Sequential file systems have the drawback that data access quickly becomes a bottleneck if clusters with multiple nodes need to get data from it. A parallel file system allows concurrent access to data by multiple nodes of a cluster. Because the amount of data is growing rapidly and the models are increasing in complexity, there is a strong growth of usage of parallel file systems.
Panasas have developed their own efficient version of a parallel file system. They solved the problem of data access with a proprietary file system and protocol. Their solution is finely balanced with respect to disk, memory and networking, leading to an industry-leading memory bandwidth.
"Our proprietary file system has almost twice the bandwidth of comparable Lustre file systems", stated Barbara Murphy, "leading to a smaller footprint, lower power usage, and a massive performance."
Another advantage of the Panasas solution is the ease of use. There is no need for extended system integration experience. Just plug in, configure the network and IP addresses and you are done.
In addition, Panasas is making their client open-sourced, allowing customers to link in other file systems, and allowing them to be not wholly dependent upon Panasas systems.
A striking example of the performance is seen at a client in the financial market, who has enormous amount of stock market data, and needed to perform market simulations with these data with results on the same day. The data transfer bottleneck made this impossible, until they used the Panasas solution, which allowed daily market simulations.
Clients are in fields including amongst others oil and gas, finance, aeronautics, genomics, and weather modelling. They focus on commercial clients, where they have 70% of clients.