Advanced Technology for Maximum Performance


WFFS Extreme Performance File Systems

WildFire File System (WFFS) offers dramatically faster read and write performance than other file systems, while also offering state of the art data protection and significant cost savings.


WFFS Product Advantages:


Dramatic increase in random write performance.  Relying upon our patented ESS technology, WFFS coalesces random writes into atomic linear writes.  As a consequence, it always writes at the composite linear write rate of the SSDs or hard disks in a storage array.  Write performance in an ESS environment can be as much as a hundred times faster than alternatives.

Dramatic performance increase in reading files or parts of files.  Because block reads are hardware dependent, WFFS cannot improve block read performance.  But it often improves file reads dramatically by reducing the number of reads required to perform a data read through its proprietary flat metadata model.

Superior data reliability.  WFFS employs a number of techniques which radically increase data and storage system reliability.  Our atomic write model totally avoids read-modify-write contradictions and prevents premature overwrite of old data.  All data is written FIFO, assuring any crash is clean.  Erasure code support can reduce the consequences of media failure.  The ability to write extremely fast also assures that data does not sit around in dirty buffers.  When these are combined with state of the art error recovery techniques, WFFS is as reliable as the most reliable file systems.

Profoundly extends the usable life of Flash.  The ESS component of WFFS radically reduces Flash write amplification of all types.  In a parity RAID environment, ESS typically improves the effective life of SSD media by 10x to 20x.

Concurrent block and File System.  Historically, ESS has only been available as a block device.  The new  technologies in WFFS permit concurrent use of both block and file systems.

Suitable for use with local or remote storage. WFFS significantly reduces the computational resources needed for reading and writing to mass storage.  On the one hand, this means that lower powered, less expensive CPUs can be used in storage appliances.  On the other hand, it also permits greater general performance of systems with on-board storage by reducing the storage burden on the application.

Low memory footprint technology.  Many block and file systems have high memory requirements.  WSSF gets rid of this problem.  The file system management scheme requires only one byte of memory per file.  Thus, a 20TB system consisting of a billion 20KB files requires only a gigabyte of memory, and a petabyte requires only fifty gigabytes.  Similarly, the block system supports optional memory virtualization which can typically reduce memory requirements from eight bytes per 4KB to one-eighth byte per 4KB: 32 megabytes per terabyte.

Profoundly reduces materials costs.  WFFS lets you build mass storage that costs a great deal less.  Hard disk storage costs can be reduced through the use of MSR drives.  Flash SSD costs can be reduced 75% or more.  RAM and CPU costs can be reduced to nominal levels.  All can lead to much more affordable mass storage.

Choice of Linux or Windows.  WFFS uses its own erasure code drive array management software.  This avoids use of the standard RAID drivers built into various operating systems. Consequently, WFFS is not dependant upon the drivers of others and can function in multiple operating system environments.


WFFS Technical Features:


WFFS uses a broad range of techniques in new ways not normally seen in mass storage subsystems.  We discuss some of these technologies below.

ESS atomic writes.  WFFS uses WildFire's patented ESS technology.  Enterprise Storage Stack (ESS) takes clusters of random writes written to a drive set and converts these into linear atomic writes in time received order, complete with applicable metadata.  These atomic writes are bounded by RAID stripes and are fitted to individual flash erase blocks or individual shingles of SMR hard disks. They are written at the composite linear write speed of all the drives present.  They also have virtually no write amplification and as such last much longer that do SSDs performing standard random writes.  Finally, because writes are performed atomically, the data is never inconsistent as a result of a crash.

32-bit CRCs.  Each data block written has a 32 bit CRC code associated with it.  This CRC is compared with the data block at read.  If data checking is turned on, and the CRC does not match the data pattern, the system will try to reconstruct the data block from other elements of the array.  If this is not possible, the drive will be flagged as a problem.

Erasure code "write sets".  Like many others, we have used RAID over the years.  Other than performance level issues, the problem with RAID is that each usage is tied to particular operating system or piece of hardware.

WFFS switches from simple RAID to erasure codes.  While parity raid consists of data together with pre-allocated parity stripes, erasure codes prepare a set of N data chunks and add parity stripes to this.  Unlike standard RAID, which is limited to three parity chunks, erasure codes can support many parity stripes (eight or more) thus reducing the probability of failure.  This produces higher data confidence.  Similarly, erasure codes allow the use of more media than that originally permitted by the array.  See drive modifications, below.

Use of erasure codes let us design a product that can run on both Linux and Windows.

Drive set modifications.  Erasure codes also let you insert or remove drives.

Flat metadata data model and hashed directory structure.  Many file systems use b-trees or similar methods to organize their data.  These work very well when there are small numbers of files, either  globally or locally.  But they become slower as item counts increase levels of data.  For instance, a multi-terabyte ZFS system may require the reading and scanning of ten or even fifteen b-tree levels before data can actually be read.

Conversely, WFFS uses hashes to break each directory and file into much smaller pieces, assuring that only one read is required to identify where a particular file is actually stored.

Variable file sizes.  In WFFS, file sizes are not limited to a single size.  Small files will have variable length as part of the hashed directory structure. Files that are larger than 4KB will use suitable multiples of 4KB.


WildFire Storage:
Phone:   US toll free (888) 473-7866
+1 (610) 237-2000
fax +1 (610) 672-9549
© Copyright 2007-2018 EasyCo LLC - All Rights Reserved - EasyCo LLC dba WildFire Storage