CacheIQ provides NAS acceleration with transparent caching

Modular design allows for high availability and scalability with superior price/performance

By James E. Bagley
Senior analyst
Deni Connor
Founder and senior analyst
Storage Strategies NOW

October 2011

File storage bottlenecks occur when file servers cannot meet the demands of clients. The solution to this problem is generally an expensive upgrade or expansion to the filer network. As NAS system sprawl increases, storage administrators find themselves engaged in manual file placement work and dealing with an increasing number of single points-of-failure. The Cache IQ approach is to decouple performance from capacity by adding a transparent cache for important data. This allows low-cost per TB, high-capacity Serial ATA drives to be deployed within filers rather than expensive, short-stroked rotating media or even solid state drives.

Modular architecture

Caching of file data between clients and file servers provides acceleration of information access by reducing latency and network traffic to and from the file servers. Cache IQ’s RapidCache has taken a unique approach which divides the operational functionality  into two components: Flow Directors and Data Servers. Flow Directors monitor traffic between clients and file servers, route the cacheable storage traffic  into the Data Servers, and provide direct access between clients and filers in the event of RapidCache maintenance or failure. Data Servers inspect and cache data based on policy and provide the rich set of analytics through RapidView. Systems can begin with a single Flow Director and one or more Data Servers. High availability can be achieved by adding an additional flow director to an array of Data Servers.

Data Flow

All traffic flows through the Flow Director. Based on policy set through RapidView, Network File System (NFS) traffic is sent to the Data Servers for processing, while other traffic is passed through untouched. While numerous Data Servers can be added to a cluster (up to eight), only one Flow Director is necessary, although a second can be added for high availability requirements, as shown in the above drawing. Traffic is load-balanced between data servers. In the unlikely event of a cluster failure, the data is passed directly back to the file servers. Data Servers provide a rich set of statistics on cache effectiveness, client and filer performance, and file activity which is collected and reported with RapidView. Data and metadata are cached according to usage or through simple user-defined policy.

Advantages of transparency

RapidCache is transparent to clients and file servers. No configuration changes are required on the clients, filer servers, or applications to take advantage of the cache. Unlike approaches that require the client to access the cache as a mount point, this transparency minimizes latency in the case of a cache miss, and requires no action in the case of a RapidCache cluster outage. When the cluster comes back on-line, transparent housekeeping determines the validity of cached data by checking its signature against the filer. Validation also occurs after a certain amount of time has passed between cached data accesses. This ensures that any changes made to cached data outside of RapidCache are picked up and reflected in the cache. The validation of the cache entry is done quickly and transparently by the Data Servers. Another advantage of transparency is that RapidView can provide unequaled analytics on both the storage and client sides of the equation. Detection of potential problems and clear before-and-after performance reporting is standard, along with troubleshooting capabilities.

Configuration and capacities

Flow Directors support both 1Gbps Ethernet and 10Gbps Ethernet. Data Servers contain 144GB of DRAM and up to 3.2TB of SSD. Each Data Server delivers over 10Gbps throughput and scales linearly to a maximum of eight data server nodes per cluster. Nodes can be added to or removed from the cluster without service interruption.  A single Flow Director and Data Server without SSD starting list price is $85,000.

Our Take

CacheIQ has added network flexibility to NFS NAS acceleration, creating a unique architecture with both Flow Director and Data Server nodes. In addition, systems can be configured economically and can grow without downtime. The addition of unequalled analytics and policy flexibility from RapidView make the RapidCache solution easy to install and tune. The transparent nature of RapidCache makes it viable for both local and cloud storage applications.