slashes operational costs by over 80%, by using SSDs as an extension of memory. With its breakthrough approach of tiering access to data, with keys and hot values in RAM, while cold values are stored on Flash, this technology delivers the blazing fast sub-millisecond latencies and high throughput of Redis, with a lower cost infrastructure.
Redise Flash uses Flash as a RAM extension, rather than as persistent storage. The Redis internal dictionary, keys objects and ‘hot’ values are kept in RAM, while ‘cold’ values (which typically account for the larger part of the dataset) are kept in Flash. Multi- threaded and asynchronous Redis is used when accessing objects on Flash for optimal Flash performance. The product is based completely on the Redis core architecture and is compatible with all Redis clients, data types and commands.
Optimize costs of large transactional, operational and analytic processing, with Redise Flash’s ability to tune the RAM/Flash ratio for a configurable performance/cost profile.
Redise Flash performance varies by hardware and storage choices, and by the ratio of RAM : Flash. Typical throughput with sub-millisecond latencies:
A single SATA-based SSD instance in the cloud
A single NVMe-based SSD instance in the cloud
A single x86 bare metal server using Samsung or Intel NVMe Flash
> 1M ops/sec
A single Power8 bare metal server using IBM CAPI
> 1M ops/sec
Learn more about the benchmarks.
When to use Redise Flash
Redise Flash is suitable for use cases in which the dataset size is above 100GB, the values are larger than the keys and “hot” objects (key/value pairs) account for 10-40% of all objects.
Redise Flash offers tremendous savings over RAM for large datasets. The range of savings you can expect will vary depending on data size and throughput.
The table below describes the maximum throughput at which you can cut costs by 50% for typical dataset size scenarios. Figures are calculated for AWS, but are indicative for most other clouds/ data centers. You can also calculate cost savings specific to your scenario using the calculator below.