View source for Fusion-io NVMFS

[[Company:SanDisk|SanDisk]]/[[Company:Fusion-io|Fusion-io]]'s NVMFS [[File system|file system]], formerly known as ''Direct File System'' (DFS),<ref name=dfs>{{cite journal |first1=William K. |last1=Josephson |first2=Lars A. |last2=Bongo |first3=David |last3=Flynn |first4=Kai |last4=Li |title=Dfs: A file system for virtualized flash storage. |journal=ACM Transactions on Storage |volume=6 |issue=3 |date=September 2010 |url=https://www.usenix.org/legacy/events/fast10/tech/full_papers/josephson.pdf|doi=10.1145/1837915.1837922 |s2cid=1715382 }}</ref><ref>{{cite journal |first=Nisha |last=Talagala |title=Native Flash Support For Applications |journal=Flash Memory Summit |date=24 August 2012 |url=http://www.flashmemorysummit.com/English/Collaterals/Proceedings/2012/20120823_S304B_Talagala.pdf}}</ref> accesses [[Flash memory|flash memory]] via a virtual flash storage layer instead of using the traditional block layer API. This [[File system|file system]] has two main novel features. First, it lays out files directly in a very large virtual storage address space. Second, it leverages the virtual flash storage layer to perform block allocations and atomic updates. As a result, NVMFS performs better and is much simpler than a traditional Unix file system with similar functionalities. Additionally, this approach avoids the log-on-log performance issues triggered by [[Log-structured file system|log-structured file system]]s.<ref>{{cite journal |title=Don't stack your Log on my Log |first1=Jingpei |last1=Yang |first2=Ned |last2=Plasson |first3=Greg |last3=Gillis |first4=Nisha |last4=Talagala |first5=Swaminathan |last5=Sundararaman |journal=2nd Workshop on Interactions of NVM/Flash with Operating Systems and Workloads (INFLOW 14) |date=5 October 2014 |url=https://www.usenix.org/system/files/conference/inflow14/inflow14-yang.pdf}}</ref> Microbenchmark results show that NVMFS can deliver 94,000 I/O operations per second (IOPS) for direct reads and 71,000 IOPS for direct writes with the virtualized flash storage layer on top of a first-generation [[Company:Fusion-io|Fusion-io]] ioDrive. For direct access performance, NVMFS is consistently better than [[Ext3|ext3]] on the same platform, sometimes by 20%. For buffered access performance, NVMFS is also consistently better than [[Ext3|ext3]], and sometimes by over 149%. Application benchmarks show that NVMFS outperforms [[Ext3|ext3]] by 7% to 250% while requiring less CPU power.<ref name=dfs/> Additionally, I/O latency is lower with NVMFS compared to [[Ext3|ext3]].<ref>{{cite journal |first=Thomas |last=Rochner |title=Running NoSQL natively on flash |journal=NoSQL Search Roadshow Zurich |url=http://nosqlroadshow.com/dl/basho-roadshow-zurich-2013/slides/ThomasRochner_RunningNoSQLNativelyOnFlash.pdf |date=19 September 2013}}</ref>

==Flash Memory API==
The API used by NVMFS to access flash memory consists of:<ref>{{cite web |first=Dhananjoy |last=Das |title=In a Battle of Hardware, Software Innovation Comes Out On Top |date=14 November 2014 |website=SanDisk |url=http://itblog.sandisk.com/in-a-battle-of-hardware-software-innovation-comes-out-on-top/|archive-url=https://web.archive.org/web/20141129183506/http://itblog.sandisk.com/in-a-battle-of-hardware-software-innovation-comes-out-on-top/ |archive-date=2014-11-29 }}</ref>
* An address space that is several orders of magnitude larger than the storage capacity of the flash memory.
* Read, append and trim/deallocate/discard primitives.
* Atomic writes.<ref>{{cite book |first1=Xiangyong |last1=Ouyang |first2=David |last2=Nellans |first3=Robert |last3=Wipfel |first4=David |last4=Flynn |first5=Dhabaleswar K. |last5=Panda |title=2011 IEEE 17th International Symposium on High Performance Computer Architecture |chapter=Beyond block I/O: Rethinking traditional storage primitives |date=February 2011 |pages=301–311 |citeseerx=10.1.1.300.4140 |doi=10.1109/HPCA.2011.5749738 |isbn=978-1-4244-9432-3 |s2cid=6214993 }}</ref>

The layer that provides this API is called the ''virtualized flash storage layer'' in the DFS paper.<ref name=dfs/> It is the responsibility of this layer to perform block allocation, [[Wear leveling|wear leveling]], [[Garbage collection (computer science)|garbage collection]], crash recovery, address translation and also to make the address translation data structures persistent.

==References==
{{reflist}}

{{Filesystem}}

[[Category:Compression file systems]]
[[Category:Flash file systems]]
[[Category:File systems supported by the Linux kernel]]

{{Sourceattribution|Fusion-io NVMFS}}