A Look at PVFS, a Parallel File System for Linux

1. A Look at PVFS, a Parallel File System for Linux Talk originally given by Will Arensman and Anila Pillai

2. Overview • Network File Systems (NFS) • Drawbacks of NFS • Parallel Virtual File Systems (PVFS) • Using PVFS • Conclusion • References

3. 1. Network File System (NFS) • NFS is a client/server application developed by Sun Microsystems • It lets a user view, store and update files on a remote computer as though the files were on the user's local machine. • The basic function of the NFS server is to allow its file systems to be accessed by any computer on an IP network. • NFS clients access the server files by mounting the servers exported file systems. For example: where the files really are Your view /home/ann server1:/export/home/ann

4. 2. Drawbacks of NFS • Having all your data stored in a central location presents a number of problems: • Scalability: arises when thenumber of computing nodes exceeds the performance capacity of the machine exporting the file system; could add more memory, processing power and network interfaces at the NFS server, but you will soon run out of CPU, memory and PCI slots; the higher the node count, the less bandwidth (file I/O) individual node processes end up with • Availability: if NFS server goes down all the processing nodes have to wait until the server comes back into life. • Solution: Parallel Virtual File System (PVFS)

5. 3. Parallel Virtual File System(PVFS) • Parallel Virtual File System (PVFS) is an open source implementation of a parallel file system developed specifically for Beowulf class parallel computers and Linux operating system • It is joint project between Clemson University and Argonne National Laboratory • PVFS has been released and supported under a GPL license since 1998

6. 3. Parallel Virtual File System(PVFS) • File System – allows users to store and retrieve data using common file access methods (open, close, read, write) • Parallel – stores data on multiple independent machines with separate network connections • Virtual – exists as a set of user-space daemons storing data on local file systems

7. PVFS… • Instead of having one server exporting a file via NFS, you have N servers exporting portions of a file to parallel application tasks running on multiple processing nodes over an existing network • The aggregate bandwidth exceeds that of a single machine exporting the same file to all processing nodes. • This works much the same way as RAID 0 – file data is striped across all I/O nodes. Data blocks 1 2 3 4 5 6 7 8 9 10 1 3 5 7 9 2 4 6 8 10 RAID 0 (Stripping)

8. PVFS… • PVFS provides the following features in one package: • allows existing binaries to operate on PVFS files without the need for recompiling • enables user-controlled striping of data across disks on the I/O nodes • robust and scalable • provides high bandwidth for concurrent read/write operations from multiple processes to a common file

9. PVFS… • PVFS provides the following features in one package: • ease of installation • easily used - provides a cluster wide consistent name space, • PVFS file systems may be mounted on all nodes in the same directory simultaneously, allowing all nodes to see and access all files on the PVFS file system through the same directory scheme. • Once mounted PVFS files and directories can be operated on with all the familiar tools, such as ls, cp, and rm.

10. PVFS Design and Implementation • Roles of nodes in PVFS: • COMPUTE NODES - on which applications are run, • MANAGEMENT NODE - which handles metadata operations • I/O NODES - which store file data for PVFS file systems. • Note:- nodes may perform more than one role

11. PVFS I/O Nodes • In order to provide high-performance access to data stored on the file system by many clients, PVFS spreads data out across multiple cluster nodes I/O nodes • By spreading data across multiple I/O nodes, applications have multiple paths to data through the network and multiple disks on which data is stored. • This eliminates single bottlenecks in the I/O path and thus increases the total potential bandwidth for multiple clients, or aggregate bandwidth.

12. PVFS System Architecture

13. PVFS Software Components • There are four major components to the PVFS software system: • Metadata server (mgr) • I/O server (iod) • PVFS native API (libpvfs) • PVFS Linux kernel support • The first two components are daemons (server types) which run on nodes in the cluster

14. PVFS Components • The metadata server(or mgr) • File manager; it manages metadata for PVFS files. • A single manager daemon is responsible for the storage of and access to all the metadata in the PVFS file system • Metadata - information describing the characteristics of a file, such as permissions, the owner and group, and, more important, the physical distribution of the file data

15. PVFS Version 1 I/O Node Configuration • Data in a file is striped across a set of I/O nodes in order to facilitate parallel access. • The specifics of a given file distribution are described with three metadata parameters: • base I/O node number • number of I/O nodes • stripe size • These parameters, together with an ordering of the I/O nodes for the file system, allow the file distribution to be completely specified

16. PVFS Components • Example: • pcount - field specifies that the the number of I/O nodes used for storing data • base - specifies that the first (or base) I/O node (is node 2 here) • ssize - specifies that the stripe size--the unit by which the file is divided among the I/O nodes—here it is 64 Kbytes • The user can set these parameters when the file is created, or PVFS will use a default set of values Meta data example for file

17. PVFS Components PVFS file striping done in a round-robin fashion • Though there are six I/O nodes in this example, the file is striped across only three I/O nodes, starting from node 2, because the metadata file specifies such a striping. • Each I/O daemon stores its portion of the PVFS file in a file on the local file system on the I/O node. • The name of this file is based on the inode number that the manager assigned to the PVFS file (in our example, 1092157504).

18. PVFS Components • 2.0 The I/O server (or iod) • It handles storing and retrieving file data stored on local disks connected to the node. • when application processes (clients) open a PVFS file, the PVFS manager informs them of the locations of the I/O daemons • the clients then establish connections with the I/O daemons directly

19. PVFS Components • the daemons determine when a client wishes to access file data, the client library sends a descriptor of the file region being accessed to the I/O daemons holding data in the region • what portions of the requested region they have locally and perform the necessary I/O and data transfers.

20. PVFS Components • PVFS native API (libpvfs) • It provides user-space access to the PVFS servers • This library handles the scatter/gather operations necessary to move data between user buffers and PVFS servers, keeping these operations transparent to the user • For metadata operations, applications communicate through the library with the metadata server

21. PVFS Components • PVFS native API (libpvfs) • For data access the metadata server is eliminated from the access path and instead I/O servers are contacted directly • This is key to providing scalable aggregate performance • The figure shows data flow in the PVFS system for metadata operations and data access

22. PVFS Components Metadata access For metadata operations applications communicate through the library with the metadata server

23. PVFS Components Data access Metadata server is eliminated from the access path; instead I/O servers are contacted directly; libpvfs reconstructs file data from pieces received from iods

24. PVFS Components • PVFS Linux kernel support • The PVFS Linux kernel support provides the functionality necessary to mount PVFS file systems on Linux nodes • This allows existing programs to access PVFS files without any modification • This support is not necessary for PVFS use by applications, but it provides an extremely convenient means for interacting with the system

25. PVFS Components • PVFS Linux kernel support • The PVFS Linux kernel support includes: • a loadable module • an optional kernel patch to eliminate a memory copy • a daemon (pvfsd) that accesses the PVFS file system on behalf of applications • It uses functions from libpvfs to perform these operations.

26. PVFS Components Data flow through kernel toPVFS servers app pvfsd user space VFS /dev/pvfsd kernel space • The figure shows data flow through the kernel when the Linux kernel support is used • Operations are passed through system calls to the Linux VFS layer. Here they are queued for service by the pvfsd, which receives operations from the kernel through a device file • It then communicates with the PVFS servers and returns data through the kernel to the application

27. PVFS Application Interfaces • Applications on client nodes can access PVFS data on I/O nodes using one of the three methods: • Linux kernel interface:- • The Linux kernel VFS Module provides the functionality for adding new file-system support via loadable modules without recompiling the kernel. These modules allow PVFS file systems to be mounted in a manner similar to NFS. Once mounted, the PVFS file system can be traversed and accessed with existing binaries just as any other file system. .

28. PVFS Application Interfaces • Applications on client nodes can access PVFS data on I/O nodes using one of the three methods: • PVFS native API:- • The PVFS native API provides a UNIX-like interface for accessing PVFS files. It also allows users to specify how files will be striped across the I/O nodes in the PVFS system.

29. PVFS Application Interfaces • ROMIO MPI-IO interface:- • ROMIO implements the MPI2 I/O calls in a portable library. This allows parallel programmers using MPI to access PVFS files through the MPI-IO interface

30. 5. Using PVFS 1. Download, untar pvfs, pvfs-kernel files. Available at: http://parlweb.parl.clemson.edu/pvfs/ 2. Go to PVFS directory: • ./configure, make, make install • make install on each node 3. Go to PVFS kernel directory: • ./configure –with-libpvfs-dir=../pvfs/lib • make, make install • cp pvfs.o /lib/modules/<kernel-version>/misc/ • make install, cp pvfs.o on each node

31. PVFS Installation • Metadata server needs: • mgr executable • .iodtab file: contains IP addresses and ports of I/O daemons • .pvfsdir file: permissions of the directory where metadata is stored • Run mkmgrconf to create .iodtab and .pvfsdir

32. PVFS Installation • I/O server needs • iod executable • iod.conf file: describes the location of the pvfs data directory on the machine • Each client needs • pvfsd executable • pvfs.o kernel module • /dev/pvfsd device file • mount.pvfs executable • mount point

33. PVFS Installation • After installation • Start iod’s • Start mgr • Run pvfs-mount

34. PVFS API 1.Initialize a pvfs_filestat struct: struct pvfs_filestat { int base; /* First node. -1 = default = 0 */ int pcount; /* # I/O Nodes. default = all*/ int ssize; /* Stripe size. default = 64K */ int soff; /* Not used. */ int bsize; /* Not used. */ }

35. PVFS API 2.Open the file: pvfs_open(char *pathname, int flag, mode_t mode, struct pvfs_filestat *dist); 3. Have a look at your metadata: pvfs_iocctl(int fd, GETMETA, struct pvfs_filestat *dist);

36. PVFS Utilities • Copy files to PVFS: u2p –s <ssize> -b <base> -n <#nodes> <src> <dest> • Examine file distribution: pvstat <file> <file>: base = 0, pcount = 8, ssize = 65536

37. PVFS at U of A • Hawk • Gigabit Ethernet (100Mbps) • 4 nodes, each with local drives • Raven • Myrinet (~160 MB per second) • 7 40 GB Drives (~60 MB per second)

40. 7. Conclusions • Pros: • Higher cluster performance than NFS. • Many hard drives to act a one large hard drive. • Works with current software. • Best when reading/writing large amounts of data • Cons: • Multiple points of failure. • Poor performance when using kernel module. • Not as good for “interactive” work.

41. 8. References • The Parallel Virtual File System, Available at: http://www.parl.clemson.edu/pvfs/ • P. H. Carns, W. B. Ligon III, R. B. Ross, and R. Thakur, ``PVFS: A Parallel File System For Linux Clusters'', Proceedings of the 4th Annual Linux Showcase and Conference, Atlanta, GA, October 2000, pp. 317-327 • Thomas Sterling, “Beowulf Cluster Computing with Linux”, The MIT Press, 2002 • W. B. Ligon III and R. B. Ross, ``An Overview of the Parallel Virtual File System'', Proceedings of the 1999 Extreme Linux Workshop, June, 1999. • Network File System, Available at: http://www.redhat.com/docs/manuals/linux/RHL-7.2-Manual/ref-guide/ch-nfs.html • http://www.linuxuser.co.uk/articles/issue14/lu14-All_you_need_to_know_about-PVFS.pdf