Best Practices To Eliminate Fragmentation On Modern Storage Technologies
A SAN allows the network administrator to have the remote disk appear locally. Regardless of the protocol used to connect: ISCSI, Fiber Channel, etc.
SAN storage units are called LUNs. Originally, the term LUNs (Logical Unit Number) simply means that SCSI disks are addressed to a particular array of a specific disk, but this is now used to define the physical hard disk array included in a SAN. as logical volume (s)
Storage virtualization usually involves a very large pool of logical data. Through the software, this reserve appears physically stored on all servers. In reality, this data can be stored on hundreds of physical disks distributed over dozens of servers. This is the concept of network-based storage called “Storage Area Networks”.
This technology makes “logical storage” (in the case system – what the OS sees and uses) obsolete from the physical storage device and amalgam into a large group on which a virtual storage container is created.
SAN file systems such as VMFS from VMware or EMC Celerra are known to be system disk share systems and are the backbone of virtual storage. A defragmenter OS only recognizes the files of the “local” disk that it naturally identifies. Vendors of industrial file systems typically include specialized technologies to optimize performance. These file systems are the foundation of virtual storage.
Virtual storage uses metadata to properly channel I / O. Software on a virtual storage device (such as a SAN Switch) will translate logical disk locations on physical disks.
A virtual storage device receives a query for a logical location on LUN # 1, LBA32. The device then proceeds to a course of the metadata at this attention and maps the LUN # 4, LBA16. Then, the device redirects the request to the physical server where the data is stored.
When the data is retrieved, it retransmits the original without the original having transmitted the request never know that it has been resolved from a location different from what he knows.
The fact that there is no one-to-one mapping of file system clusters to LBAs (because of LUN virtualization) is not a problem. Fragmented logical file systems force the OS to generate additional I / O to the virtualization software. By using metadata, the SAN then redirects the I / O from the logical drive to their physical location. The local disk file does not know it and can not control the physical distribution or locations in a virtual storage environment . The consequence of fragmentation is that the NTFS must make multiple requests regardless of whether it is a virtual or physical storage environment.
In the SAN file system, a block size (the smallest virtual unit of measurement) is a parameterizable measure and varies according to the software used. VMware VMFS, for example, is compatible with blocks from 1 MB to 8 M.
Logical Cluster Numbers (LCNs) are a file system built to be used by mapping to the volume of LBAs. The disk controllers take these blocks into account and assign them a physical location. Disk controllers, regardless of their degree of evolution, are not able to record files in a linear fashion. They can not “collect” incoming data based on blocks in a file.
SANs can offer extremely efficient storage systems, but it is not the work of the SAN to deal with fragmented files and evaluate their degree. Industrial technologies employed by one provider may be more efficient at recovering blocks of data than others. The architectures are also different, no matter how efficient the data recovery and combine physical disk limitations can be mitigated, fragmentation will reach the system no matter what happens and well beyond the scope of the technology.
Industrial technologies such as InvisiTasking that eliminates fragmentation directly on the storage device, will bring more sensitive resource for the storage network through a granularity of its actions. With the potential for contention when I / O demand is too large on SANs, it is always recommended to undertake an environmental assessment to determine whether defragmentation times are better timed out of work time.
For more information on fragmentation issues on modern servers, you can download our new white paper: “Best Practices for Removing Fragmentation on Modern Storage Technologies (SAN / RAID / VIRTUALIZATION)”Tags: VMware