***hubspost stuff below

Inline deduplication enables optimal use of RAM

Infinio is a content-based cache, which means that we track commonly requested content within our cache rather than what other caches do: track commonly requested locations. So if any node requests a particular piece of content, it may already be in cache, even if it is that node’s first request. The net effect of this is a deduplication rate that can increase the effective cache size by several times the physical RAM. This tracking of content enables Infinio to deduplicate at a very granular level, which is the key to best utilizing small amounts of RAM from each host.  

In fact, because our cache is distributed and shared, the amount of duplicate data across an Infinio cluster can be substantial. Enterprise applications can yield a deduplication rate upwards of 5X. In specialized environments (such as VDI), these ratios can be even greater: 32GB of physical RAM can suddenly become over 320GB of effective cache size, enabling a significant percentage of the working set to reside in RAM.

Infinio’s content-based design also ensures that the cache is distributed equally across nodes in the cluster. We use a consistent hashing function that ensures the entire cluster shares the caching load equally, and that there are no hotspots or bottlenecks on the hosts.

Scale-out design grows with your applications

While localized storage performance enhancements can be a good solution for an isolated performance problem, most organizations find that storage latency is a problem that emerges in different places throughout the environment. Infinio’s distributed and shared cache design ensures that all the resources in a cluster can be put toward improving performance for the entire cluster. The RAM on each node is part of the shared cache, and all the nodes can access all the content in the cache.

As your cluster gets bigger, performance gets better. Each additional host contributes more RAM to the pool of cache, and each additional application provides more opportunity for deduplication. Nodes and datastores can be added non-disruptively, and our uniform distribution of work ensures that all the nodes contribute equally to the caching work.  When you add nodes or datastores, there are no workloads to migrate, no indexes to move, and no hotspots of bad performance.  

Hybrid VM kernel integration is optimized for performance

For deploying Infinio, we’ve developed a unique hybrid architecture which combines our virtual appliance with a kernel module for VMware’s Pluggable Storage Architecture. While a virtual appliance is a great way to manage our sophisticated cluster-wide architecture, intercepting I/O and responding to cache misses directly in the kernel enables us to provide the best possible latency across all operations. The trade-offs of performance and resource allocation are avoided in this model, giving you the “best of both worlds.”  This hybrid architecture deploys in your environment with no disruptions or downtime, providing optimal performance and capabilities without operational tradeoffs.


Joomla25 Appliance - Powered by TurnKey Linux