The Voystor storage server line ranges from a small SMB quiet (22 dB A) NAS micro tower with either two or four drives 3.5” to a 4U 72 3.5” drive enterprise class storage server. Ask your Nor-Tech sales person about the new Voystor 5000 with ZFS. At the very top end ZFS is now incorporated into Intel® Enterprise Edition Lustre® for distributed parallel storage environments. The ZFS Lustre product comes from Intel’s recent acquisition of Whamcloud and work done at Lawrence Livermore National Laboratory (LLNL) to use Lustre® + ZFS on their 55PB storage for the Sequoia cluster. LLNL did much of the work to port ZFS from the Solaris OS to Linux for what is called ZFS on Linux (ZOL). ZOL is now considered ready for production, see the news release here.
So what is so neat about ZFS™?
- ZFS is one of the most advanced file systems available today
- ZFS was developed by Sun Microsystems which later merged with Oracle Corp.
- ZFS was released to open-source under a CDDL license
- ZFS is both a file system and a logical volume manager (LVM)
- ZFS file systems can store up to 256 quadrillion zettabytes with a zettabyte (ZB) being 270 bytes
- ZFS requires no file system check because it fixes corruption automatically
- ZFS snapshot capability can help incremental backups
- ZFS has RAID0-stripping, mirroring, RAIDz1 single parity like RAID5, RAID-Z2-double parity like RAID6
- ZFS introduced RAID-Z3 triple parity
- ZFS can use dynamic striping that adjusts to the workload
- ZFS has built in data compression
- ZFS has built-in de-duplication
- ZFS uses Copy-on-Write which is resilient against corruption and has no write-hole like normal RAID
- ZFS has tiered storage capability built in so it can use SSD’s for caching
Data Integrity is Priority 1 for ZFS
Integrity of data is checked by using a checksum hash throughout the file system tree. Each block of data has a checksum made for it and that checksum is stored in the pointer to that block. Then the pointer is check-summed and it is saved in the pointer to that data. This checksum operation continues all the way up the data tree to the root which is also check-summed which creates a Merkle tree. Data corruption while data is in-flight is undetectable in most file systems because the data is stored with its checksum. In-flight data corruption is detectable with ZFS because the checksum is stored with the pointer to the data and so on so the entire pool self-validates.
Nor-Tech acquired a lot of expertise in storage because of the complex storage required by our HPC cluster products that we build for customers like the FAA, VA, U of M, U of W, Boeing and Gulfstream. This link talks about some of the storage used for one of the HPC projects. Whether your storage needs are big or small contact Nor-Tech today to learn how we can help you with one of our Voystor storage servers.
Click here for more information about Nor-Tech storage.
Lustre® is a registered trademark of Xyratex
