Infinio is built for the use of both RAM and flash devices or SSDs. When performance is most important to you, RAM is the best option, because it can achieve 100X the performance of flash. In fact, Infinio has always designed around a memory-first approach to caching; we’ve always had including industry-leading inline deduplication natively included in the product.
We also offer the option of using persistent media (like SSDs or flash devices) to act as another layer of cache within the host. When cold data is evicted from Infinio’s RAM cache, it is stored on persistent media. This can greatly improve your application performance by enabling so-called “warm” data that doesn’t reside in RAM to still benefit from fast server-side resources.
Infinio is a content-based cache which tracks commonly requested content, enabling highly efficient and granular deduplication. This means that if any VM on an Infinio host requests a piece of content, it may already be in cache, even if it is that VM’s first request.
Enterprise applications can yield a deduplication rate of 5X and higher. In specialized environments, such as VDI, these ratios can be even greater. Whether you’re using RAM, SSDs, flash, or a mix of these, the content-based design ensures the most efficient use of these resources.
A common concern when implementing a cache is the speed with which it can “re-warm” after downtime. If there is a flash device on the host, a significant amount of data that is requested often will exist in both RAM and flash. After a shutdown or power failure, when the host is powered back on, Infinio will have immediate access to the cache on flash, leading to a quick warming of the cache. This minimizes the impact of the reboot or power failure to performance.
Equally concerning is the issue of data integrity. However, the data stored in Infinio’s cache is always protected from failure or corruption. Infinio software doesn’t keep any unique content in cache and immediately writes all data to disk. So if the cache itself fails, or if a host fails, all the data is still safely on disk.
As the cluster gets bigger, performance gets better. Each additional host contributes more resources to cache, and each additional application provides more opportunity for deduplication. I/O scales out with application growth, increasing performance linearly as more nodes are added. Nodes and datastores can be added non-disruptively. There is no requirement for nodes within an Infinio cluster to have symmetry in their resources, enabling different hosts to leverage whatever server-side resources are available for providing performance.