Multiple quality of service file system
The invention relates to a multiple QoS file system and methods of processing files at different QoS according to rules. The invention allocates multiple VLUNs at different qualities of service to the multiQoS file system. Using the rules, the file system chooses an initial QoS for a file when created. Thereafter, the file system moves files to different QoS using rules. Users of the file system see a single unified space of files, while administrators place files on storage with the new cost and performance according to attributes of the files. A multiQoS file system enhances the descriptive information for each file to contain the chosen QoS for the file.
The present invention relates to management of file systems and large files.
This application incorporates by reference herein as follows:
U.S. application Ser. No. 10/264,603, Systems and Methods of Multiple Access Paths to Single Ported Storage Devices, filed on Oct. 3, 2002;
U.S. application Ser. No. 10/354,797, Methods and Systems of Host Caching, filed on Jan. 29, 2003;
U.S. application Ser. No. 10/397,610, Methods and Systems for Management of System Metadata, filed on Mar. 26, 2003;
U.S. application Ser. No. 10/440,347, Methods and Systems of Cache Memory Management and Snapshot Operations, filed on May 16, 2003;
U.S. application Ser. No. 10/600,417, Systems and Methods of Data Migration in Snapshot Operations, filed on Jun. 19, 2003;
U.S. application Ser. No. 10/616,128, Snapshots of File Systems in Data Storage Systems, filed on Jul. 8, 2003;
U.S. application Ser. No. 10/677,560, Systems and Methods of Multiple Access Paths to Single Ported Storage Devices, filed on Oct. 1, 2003;
U.S. application Ser. No. 10/696,327, Data Replication in Data Storage Systems, filed on Oct. 28, 2003;
U.S. application Ser. No. 10/837,322, Guided Configuration of Data Storage Systems, filed on Apr. 30, 2004;
U.S. application Ser. No. 10/975,290, Staggered Writing for Data Storage Systems, filed on Oct. 27, 2004;
U.S. application Ser. No. 10/976,430, Management of I/O Operations in Data Storage Systems, filed on Oct. 29, 2004; and
U.S. application Ser. No. 11/122,495, Quality of Service for Data Storage Volumes, filed on May 4, 2005.
Data storage systems today must handle larger and more numerous files for longer periods of time than in the past. Thus, more than in the past active data is a shrinking part of the entire data set of a file system leading to inefficient use of expensive high performance storage. This impacts data storage backups and lifecycle management/compliance.
As background, a file is a unit of information stored and retrieved from storage devices (e.g., magnetic disks). A file has a name, data, and attributes (e.g, the last time it was modified, its size, etc.). A file system is that part of the operating system that handles files. To keep track of the files, the file system has directories. The directory contains directory entries which in turn consist of file names, file attributes, and addresses of the data blocks. Unix operating systems split this information into two separate structures: an i-node containing the file attributes and addresses of the data blocks and directory entries containing file names and where to find the i-nodes. If the file system uses i-nodes, the directory entry contains just a file name and an i-node number. An i-node is a data structure associated with exactly one file and lists that file's attributes and addresses of the data blocks. File systems are often organized in a tree of directories and each file may be specified by giving the path from the root directory to the file name.
To address inefficient use of expensive high performance data storage, third party archiving and hierarchical storage management (HSM) software migrate data from expensive high performance storage devices (e.g., Fibre channel) to lower cost storage devices such as tape or Serial ATA storage devices.
Archival and HSM software must manage separate storage volumes and file systems. Archival software not only physically moves old data but removes the file from the original file namespace. Although symbolic links can simulate the original namespace, this approach requires the target storage be provisioned as another file system thus increasing the IT administrator workload.
Archival and HSM software also don't integrate well with snapshots. The older the data, the more likely it is to be part of multiple snapshots. Archival software that moves old data does not free snapshot space on high performance storage. HSM software works at the virtual file system and i-node level, and is unaware of the block layout of the underlying file system or the block sharing among snapshots when it truncates the file in the original file system. With the two data stores approach, the user quota is typically enforced on only one data store, that is, the primary data store. Also, usually each data store has its own snapshots and these snapshots are not coordinated.
Archival software also does not control initial file placement and is inefficient for a large class of data that ultimately ends up being archived. Since archival software is not privy to initial placement decisions, it will not provide different quality of service (QoS) in a file system to multiple users and data types.
Archiving software also ends up consuming production bandwidth to migrate the data. To minimize interference with production, archiving software typically is scheduled during non-production hours. They are not optimized to leverage idle bandwidth of a storage system.
NAS applications may create large files with small active data sets. Some examples include large databases and digital video post-production storage. The large file uses high performance storage even if only a small part of the data is active.
Archiving software has integration issues, high administrative overhead and may even require application redesign. It may also require reconsideration of system issues like high availability, interoperability, and upgrade processes. It would be desirable to eliminate cost, administrative overhead, and provide different QoS in an integrated manner.
SUMMARY OF THE INVENTIONThe invention relates to a multiple QoS (multiQoS) file system and methods of processing files at different QoS according to IT administrator-specified rules. The invention allocates multiple VLUNs at different qualities of service to the multiQoS file system. Using the IT administrator-specified rules, the file system can assign an initial QoS for a file when created. Thereafter the file system moves files to a different QoS using IT administrator-specified rules. Users of the file system see a single unified name space of files. A multiQoS file system enhances the descriptive information for each file to contain the QoS of the file.
BRIEF DESCRIPTION OF THE DRAWINGS
The following description includes the best mode of carrying out the invention, illustrates the principles of the invention, uses illustrative values and should not be taken in a limiting sense. The scope of the invention is determined by reference to the claims. Each part or step is assigned its own number in the specification and drawings. Many features of the invention will be now described using the phrase quality of service or simply QoS. This phrase is not essential to the invention. It is merely used to distinguish between different levels of performance and/or reliability.
Each host runs an operating system such as Linux, UNIX, a Microsoft OS, or another suitable operating system. Tanenbaum, Modern Operating Systems (2001), Bovet and Cesati, Understanding the Linux Kernel (2001), and Bach, Design of the Unix Operating System (1986) describe operating systems in detail and are incorporated by reference herein.
Each host includes a bus adapter 22 between the CPU-memory bus 14 and an interface bus 24, which in turn interfaces with network adapters 17 and 26. The first host 18 communicates through the network adapter 17 over link 28 with the local area network (LAN) 30 with other hosts. The first host 18 also communicates through the network adapter 26 over a link 21 with a storage interconnect network 29. Similarly, the second host 19 communicates over links 38 and 39 with the LAN 30 and the storage interconnect network 29, respectively. The storage interconnect network 29 also communicates over links 32, 34, and 36 with the data storage subsystems 44, 46, and 48, respectively. In sum, the hosts 18, 19 and 20 communicate with each other, the LAN 30 and storage interconnect network 29 and data storage subsystems 44, 46, and 48.
The LAN 30 and the storage interconnect network 29 can be separate networks as illustrated or combined in a single network, and may be any suitable known bus, SAN, LAN, or WAN technology such as Fibre Channel, SCSI, InfiniBand, or Ethernet, and the type of interconnect is not essential to the invention. See Kembel, The FibreChannel Consultant, A Comprehensive Introduction (1998), Kembel, The FibreChannel Consultant, Arbitrated Loop (1996-1997) The FibreChannel Consultant, Fibre Channel Switched Fabric (2001), Clark, Designing Storage Area Networks (2003), Clark, IP SANs: A Guide to iSCSI, iFCP, and FCIP Protocols for Storage Area Networks (2002) and Clark, Designing Storage Area Networks (1999), which are incorporated by reference herein.
U.S. application Ser. No.10/677,560, Systems and Methods of Multiple Access Paths to Single Ported Storage Devices, filed on Oct. 1, 2003 describes suitable data storage subsystems, and is incorporated by reference herein. In alternative embodiments, any suitable controller and compatible storage device(s) can be used (e.g. tape drives or semiconductor memory) in the data storage subsystem. Massiglia, The RAID Book: A Storage System Technology Handbook (6th Edition, 1997) describing RAID technology is incorporated by reference herein.
A host may access secondary storage devices (e.g., hard disk drives) through a VLUN (virtual logical unit) that abstracts the storage device(s) as a linear array of fixed-size blocks. A logical block address (LBA) identifies each fixed-sized block. The data storage system constructs a VLUN from all or parts of several physical storage devices such as disk drives. To make a large VLUN, a data storage system may concatenate space allocated from several storage devices. To improve performance, the data storage system maps adjacent regions of VLUN space onto different physical storage devices (striping). To improve reliability, the system holds multiple copies of a VLUN on different storage devices (mirroring).
In operation, a user requests an I/O operation of one of the hosts 18, 19, or 20 which will transmit the request on the LAN 30 or the storage interconnect network 29 to one or more of the data storage subsystems 44, 46, or 48.
If a write is received, the data storage subsystem 44 can use a write-through scheme and not acknowledge the write until the data is written to nonvolatile memory (e.g., disk array 43). This ensures data consistency between the host and data storage subsystem in the event of a power failure, etc.
In a write-back scheme, the data storage subsystem 44 acknowledges the write before data is written to disk array 43 and stores the data in nonvolatile memory (e.g., battery backed RAM) until written to the disk array to ensure data consistency.
Provisioning a MultiQoS File System
A multiQoS file system can be provisioned by specifying the initial, incremental and maximum capacities of the storage or specifying the initial, incremental, and maximum storage for each QoS VLUN. Or a multiQoS file system can be provisioned by specifying the overall initial, incremental, maximum storage and providing percentages for each QoS.
The provisioning can be also driven by rules.
The UI can be implemented in client software or in a client-server architecture. If the UI is implemented as a Web application, the IT administrator can open a browser (e.g., Microsoft Internet Explorer or Firefox) on management client 112, request a Web form (
The management controller 110 next transmits the user capacity values to the first host 18 that allocates a VLUN in memory 15 at each QoS. The file system provides capacity on a VLUN to place file system core structures (e.g., boot block, super block, free space management, i-nodes, and root directory). For example, the management controller 110 can place the core file system structures in the highest QoS VLUN.
To format a multiQoS file system, the file system writes the core structures into the chosen VLUN. The file system then initializes space allocation data structures in all of the VLUNs assigned to the multiQoS file system. In an embodiment, the file system maintains a high water mark for each VLUN that indicates how far in each VLUN the file system has initialized space allocation information. In an embodiment, the multiQoS file system formats a limited amount of space allocation information such as 32 megabytes (MB). If the file system runs out of the initial 32 MB allocated to a VLUN, it can format the next 32 MB and updates the high water mark to show where to format the next increment of space for that VLUN.
Expanding a MultiQoS File System
After the IT administrator creates a multiQoS file system, a VLUN at a certain QoS and attached to the file system may run short on space. When the multiQoS file system reaches the high water mark that indicates how much capacity has been used up for a VLUN, it requests additional space be allocated to that VLUN, the management controller 110 expands the VLUN corresponding to the QoS and notifies the file system of the expansion. The file system formats the space allocation information in the VLUN to account for the new space. The IT administrator can specify a spill-over rule where instead of expanding the exhausted QoS VLUN, the new data may be spilled over into higher or lower QoS VLUNs that are already allocated to the multiQoS file system. As an example, the rule could enable spill over when allocated space utilization is below a threshold (e.g., 40% of total storage capacity).
The IT administrator can also add a new QoS to the multiQoS file system. In that case, the management controller 110 will allocate a new VLUN at the new QoS and attach it to the multiQoS file system. The file system formats all or a portion of the space allocation information in the new VLUN. The IT administrator will also need to update rules that select the QoS for files to use the new QoS. A later section describes how to change the rules.
Compacting and Shrinking a MultiQoS File System
The IT administrator can compact a multiQoS file system by migrating all files from the VLUN to be vacated to remaining VLUNs. Once a VLUN is completely empty, it can be returned to the storage pool, thus shrinking the storage allocated to the multiQoS file system. This migration can be done by adding a rule or it can be done on demand as described in the section on synthetic namespace below.
Creating a File in a MultiQoS File System
When a user creates a new file in a multiQoS file system, the file system checks the rules associated with the file system to select the initial QoS for the file and its attributes. The file system then allocates blocks for the file from the VLUN assigned to the file system with the desired QoS.
In some protocols, such as Common Internet File System (CIFS), applications can specify the amount of space to reserve for the file. The file system can use the reserved space information to estimate the eventual size of the file and in turn use that estimate in the rules. For example, if the rules place files larger than 1 gigabyte on low QoS storage and the CIFS application reserves four gigabytes (GB), the file system will place such a file on low QoS storage. Norton et al., Storage Networking Industry Association, Common Internet File System (CIFS)—Technical Reference Revision: 1.0 (2002) describe the details of CIFS and is incorporated by reference herein.
Other protocols, such as Network File System (NFS), do not permit specifying the intended size of a file. Thus, an IT administrator can specify rules storing part of a file (e.g., first gigabyte) at one QoS and another part at another level. A multiQoS file system can also indicate the QoS of a block by using the top bits of the block address so a file can have blocks at different qualities of service levels.
Establishing Initial Placement Rules
The IT administrator can specify initial placement rules that establish QoS by file type. Many operating systems support two-part file names. For example, in a file named “file1.PDF”, the extension PDF is the file type. Linux and Unix also support three-part file names such as “file1.PDF.Z.” The extensions (“PDF” and “Z”) indicate the file type is PDF compressed with the Ziv-Lempel algorithm.
Another placement rule is to place the files according to user ID or group ID. For example, an email service provider could use the rule to place emails belonging to premium customers in high QoS storage.
Another placement rule is to place files by file size. For example, a university administrator may restrict very large files typically downloaded by students to low QoS despite quota rules that might have allowed them to be placed on a higher QoS.
Another placement rule is to place files by folder. All files in a particular folder of the file system are placed in the same QoS VLUN. Placement by folder allocates differential QoS storage to projects as a single file system.
Migration Rules
The IT administrator can specify other migration rules.
It is suggested to migrate a file in a chunk (also referred to as an extent) in a background process rather than all at once to avoid adverse impact to the bandwidth of the storage interconnect network. An IT administrator can define the chunk size also referred to as the migration size in terms of MB. A single migration size can be used for all migration whether up or down as shown in
The multiQoS file system can set a file activity rule to trigger migration of a file. Reading and writing to a file over time is a measure of file activity.
To illustrate, assume the IT administrator sets a rule that if a file is not accessed once in a month, it should migrate from high to low performance storage as represented by high QoS VLUN to low QoS VLUN. We also assume if a file is accessed more than once in a month, it should migrate from low to high performance storage. We look at one month in this example, but the time period can be shorter or longer. Finally, we assume steps 1-3 and 7 occur in the month. At step 1, the first client reads file A0, the second client reads C1, the third client accesses the file F, writing versions F1-F3, and the Nth client reads file H0. At step 2, the host stages the active files in cache memory as appropriate. At step 3, the host runs a background process that checks file attributes, applies the rules and identifies all files that need to migrate.
Based on this, the host migrates inactive file B0 from high to low performance storage. To accomplish this, the host stages file B0 into cache at step 4. Further, the host writes file B0 to the low QoS VLUN at step 5. At step 6, the host updates the directory entry or i-node of file B0 to indicate it is now in the low QoS VLUN. At step 7, the host identifies file F was repeatedly accessed during the month so must migrate from low to high performance storage. At step 8, the host stages file F3 into cache, and at step 9 writes file F3 to the high QoS VLUN. At step 10, the host updates the directory entry or the i-node of F3 to indicate its blocks are in the high QoS VLUN. A background process writes the files to secondary storage when appropriate in either a write-back or write-through scheme.
MultiQoS File System Representation
A block can point to additional block addresses.
The total address space represented by the bits in the block address (e.g. 64-bits or 32-bits) can be partitioned statically among the multiple VLUNs of the multiQoS file system. A fixed or variable number of the bits in the block address is used as an index to look up the corresponding VLUN, while the remaining bits are used to determine the address of the block within that VLUN. Such static partitioning allows each volume to grow independently to a very large maximum limit. In an embodiment, the highest order bits of the block address may be used as index into a table of VLUNs and the remaining bits be used to determine the block address in that VLUN. In an embodiment with two VLUNs, the file system can map one VLUN from the lowest address and grow the second VLUN in reverse from the highest address so that they grow together and better use the entire address space.
Large File Extent Migration and Access Tracking
An IT administrator can specify that the migration rules be applied to each extent (i.e., a contiguous allocation of blocks) of a large file. A large file is larger than a certain size such as 1 GB.
To illustrate large file extent migration, assume the IT administrator sets a file activity rule that if any extent of a large file is not accessed once in a month, it migrates from high to low performance storage represented by high QoS VLUN to low QoS VLUN. Also assume if an extent of a large file is accessed more than once in a month, it migrates from low to high performance storage.
The file system uses the access tracking information in the extent attributes to select the QoS for each extent of the large file. When the file system migrates an inactive extent as defined by the IT administrator rules, the file system updates the QoS information in the extent attributes and performs the actual migration as described earlier in
In an embodiment, the file system maintains a cache of access tracking information for a large file in host main memory and only saves the information to extent attributes periodically to reduce the overhead of maintaining the information.
Migrating Files to Different Qualities of Service
As time elapses from the initial creation of a file, the rules associated with the multiQoS file system may indicate that the file should move to a different QoS. For example, the rules might state that files not accessed in a month move to low QoS storage. Likewise, the rules might state that a file in low QoS storage should move to high QoS storage if modified. Alternatively, the IT administrator can manually direct the file system to migrate a file or set of files to a different QoS.
The file system discovers the need for a change in the QoS for a file by either an access to the file or by the file system scanning its files in a low priority background operation. In an alternative embodiment, a certain percent (e.g., 5%) of the total bandwidth of the data storage system can be reserved for scanning and/or migration. In either case, the file system triggers an activity to move the file to the desired QoS while maintaining access to the file and all other files in the file system. If the background activity of migration is run at a lower priority than production data it can be preempted as required. While production activity may continue while migration is in progress, migration rules may continue to be affected. It is suggested that, once begun, the migration of a file, an extent of a large file, or a large file be allowed to complete. Further, a recently migrated file or extent is prevented from migrating again until a reasonable time period has expired to prevent “thrashing” that is constant movement of files and extents back and forth between different QoS.
Referring to step 300 of
If the variable I is not greater than the last i-node at step 302, the host tests if the file of that i-node is identified for migration at step 304. The file is identified for migration in the file attributes, for example, by setting a migration identifier. If the file is not identified for migration, the host computes a new QoS for the file using the migration rule(s). In an embodiment, the host compares migration rule(s) to rule attribute(s) at step 306. In another embodiment, the host compares migration rule(s) to a value such as file size or capacity threshhold at step 306. At step 308, the host tests if the current QoS equals the new QoS computed at step 306. If not, the host sets a migration identifier in the file attributes at step 310 to identify the file for migration. At step 312, the host migrates the file to the new QoS VLUN as illustrated in
With regard to the method of scanning, the file system may use a B-tree to scan for files requiring migration where the leaf nodes are linked to siblings. The scan visits the first (i.e., leftmost) leaf node and follows the chain of links to the right to cover all the objects in the file system. As objects are added to and deleted from the B-tree, the B-tree needs to be rebalanced to ensure all the objects are the same distance from the root (i.e., the B-tree must treat all children the same). Rebalancing can change the sibling links that connect leaf nodes. To avoid interference with such rebalancing, a scan will place a lock on the B-tree to prevent modifications. However, holding a lock on the B-tree during the entire scan can impact production I/O.
In another embodiment, a method of scanning can be implemented to eliminate the need for holding a lock on the B-tree during the entire scan. The method yields the lock repeatedly during the scan for any rebalancing that might be pending.
Referring to step 301 of
Referring to step 400 of
If the variable I is not greater than the last i-node at step 402, the host checks size of the file and/or the file attributes to determine if the file is a large file at step 404. If not, the host performs the method illustrated in
The steps below can be performed in parallel and in a different order as long as it results in migration of a file between QoS. Referring to step 200 of
Alternatively, the host can allocate at the outset the entire space required for a file (or extent) identified for migration on the target VLUN. This provides a contiguous allocation of blocks, that is, less fragmentation of the migrated file (or extent).
The steps below can be performed in parallel and in a different order as long as it results in migration of a file between QoS. Referring to step 201 of
In an embodiment, if the underlying multiQoS file system supports snapshots, the file system that migrates data to a different QoS migrates the blocks for all snapshots sharing the latest version of the data rather than allocating a whole new copy of the data as copy-on-write snapshots usually require.
While migrating a file to a different QoS, the file system may not have enough space in the new QoS to perform the migration. In that case, the file system sends an alert to trigger automatic expansion of the VLUN associated with the QoS or to notify of the space constraint.
Additional Access Patterns
As described above, a multiQoS file system uses the access time information available from file attributes to choose QoS. In an embodiment, a multiQoS file system tracks additional access information to avoid overreacting to stray references to files. For example, a multiQoS file system can associate an additional 32-bits to track reads and an additional 32-bits to track writes in the i-node information for each file. Each bit in these new fields corresponds to one day of access. The least significant bit corresponds to the day of the most recent access as indicated in the current i-node fields “atime” (read time), “mtime” (write time), or “crtime” (create time). The next bit corresponds to access of a day prior to the most recent access, and so on. Each 32-bit field shows accesses for approximately one month. In another example of access pattern tracking, a multiQoS file system can have rules such as accessed five days in a row or accessed four times in the last month.
Alternatively, the access pattern records may not be stored in the i-node, and instead may be stored in a system file or files. The system file or files will be indexed by the i-node. These system files are not visible to end-user and used by the file system.
Alternatively, the access pattern record of a file may be stored as an object in the B-tree that contains all the file system objects. The object ID for the access pattern record for a file would be associated with the file's i-node or be calculated from the file's object ID by replacing the type ID from type=i-node to type=access record.
Changing QoS Rules in a MultiQoS File System
After creating a multiQoS file system, the IT administrator may need to change the rules controlling the selection of QoS. For example, the IT administrator may add a new QoS to a multiQoS file system and need to add or change rules to make use of the new level.
After modifying the rules associated with a multiQoS file system, the existing files may no longer have the desired QoS. The file system determines the correct QoS for each file when accessed or scanned using the new rules and migrates the file if needed.
Migrating From a VLUN in a MultiQoS File System
An IT administrator may need to move the data of a multiQoS file system off a VLUN. For example, a VLUN may become badly fragmented or may be allocated on data storage subsystems that need to be removed from the data storage system. If the IT administrator wishes to remove a QoS from a multiQoS file system, he can change the rules so that no rule permits use of the obsolete QoS. After the file system has completely swept the multiQoS file system and migrated all files away from the obsolete QoS, the management software can detach the obsolete VLUN from the file system and delete the VLUN. In an embodiment, the IT administrator can create a replacement VLUN for an existing QoS in a multiQoS file system and migrate all files with blocks on the obsolete VLUN to the new VLUN.
Synthetic Namespace Views
A multiQoS file system provides a uniform view of the files as a single set to the IT administrator who may want to see which files the system has stored at each QoS. The multiQoS file system provides special subdirectories with names like “.Iowqos” and “.highqos” that show the files stored at particular QoS. At any directory level, listing the contents of “.lowqos” shows only the files in the directory level assigned to the low QoS. To implement the special directories in stateless protocols like the NFS, the multiQoS file system adds the desired QoS to some unused field in the NFS file handle for the directory of interest. For example, the file handle for the directory “/a/b/c/.highqos” lists only the files in “/a/b/c” with high QoS. The multiQoS file system synthesizes a file handle for “/a/b/c/.highqos” using the file handle for the directory “/a/b/c” and with the new field in the file handle stating that the user wants only high priority files.
The multiQoS directory reading functions (corresponding to the NFS operations READDIR and READDIRPLUS) use the new field in the file handle for a directory and if not set, return all the files in the directory and if set, return only the files for the desired QoS. Brent Callaghan, NFS Illustrated (2000) describes the details of NFS and is incorporated herein by reference. In an embodiment, the multiQoS file system does not show the special directories.
A large file that has blocks in different QoS VLUNs will appear in all the synthetic QoS folders. This is implemented by tracking all the QoS levels used by the file in its i-node. In an embodiment, this is a bitmap with each bit corresponding to a QoS level. The IT administrator can specify rules in the UI using various file's attributes including: the size of the file, the time since the file's creation, the time since any user read the file, the time since any user modified the file, the owner of the file, the folder or directory containing the file, and the amount of free space in each QoS allocated to the file system.
The IT administrator rules can be combined to develop additional rules. For example, a rule may specify “.mp3” files go to low priority storage, and all other files created, read, or modified in the last month to high priority storage. The rules can select different qualities of service for user data as opposed to file system metadata (e.g., directories, indirect blocks, and i-nodes). IT administrators may save a set of rules so they can use them on many multiQoS file systems to enforce uniform policies.
Claims
1. A method of provisioning a multiQoS file system on a host, comprising:
- allocating a high QoS VLUN and a low QoS VLUN;
- creating file system core structures in the high or low QoS VLUN;
- initializing space allocation structures in the high QoS VLUN and the low QoS VLUN; and
- creating a fixed table that maps indexes into VLUN identifiers.
2. The method of claim 1, wherein the step of allocating includes computing user capacity required for the high QoS VLUN and the low QoS VLUN.
3. The method of claim 1, further comprising detecting the high QoS VLUN or the low QoS VLUN are running out of space and notifying the file system of the need for expansion of the high QoS VLUN or the low QoS VLUN.
4. The method of claim 1, further comprising adding a new QoS to the file system, allocating a new VLUN at the new QoS, initializing a new space allocation structure in the new VLUN and updating a rule to use the new QoS.
5. The method of claim 4, wherein the new QoS is between the high QoS and low QoS.
6. The method of claim 4, wherein the new QoS is below the low QoS.
7. The method of claim 4, wherein the new QoS is above the high QoS.
8. A method of processing files in a multiQoS file system, comprising:
- (a) reading a migration rule;
- (b) reading a file having an attribute, a current QoS, and blocks;
- (c) testing if the file is identified for migration and if not:
- (d) computing a new QoS by comparing the migration rule to the attribute; and
- (e) testing if the current QoS equals the new QoS and if not, indicating that migration is in progress.
9. The method of claim 8, wherein the migration rule includes a value of file activity.
10. The method of claim 8, wherein the migration rule includes a value of capacity threshold.
11. The method of claim 8, wherein the migration rule includes a value of file size.
12. The method of claim 8, wherein the extension of the file name defines the current QoS.
13. The method of claim 8, wherein the file attribute includes a migration flag and step (e) includes setting the migration flag.
14. The method of claim 8, wherein the file attributes and the addresses of the blocks are held in an i-node of the multiQoS file system.
15. The method of claim 8, further comprising a step of (f) migrating the file to a VLUN having an identifier corresponding to the new QoS.
16. The method of claim 15, wherein the method repeats steps (a)-(f) for another file.
17. The method of claim 15, wherein the blocks have addresses, further comprising a step (g) allocating blocks in the new QoS VLUN; (h) copying the blocks from the old QoS VLUN to the blocks allocated in the new QoS VLUN; and (i) releasing the blocks in the old QoS VLUN.
18. The method of claim 8, wherein step (b) uses a preallocation size from the CIFS protocol as the file size.
19. A multiQoS file system in a host, wherein each file has an attribute, a migration flag, and a QoS, arid blocks, comprising:
- a multiQoS file system; and
- a host coupled to the multiQoS file system to receive a migration rule, to determine if a file is identified for migration and if not compute a new QoS by comparing the migration rule to the attribute, compare the current QoS to the new QoS and if not equal set the migration flag and migrate the file from the current QoS VLUN to the new QoS VLUN.
20. The system of claim 19, wherein the migration rule uses a value associated with a capacity of the current QoS VLUN, a file activity, a file size, or a file type.
21. The system of claim 19, wherein the file attributes and addresses of the blocks are located in an i-node of the multiQoS file system.
22. The system of claim 19, wherein the host reads the current QoS in the block addresses before the file migrates and the host writes the new QoS in the block addresses after the file migrates.
23. The system of claim 19, wherein the host migrates the file in chunks from the current QoS VLUN to the new QoS VLUN.
24. The system of claim 19, wherein the host writes the file to a performance band of a data storage subsystem that corresponds to the new QoS VLUN.
25. The system of claim 19, wherein the host writes the file to an array of storage devices in a data storage subsystem that corresponds to the new QoS VLUN.
26. The system of claim 19, further comprising a management client coupled to a management controller to receive and transmit IT administrator input to the host.
27. The system of claim 26, wherein the management client is configured to receive IT administrator input for user capacity, file type, capacity threshold, migration size, and/or a migration rule.
28. The system of claim 19, wherein the migration rule includes a value of file activity, capacity threshold, file size, and/or file type.
29. A method of identifying files for migration between different QoS in a multiQoS file system, comprising:
- (a) testing if a file of an i-node is identified for migration and if not, computing a new QoS of the file by comparing a migration rule to a rule attribute of the file; and
- (b) testing if a current QoS of the file equals the new QoS computed at step (b) and if not, setting a migration flag to identify the file for migration.
30. The method of claim 29, further comprising a step (c) migrating the file from the current QoS to the new QoS.
31. The method of claim 30, wherein the method performs step (c) after step (b) before performing the method for another file.
32. The method of claim 30, wherein steps (a) and (b) repeat for all files in the multiQoS file system before performing step (c).
33. A method of migration of files from a current QoS VLUN to a new QoS VLUN in a multiQoS file system, comprising:
- (a) assigning a first i-node of the multiQoS file system to a variable I;
- (b) testing if the variable I is greater than the last i-node of the multiQoS file system, and if greater, waiting for the next scan of all of the i-nodes of the multiQoS file system, and if not greater, testing if the file is identified for migration and if not, computing a new QoS for the file using migration rule(s); and
- (c) testing if the current QoS of the file equals the new QoS of the file and if not, identifying the file for migration.
34. The method of claim 33, wherein using the migration rule includes reading a value of file activity, capacity threshold, file size, or file type.
35. The method of claim 33, wherein step (c) includes setting a migration flag.
36. The method of claim 33, further comprising step (d) migrating the file from the current QoS VLUN to the new QoS VLUN.
37. The method of claim 36, wherein steps (a) through (c) are performed on all files in the multiQoS file system before performing step (d).
38. The method of claim 33, wherein the next scan of step (b) runs as a background process, starts after a predetermined time, or starts when a condition is met.
39. The method of claim 38, wherein the condition is based on relative priority of the next scan with respect to other processes running in the host, recent consumption of host CPU time for the method falls below a value, or the amount of time the method has slept.
40. The method of claim 33, wherein the file is identified for migration in the file attributes.
41. A method of migration of a file between QoS in a multiQoS file system when a file offset is less than the total number of blocks of the file, comprising
- (a) placing a read lock on the file;
- (b) finding the block addresses for a chunk of the file starting with the file offset;
- (c) unlocking the read lock and reading the blocks found in step (b);
- (d) allocating new blocks for the chunk of the file in a new QoS VLUN;
- (e) placing a write lock on the file;
- (f) copying the old blocks to the new blocks;
- (g) updating the file attributes to the new QoS;
- (h) putting the old blocks on the free list;
- (i) committing the transaction and unlocking writes; and
- (j) adding the file offset to the chunk size to get the value of file offset.
42. The method of claim 41, wherein the step (g) includes updating the file attributes to point to the new blocks.
43. A method of migration of a file between QoS in a multiQoS file system, comprising:
- iterating through the blocks of a file;
- allocating new blocks in a new QoS, wherein the blocks in each QoS contain an index in certain bits of their block address indicating the QoS;
- copying the data from the old blocks to the new blocks for each block of the file;
- adjusting the attributes of the file to point to the new blocks; and
- freeing the old blocks.
44. The method of claim 43, wherein the step of allocating new blocks includes allocating copying and freeing small chunks at a time.
45. A method of migrating extents of a large file between QoS of a multiQoS file system, comprising:
- (a) testing if the large file is identified for migration, and if not, performing the following steps:
- (b) testing if the extent is part of the large file and if so, performing the following steps:
- (c) computing a new QoS of the extent using a migration rule; and
- (d) testing if the current QoS equals the new QoS, and if not, identifying the extent as in migration; and
- (e) migrating the extent from the current QoS VLUN to the new QoS VLUN.
46. The method of claim 45, wherein using the migration rule includes reading a value of file activity, capacity threshold, file size, or file type.
47. The method of claim 45, wherein step (d) includes setting a migration flag.
48. The method of claim 45, wherein steps (a) through (e) are performed for the extent before steps (a) through (d) are performed on another extent.
49. The method of claim 45, wherein steps (a) through (d) are performed on all extents of the large file before performing step (e) for any extent.
50. The method of claim 45, wherein steps (a) through (e) run as a background process, start after a predetermined time, or start when a condition is met.
51. The method of claim 50, wherein the condition is based on the relative priority of the method with respect to other processes, recent consumption of CPU time by the method falls below a value, or the amount of time the method has slept.
52. The method of claim 45, wherein the extent is identified for migration in the extent attributes.
53. The method of claim 45, wherein the step (b) includes evaluating a migration rule using an extent attribute.
54. The method of claim 45, wherein after steps (a) through (d) are performed on all extents of the large file, the method repeats for the next file in the file system.
55. The method of claim 45, wherein after steps (a) through (e) are performed on all extents of the large file the method repeats for the next file in the file system.
56. A method of migration of a large file including extents, between different QoS VLUNs in a multiQoS file system, comprising
- (a) placing a read lock on each extent;
- (b) finding the block addresses for each extent;
- (c) unlocking the read lock and reading the blocks found in step (b);
- (d) allocating new blocks for each extent in a new QoS VLUN;
- (e) placing a write lock on each extent;
- (f) copying the old blocks to the new blocks;
- (g) updating each extent's attributes to the new QoS;
- (h) putting the old blocks on the free list;
- (i) committing the transaction and unlocking writes; and
- (j)resetting the migration flag.
57. The method of claim 56, wherein the step (h) includes resetting the rule attribute and updating the extent attributes to point to the new blocks.
Type: Application
Filed: Oct 8, 2005
Publication Date: Apr 12, 2007
Inventors: Unmesh Rathi (Sunnyvale, CA), Rex Hamilton (Ben Lomond, CA), Kurt Shoens (Los Altos, CA)
Application Number: 11/245,718
International Classification: G06F 17/30 (20060101);