BlueArc's Titan is a highly scalable system, delivering customers a solution that meets both their capacity and performance needs. Titan scales to 256 terabytes in a single file system, or up to 4 petabytes in a Cluster Name Space (CNS) environment which creates a unified name space and directory structure across many different file systems.
The ability to scale the storage solution behind a single Titan while maintaining the ease of management of such a system reduces the total cost of ownership. Combined with the ability to dynamically grow the storage on an existing storage system, BlueArc provides a scalable storage solution that can grow in capacity as needs expand.
The Titan Server supports volumes of up to 256TB in size. This enables customers with large data sets to create a single volume that users can mount reducing data management, versus creating multiple volumes that must be mounted individually by a user to access the required data and requires excessive data management and partitioning.
Increasing the number of external hardware RAID controllers installed in the BlueArc Storage System can scale the performance of the system as well as capacity, as the act of adding these hardware RAID controllers means that bandwidth and transactional performance is increased on the back-end SAN storage sub-system. This increases the total available processing power of the storage sub-system, enabling the system to handle more client IO's or higher bandwidth read/write requests. This is in stark comparison to other systems that perform the RAID calculations within the filer, and thus gets slower as capacity is increased. Titan provides a scalable environment and allows storage solutions to be built behind the Titan that can scale to meet BlueArc's clients needs.
Titan's modular storage allows a SAN backend to be designed to meet the varying requirements of the customer. By optimizing the number of disks and the type of disk technology behind each RAID controller pair a SAN can be designed that can support both high performance and higher capacity lower cost requirements. RAID controllers usually support many more disk drives than it takes to saturate a controller, thus if performance is the goal, optimization is with fewer drives and more controllers, to provide higher performance and less latency, usually using high performance Fibre Channel drives. If capacity and costs are the goal, then optimization is with more drives and Serial ATA drives, providing higher capacity and reduced costs. The trade off is transactional and throughput performance as well as slightly higher latency. Titan Tiered Storage capability is designed to meet both requirements within a single unified storage system.
The user scalability as well as the storage scalability of Titan is only achievable as a result of its hardware architecture and hardware-based file system as the task processing is very high and the latency associated with the server are very low, enabling more clients to be served at any given point in time. Server latency is key to the user scalability of a server. Titan maintains this rapid response even under load as each part of the architecture is designed to handle the maximum load simultaneously with distributed memory, state engines and pipelines that enable this massive parallelism.
The key for a server to be able to scale in performance, capacity and users lies in the latency of the servers. Latency is a measure of how quickly a server can respond to clients' requests - the higher the latency, the lower the number of clients that can be supported by a single server as the client has to wait longer time periods. If a server has high latency and a large user population, client requests may timeout, resulting in the client machine re-issuing the request; the result is an increase in network and server load.
Latency and throughput are inversely proportional to each other, meaning that the higher the latency of a server, the lower the throughput (performance) of the server, and visa-versa. A low latency server means that the server is available to service client requests as clients do not have to wait long periods for server responses. A low latency server can therefore service more client requests and provide a higher sustained throughput, versus a server that has higher latency figures.
Server latency is a key measurement of the scalability and performance of a storage system. The design of the Titan provides a very low latency server as all data movement is in hardware, with dedicated resources per function. This means that the Titan does not suffer from what is the normal wait times that a traditional server based on the PC-based architecture has for it to access the central memory pool, central CPU and participation in IO bus arbitration.