Hitachi NAS Architecture for HUS and HUS VM

The Hitachi NAS (HNAS) platform is unique in the industry since it uses a patented hardware accelerated architecture which delivers massive data path parallelization and provides the foundation for Hitachi’s Silicon File System hardware based filesystem. The key is that it separates data movement from data management and utilizes the most appropriate processing resource for a given task. It is not an appliance that runs software on a processor that is responsible for driving all the functions of a file server. This architecture is very similar to the architecture in our enterprise block based storage systems, VSP and HUS VM, which I’ve described in a previous post.

HNAS architecture consists of a Filesystem Board and a Management Board. On the Filesystem Board there are a series of Field Programmable Gate Arrays, FPGA that can pipeline a file request through the Network Interface (NI) to the Data Mover (TFL) to the acceleratedfile system (WFS) and to the Disk Interface (DI). The Management board has CPU and memory to manage the different functions that support the file like clones, tiering and deduplication. The SiliconFS file system is an Object Based File System with meta data that enables it to provide faster cloning of file systems, tiering of files to different pools of storage as well as to external storage including Hitachi Content Platform, and object based replication for faster fail-over and fail-back disaster recovery.

The separation of data management and data movement provides the ability to do active dedupe with no impact to the primary I/O performance. The automatic quality of service of the deduplication function will self-throttle itself at 50% of available system performance (IOPS, throughput), thus prioritizing user and application I/O. Therefore, use and application I/O performance will not be impacted. In the case of reads of deduplicated data, the performance impact should be negligible, due to the pointer-based implementation of the deduplicated file system. Michael Hay provides an easy to understand description of this process in his blog pointing out the advantages of this architecture in dedupe as well as in rehydration and dedupe of the dedupe data base.

This is active dedupe that you can set and forget, and this is only possible with the unique architecture of the Hitachi NAS system.

The latest Hitachi NAS systems are available in three models, which can be deployed as gateways for our block storage systems like VSP or configured in the rack with our HUS and HUS VM storage systems.

Recently we completed SPECsfs measurements using HUS VM with all flash drives for the 2-node  and 4-node HNAS 4100 configurations and the results were outstanding.

See what Bob Madaio has to say about this in his post.

Hitachi NAS platform with its unique Hybrid-Core architecture that separates data management from data movement is similar to the architecture principles of VSP and HUS VM. This architecture provides very high performance due to its ability to pipeline data movement through the FPGAs. It also provides the high levels of scalability that are needed to meet the explosive growth of unstructured data. Active dedupe increases the efficiency of storage capacity with no impact to quality of service. It unifies easily with the Hitachi Data Systems family of block storage systems and can archive, based on policies, to Hitachi Content platform. The Hitachi Data Discovery Suite can perform a content aware search across HNAS object-based file systems without the need to do a tedious directory crawl.  HNAS management is integrated into the Hitachi Command Suite and provides one pane of glass for management and monitoring of Hitachi Data Systems suite of server, file and block products.

For more information, visit our HNAS website.