By a quorum or a Voting disk one understands a component cluster of the manager of a computer cluster for the keeping of the data integrity in the case of a partial failure. In case of failure cluster the Interconnects (the connection between the cluster knots), exists the risk of a fragmentation of the overall system in unwanted-proves autonomously acting units, which nearly always threatens the data integrity (Split Brain problem). By alternate or competitive letter into the logical structure of the Voting disk in case of an interrupted Interconnects one decides, which part of the cluster is to survive. The Voting disk lies on Shared STORAGE.

An example of the case Oracle of the RAC, it survives:

• when asymmetrical division (e.g. 2:3 knot) the larger part
• when straight lines division (e.g. 2:2 knot) the part with the larger Workload.

A such distinction after a loss of the Interconnects as communication channel would be impossible without a "“tuning"” on mass storages. Since nearly all cluster managers react to a loss of the intermediate connection with the restart at least one knot, also the persistente storage of the cluster status is in the Voting disk of advantage: It is void a property part of the renegotiations over availabilities and Masterstati. These negotiations often cause several restarts without the persistente Voting disk. Thus the availability of the single knots rises by being void restart cycles on use of a quorum.

Problems and solutions

The Voting disk actual as soon as it is used - an integral component of the cluster. If a before times available quorum is suddenly no longer seizable during the cluster enterprise, the entire system precipitates. This applies naturally in particular in case of failure for an individual Shared STORAGE. The single developed thereby POINT OF failure to avoid, is at present effort of all manufacturers of cluster commodity.

The common beginning for the solution of these problems is the reflection of the Voting disk over several physical media. Here however again new abysses open themselves:

• The Voting disks must be afflicted with as small a latency as possible guaranteed consistently and.
• A Split Brain scenario with allocation of the Voting disks between the potentially autonomous subunits would be likewise counter productive. Oracle cluster commodity with an odd number of quorums e.g. solves this conflict.

The king solution for consistency of latency and availability problems represents (possibly very expensive) the STORAGE-lateral Replikation in the SAN. It presents all cluster Membern transparency only one made a replication equipment and relieves thereby cluster commodity, cluster Member and administrators.

See also

• Oracle cluster ready services
• Cluster file system

Articles in category "Quorum (computer science)"

We found here 1 articles.

Related Websites

We found here 4 related websites.

• Citations: and availability of quorum systems - Naor, Wool ...
  M. Naor and A. Wool. The load, capacity, and availability of quorum systems. In Proceedings of the 35th IEEE Symposium on Foundations of Computer Science, ...


• Citations: th IEEE Symposium on the Foundations of Computer ...
  The load, capacity and availability of quorum systems. 35th IEEE Symposium on the Foundations of Computer Science, 1994, pp. 214--225.


• Computer Science @ UC Davis | Bylaws of the GGCS
  The Graduate Group in Computer Science (hereafter referred to as the GGCS) is ... 50+% of the members of the GGCS will constitute a quorum for the meeting. ...


• Computer Science: Publication: Clustering in sensor networks using ...
  Clustering in sensor networks using quorum sensing, I Wokoma and L Sacks and IW Marshall, 2003, Computer Science, University of Kent at Canterbury.