Distributed lock manager using a passive, state-full control-server

A locking apparatus for use in a multicomputer system wherein multiple data processing nodes, each under control of its own operating system, are joined by a communications network and wherein at least some of the processing nodes control access to shared resources, the locking apparatus comprising:

• a) a lock control server associated with each processing node owning a shared resource, said lock control server comprising:
• i) table means to record lock control status of data objects of the shared resource;
• ii) means responsive to a requested lock on a data object, to report conflicting lock status of the data object;
• b) a plurality of lock managers, each of the lock managers being associated with a node which requires access to a data object whose access is controlled by another node, each of said lock managers comprising:
• i) means to request that lock control in a specified mode be granted on the data object from the lock control server, and if not granted, to receive a list of the lock managers with conflicting control for the data object;
• ii) means to request of and receive acknowledgement from other lock managers of the giving-up of conflicting control of the data object;
• iii) means to locally grant locks in a specified mode on the data object after lock control in an appropriate mode for the data object has been acquired from the lock control server.

A locking apparatus for use in a client node of a multicomputer system wherein multiple data processing nodes, each under control of its own operating system, are joined by a communications network and wherein at least one of the data processing nodes controls access to a shared resource having a plurality of data objects stored therein, the locking apparatus comprising:

• request receiving means, for receiving a lock request on a specified data object from a process executing on the client node;
• lock state determination means for determining a present lock state of the data object;
• lock comparator means for determining whether the present lock state is compatible with the lock request;
• lock request generator means, responsive to a determination by the lock comparator means that the present lock state is incompatible with the requested lock, for sending a message on the network, from the client node to another node which has the lock, requesting that the another node give up the incompatible lock and acknowledge when it has done so;
• and,
• update means, for analyzing responses from the network and, responsive to acknowledgement from the another node, for updating the lock state to indicate that the node has the lock on the specified data object.

In a multicomputer system wherein multiple data processing nodes, each under control of its own operating system, are joined by a communications network and wherein at least one server controls access to a shared resource, a locking method comprising the steps of:

• recording, at the server, lock control status of data objects of the shared data resource;
• requesting, by a first node, that lock control in a specified mode be granted on a particular data object;
• responsive to a requested lock on a data object, reporting by the server, a list of nodes having conflicting lock status of the data object;
• responsive to receipt of the list, requesting by the first node that other nodes on the list give up of conflicting control of the data object;
• and,
• responsive to acknowledgement from the other nodes that conflicting control has been given up, request by the first node that the server update the lock control status to reflect that the first node has lock control in the particular mode for the data object.

A method for determining granularity of a lock request on a data object shared by two or more processes executing on a least one processor, comprising the steps of:

• identifying the particular data object on which a lock is needed;
• (a) determining if the data object is a directory type file;
• (b) determining if the data object is smaller than a predetermined size;
• (c) determining if access is needed to a first page of the data object;
• (d) determining if the access is consecutive to a previous access;
• (e) determining if the access is to greater than a predetermined number of pages;
• and,
• (f) when it is determined that any of steps (a), (b), (c), (d) and (e) are answered in the affirmative, requesting a lock on the entire data object and otherwise requesting the lock on particular pages to be accessed.

A system for determining granularity of a lock request on an data object shared by two or more processes executing on a least one processor, comprising:

• process means for identifying a particular data object on which a lock is needed;
• granularity means, coupled to the means for identifying, for determining a granularity of a lock to be requested;
• the granularity means determining the granularity responsive to whether the data object is a directory type file;
• size of the data object;
• whether access is needed to a first page of the data object;
• whether access is consecutive to a previous access;
• and whether access is to greater than a predetermined number of pages;
• and,
• lock request means, coupled to the granularity means, for requesting a lock on pages of the data object specified by the granularity means.

A system for determining granularity of a lock request on a data object shared by two or more processes executing on at least one processor, comprising:

• process means for identifying a particular data object on which a lock is needed;
• granularity means, coupled to the process means for determining a granularity of a lock to be requested;
• the granularity means determining the granularity responsive to a set of heuristic rules applied to at least one attribute of the data object and at least one attribute of a requested access;
• lock request means, coupled to the granularity means, for requesting a lock on pages of the data object specified by the granularity means.

In a multicomputer system wherein multiple data processing nodes, each under control of its own operating system, are joined by a communications network and wherein at least one server controls access to a shared resource, a locking method comprising the steps of:

• recording, at the server, lock control status of data objects of the shared data resource;
• identifying a data object on which a lock is required;
• determining a portion of the data object on which to request the lock;
• requesting, by a first node, that lock control in a specified mode be granted on the portion of a data object;
• responsive to a requested lock on a data object, reporting by the server, a list of nodes having conflicting lock status of the data object;
• responsive to receipt of the list, requesting by the first node that other nodes on the list give up conflicting control of the data object;
• and,
• responsive to acknowledgement from the other nodes that conflicting control has been given up, request by the first node that the server update the lock control status to reflect that the first node has lock control in the particular mode for the data object.