List of terms

Here are some short descriptions of terms, names and acronyms that I've encountered during the litterature study. More terms might be added in the future.

AFS
Originaly Andrew File System, now only AFS. A distributed file system writen for the Andrew project att Carnegie-Mellon Univeristy, now maintained and distributed by Transarc.

Arla
Arla is a free AFS clone, mainly under development by the computer club Stacken at KTH.

Cache Manager
The Cache Manager is the client side of AFS. There is one Cache Manager running at each workstation.

Coda
Coda is an optimisticaly replicated distributed file system. Coda has support both for r/w replication and for disconnected operation, making it highly available.

One-copy serializability
One-copy serializability means that the result of a read or write to a file is the same regardless if the file is replicated or not. A goal for some replicated file systems, but abandoned in others.

Optimistic replication
An optimistic replication protocol allows writes in both parts of a partitioned distributed file system (or database). This can lead to conflicts in the file system, since conflicting updates can be done. When conflicts occur they are detected and corrected, either automaticaly, or as last resort, by hand.

Pessimistic replication
As opposed to optimistic replication a pessimistic protocol will guarantee consistency and one-copy serializability at all times. This means that in the case of a network partitioning only one partition can allow updates. The main problems in pessimistic protocols are to determine the active partition and to give as high availability as possible in the case of several concurrent partitionings.

Quorum
A quorum is a part of a group large enough to "make a decision". In distributed systems quorums are used to determine wether an operation like a read or a write is allowed. There is often a read quorum (R) and a write quorum (W), so that you must contact at least R nodes to be allowed to read a file and at least W nodes to write to a file. Read quorums and qrite quorums must be assigned in such a way that you are guaranteed to have at least one up to date node in a read quorum. For example in a system with N nodes you can choose the write quorum to be N and the read quorum to be one, or you can set both quorums to N/2 + 1.

VLDB
The Volume Location Database is a database in AFS. The VLDB contains information about which File Server each volume can be found at, the volume's status etc. The VLDB is maintained by the VL Server.

VL Server
The Volume Location Server maintains and exports the VLDB. When a AFS client (Cache Manager) tries to open a file in a certain volume it first asks the VL Server about the volume's location.

[Main]Back to main page
Page maintained by Noora Peura (noora@nada.kth.se).