Brian S. White, Michael Walker, Marty Humphrey, Andrew S. Grimshaw

1. LegionFS : A Secure and Scalable File System Supporting Cross-Domain High-Performance Applications. Brian S. White, Michael Walker, Marty Humphrey, Andrew S. Grimshaw Department of Computer Science, University of Virginia, Charlottesville.

2. Agenda • The Problem in Focus. • Design of Legion FS. • Fundamental tenets in detail. • A Research Perspective • An Overview of related work. • Conclusion

3. The Problem in focus: • Clients today, are seeking file system to dynamically adapt to varying connectivity, security and latency requirements. • Wide area environments are fraught with insecurity and resource failure due to increasing resources. • Added, the file system should also relieve the user of coping with failures caused due to increasing number of resources incorporated into wide area environments. • A fully-integrated file system infrastructure is in need. • Legion File system architecture supports 5 tenets [Naming, Security, Scalability, Adaptability and Extensibility] which is considered to be fundamental to any system required to meet the above goals.

4. The Object Model Legion is an object- based system comprised of independent, logically address space-disjoint, active objects that communicate via remote procedure calls. Represent coarse grained resources and entities such as users, host, schedulers, files and directories. belongs Class Legion Object Active Inactive State is stored on disk • The complete method signatures exported by an object defines its interface. • Objects may be migrated simply by transferring the internal state to another host. The object’s class spawns a process that is instantiated with migrated internal state.

5. The legion File abstraction is a BasicFileObject, whose methods closely resemble UNIX system calls. • Files and Contexts residing on one host have been aggregated into container processes, called ProxyMultiObjects. • A ProxyMultiObject polls for request and demultiplexes them to the corresponding contained file or context. • Files store data in a Legion Buffer, which achieves persistence through the underlying UNIX file system. • ProxyMultiObjects are lightweight and intended to be distributed throughout the system. They service only a portion of the name space, rather than comprising it in its entirety.

6. The Object Model :ProxyMultiObject Server Loop Address Space BasicFileObject LegionBuffer (Data) BasicFileObject LegionBuffer (data) Native File System

7. Naming • Context Names: Identify Legion objects. • Mapped to unique, location-independent binary names called Legion object identifiers (LOIDs) by the directory service called Context Space. • LOIDS are bound (via a binding process) to low-level Object Addresses for direct object-to-object communication. • Object Address (OA) represents an arbitrary communication endpoint, such as a TCP Socket. • A LOID records the class of an object, its instance number, and a public key to enable encrypted communication. New LOID types can be constructed to contain additional security information. • Context Space is similar to a globally distributed, rooted directory. It is comprised of ContextObjects, which provide mappings from context names to LOIDs in the same fashion that directories map path names to inode numbers. • Unlike directories, ContextObjects may contain references to arbitrary objects.

8. Security • Legion’s distributed, extensible nature and user-level implementation prevent it from relying on a trusted code base or kernel. • No Super user Concept: Individual objects are responsible for legislating and enforcing their own security policies. • Embedding of the public key: Enables secure communication among objects. • Objects are free to negotiate the per-transaction security level on messages, such as full encryption, digital signatures, or clear text. • ACL: Access Control Lists: The authorization is determined by ACLs, an ACL enumerates the operations on an object and the associated access rights of specific principals (or groups of them). • A client can dynamically modify the level of security employed for communication.

9. Scalability • Legion FS distributes files and contexts across the available resources in the system: • No encounter with centralized servers. • Applications enjoy larger bandwidth without contention. • The fully distributed design of Legion FS allows the user to remain ignorant of the constraints of physical disk enclosures, available disk space and file system allocations. • Multiple levels of caching. • Facilitates efficient file and directory lookups and employs limited forms of replication. • Manager objects such as classes become hot spots.

10. Extensibility • Legion FS provides an extensible set of services which can be specialized on an application or domain- specific basis. • Event based protocol stack. • Framework to extend semantics. • Interface inheritance was useful in implementing ProxyMulti-Objects, TwoDFileObjects and Simple K- Copy Classes (SKCC). • SKCC wrap standard classes to provide fault tolerance, by replicating an object’s internal state across a number of user specified storage elements.

11. Adaptability • LegionFS maintains system wide metadata. [<key, value> pairs]. • Attributes are available directly from the object and are also stored in a metadata repository, called the Collection. • Collection allows applications to track the dynamics of the system as well as capitalize on its more stable, inherent diversity. • Legion FS allows a user to explicitly migrate or deactivate an object. • A Legion file retains more exact data concerning access patterns and prefetch schedules.

12. A Research Perspective • While evaluating Legion FS, the paper adopts comparison strategy (between PDB and Legion object), but the comparison would not be a fair one if both the File systems are in different conditions. • The paper begins saying LegionFS was designed to handle hundreds of systems but eventually an experiment showed the fall of the slope in bandwidth vs. number of clients graph. The curve fell at 16 readers itself. A more detailed results of the data should be given. • Also data collected in order to evaluate the performance of LegionFS, result in conclusions which do not support the claims made in the beginning of the paper.

13. Related work---An Overview. • SRB – A middleware that provides access to data stored on heterogeneous resources residing within a distributed system. • WebFS – a kernel-resident file system that provides access to the global HTTP name space. • Ufo employs the UNIX tracing facility to intercept open system calls and transfers whole files from FTP and HTTP Servers. • The PUNCH Virtual File System (PVFS) allows a client executing on a compute server to access files stored within another security domain. • Coda and Odyssey which support application-transparent and application-aware adaptation respectively. • Hurricane File system: employs building blocks to encapsulate file system policies, such as prefetching and distribution.

14. Ficus is a replicated file system that allows kernel or user –level file system modules exporting the vnode interface to be stacked. • The Spring object-oriented operating system is composed of cooperating servers running on a micro kernel. • The FiST language is a high level language for describing stackable file systems. • Self-certifying File System (SFS) embeds a public key in the name of a file, making “self-certifying” pathnames. LegionFS leverages a similar, distributed key management system.

15. Conclusion • The paper proposes a quantitative study of a system that supports some fundamental tenets, anticipating the future of wide area application environment. • While explaining the tenets in detail and evaluating the performance of the File system, some speculations have been made which have to be substantiated with further proofs. • Legion FS is described as a framework promoting extensibility helps wide area applications to exploit dynamics of the system, however these statements are based on certain conditions which are mere speculations. • The means of incorporating application-specific policies is enabled the set of mechanisms afforded by legion.