This disclosure relates generally to the field of database management, and more particularly, to apparatus and methods for concurrent session management.
In a distributed computing system having a content management system including a data repository, users (through client devices) establish sessions with the content management system in order to access the data repository for read and/or write operations. However, in systems where concurrent access to data resources is allowed, data errors can occur unless access by concurrent users to shared resources is controlled.
One control technique employs a reference counter which is updated every time an object type referenced or de-referenced. However, in order to update an object is exclusively locked so that no other reads or writes can be committed to avoid concurrency errors. However, users must wait until the object is unlocked before their session can proceed.
Therefore, it would be desirable to provide a mechanism to allow updates to objects in a more efficient manner.
This disclosure describes methods and systems for concurrent collection and/or generation of shared information.
In a distributed computing system, concurrent access to a shared resource is controlled by a status indicator. For example, in a single reader system, a flag is set to TRUE to indicate concurrent access by the single reader. When the flag is FALSE, then there are no concurrent sessions. In a multiple reader system, a counter is incremented each time a new concurrent session is initiated and decremented each time a session is terminated and releases the shared resource. When the counter has a value of zero, there are no concurrent sessions.
Typically, a shared resource has a defined size, for example, as a linked list. In one embodiment, after a user writes new data to the shared resource (data push), the status indicator is checked to see if there are any concurrent sessions active. If so, then older data is not removed from the shared resource (data pop) to make room for the new data until such time as there are no more concurrent sessions for that resource. If there are no active concurrent sessions, then older data is removed and the resource returns to its fixed size.
1. Operating Environment
Communication network 5 may include many interconnected computer systems and communication links. Communication links 6, 7, 8, 9 may be hardwire links, optical links, satellite or other wireless communications links, wave propagation links, or any other mechanisms for communication of information, and may be the same or different for the various hardware components that are coupled to the network 5. Various communication protocols may be used to facilitate communication between the various systems shown in
Distributed computer network 10 is merely one illustrative embodiment and is not intended to limit the scope of the disclosure. One of ordinary skill in the art would recognize many other variations, modifications and alternatives in how to effectively arrange such components and systems.
In a typical client/server arrangement as shown in
Server system 20 receives requests for information from one or more of client systems 11, 12, 13 and performs processing in order to satisfy the requests, then forwards the resultant information back to the requesting client system. The processing required to satisfy the request may be performed by server system 20 or may alternatively be delegated to or obtained with the help of other servers connected to communication network 5 or to server system 20.
Client systems 11, 12, 13 enable users to access and query information stored or managed by server system 20. In one embodiment, a web browser application executing on a client system enables users to select, access, retrieve and/or query information stored or managed by server system 20. Examples of web browser programs include the Internet Explorer browser by Microsoft Corporation, the Firefox browser by Mozilla Foundation, and others.
Mass storage devices 57 may include disk drives, floppy disks, magnetic disks, optical disks, magneto-optical disks, fixed disks, hard disks, CD-ROMs, recordable CDs, DVDs, recordable DVDs (e.g., DVD-R, DVD+R, DVD-RW, DVD+RW, HD-DVD, or Blu-ray Disc), flash and other nonvolatile solid-state storage (e.g., USB flash drive), battery-backed-up volatile memory, tape storage, reader, and other similar media, and combinations of these.
The subject matter described herein may be computer-implemented in a suitable system by providing computer-executable instructions stored on a non-transitory computer-readable medium. A computer-readable medium may include any medium that participates in providing instructions to one or more processors for execution. Such a medium may take many forms including, but not limited to, nonvolatile, volatile, and transmission media. Nonvolatile media includes, for example, flash memory, or optical or magnetic disks. Volatile media includes static or dynamic memory, such as cache memory or RAM. Transmission media includes coaxial cables, copper wire, fiber optic lines, and wires arranged in a bus. Transmission media can also take the form of electromagnetic energy, radio frequency signals, acoustic or light waves, such as those generated during radio wave and infrared data communications.
For example, a binary machine-executable version of software instructions for providing the features described herein may be stored or reside in RAM or cache memory, or on mass storage device 57. The source code of the software may also be stored or reside on mass storage device 57 (e.g., hard disk, magnetic disk, tape or CD-ROM). As a further example, software code may be transmitted via wires or radio waves, or downloaded through a network such as the Internet.
A system block diagram of the computer system/workstation 50 is shown in
The system bus 70 of computer system 50 are exemplary of an interconnection scheme serving to link the subsystems. For example, speaker 67 could be connected to the other subsystems through a port or have an internal direct connection to central processor 61. The processor 61 may include multiple processors or a multi-core processor, which may permit parallel processing of information. Computer system 50 is only one example of a suitable computer system. Other configurations of components and/or subsystems suitable for use with the methods described herein will be readily apparent to one of ordinary skill in the art.
Computer software products may be written in any of various suitable programming languages, such as C, C++, C#, Pascal, Fortran, Perl, Matlab (from MathWorks), SAS, SPSS, JavaScript, AJAX, Java, SQL, and XQuery. A computer software product may be an independent application with data input and data display modules. Alternatively, computer software products may be classes that are instantiated as distributed objects, or component software such as Java Beans or Enterprise Java Beans (both by Oracle Corporation). In one embodiment, the subject matter described herein is embodied as a computer program product which stores instructions, such as computer code, that when executed by a computer cause the computer to perform the processes and/or techniques described below.
The computer system 50 runs an operating system (OS) to manage hardware resources and provide common application services for application software. As an example, the computer 50 may run a Microsoft Windows® operating system (e.g., Windows NT, Windows XP, Windows Vista, Windows 7, Windows CE, Windows Mobile), or Linux, UNIX, Sun OS, Solaris, Mac OS X, Alpha OS, AIX, IRIX32, or IRIX64. Other operating systems may also be used.
2. Content Management Platform
The client applications 100 originate from client devices, such as devices 101, 102, 103, which communicate through a network 104 with an application server 121 in order to use one or more applications hosted on the application server. In one embodiment, the network 104 is an Ethernet connection using conventional TCP/IP communication methods with both client devices and server devices. However, other types of network connection technologies are well known and may also be used to full advantage, including local area network (LAN), wide area network (WAN), storage area network (SAN), the Internet, etc. Client devices may be any type of processor-based digital device, such as desktop or laptop computer, iPhone@ or other smart hand-held device, or other.
The application server 121 hosts one or more applications running on a content management platform that provide access to repository 140 for clients hosted on content server 122, which is also accessible to clients via network 104. An HTTP server 123 is also coupled to the network 104 to provide an administrative terminal for system operators to perform system maintenance, logging, access control, etc.
The repository 140 is provided for use by client applications 100, through the content server 122, which is accessible to the client devices through a network 104. The repository 140 may include a file store 141 for storing content items, a relational database 142 for storing attributes/metadata associated with the content items, a full-text index store 143 for the content items, and directory services 144.
A software platform runs on each of the various components, such as the EMC Documentum Platform software and related EMC Enterprise Content Management software, distributed by EMC Corporation, Hopkington, Mass. The Documentum Platform provides capabilities for managing enterprise content and is the foundation for related content management and case management product offerings.
For example,
DFC is the programming interface for accessing the functionality of the Documentum platform through Java and COM class libraries and underlies all Documentum platforms. DFS is an SOA development framework and application programming interface (API), based on web services, and designed to support interoperable machine-to-machine interaction over a network. DFS is now the primary interface for most developers and users of the Documentum platform, although use the terms DFC and DFS are frequently (although incorrectly) used interchangeably.
Each of the clients 101, 102, 103 also runs a content management framework, with the API provided by the Documentum Client Library (DMCL), which is a linked library inside the client application (a DLL in Windows; a shared library in Unix). The DMCL takes documented API calls made by the client application and maps them to Server RPC functions that can be executed by the Content Server. Once the API call has arrived inside the Content Server, processing is no longer affected by the DMCL environment. In fact the DMCL thread that made the call will simply be sleeping awaiting the results of the processing from the server.
The Documentum platform also includes an extensible Business Objects Framework (BOF) which is built into DFC. BOF provides a useful framework and methodology for developing reusable business logic components called Business Objects that can be plugged into middle tier or client applications. More information about the Documentum Platform and related content management and case management product offerings can be found on the EMC website located at URL www.emc.com.
3. Session Management
A session is established between a client and the data repository when the client software issues an API call to connect to the content server. The DMCL on the client device creates a thread (Windows) or a process (Unix) representing the session in client memory, and likewise, the content server creates a thread/process representing the session on the server. A session created on the client device typically does not have a timeout associated with it, but will remain in client memory until disconnected. A session on the server, however, will usually have a timeout set in the configuration file.
Although concurrent sessions are generally permitted in most systems, there are typically limits set in configuration files for the number of sessions that may be handled at one time, and these limits are usually different for the client (fewer sessions) and the server (more sessions). In the client, the DMCL layer defines a key [max_session_count] in the configuration file [dmcl.ini] that controls the number of sessions which may be active on the client at one time. The server also includes a defined key [concurrent_sessions] in its configuration file [server.ini] that controls the number of sessions that may be active on the server.
4. Object Types and Type Cache
Many databases and content management systems, such as the EMC Documentum platform, are object-oriented systems wherein each individual item is defined as an object. An object type represents a class of objects, and the definition of an object type is a set of properties whose values describe individual objects of that type. Thus, object types are like templates, and the content server uses the type definition as a template to create an object and then set the properties of the object to values specific to that object instance.
While objects are actually stored in the data repository, the content server allocates server resources, including pointers and caches, to facilitate session management of objects. For example, an Object Cache may be defined to deal with objects based on a set of policies, including fetching, finding, naming, mapping, detaching, and cloning objects. A Global Cache may be defined as the object cache of the root session, and is used as the basis for spawning new user sessions. A pointer may be defined to point to a specified object in the global cache. A Session Cache may be defined as the object cache for the session. When a new session is created, the session cache is populated by the global cache. One or more Change Checker routines may be defined to periodically wake up and refresh the global cache and other caches as required to ensure that entries are up to date.
However, considerable overhead may be consumed with cache operations. For example, a subtype usually repeats the definition of its inherited attributes from its parent type. Another example from the Windows platform is that type definitions are copied from the main thread to the session thread even though they are identical. Also, in Unix platforms, object types are constantly updated which causes the session to create its own copy, known as the “copy-on-write” (COW) mechanism.
Current implementations of the Type Cache are different on thread-based Windows platforms (see
If there is a parent/child cache set up in a Windows thread, the global (parent) cache actually issues a call back into the session (child) cache and forces it to load the type object, and then create a copy allocated from the global heap to store in its own object store. This is done to avoid lengthy locking times on the global cache which can block other sessions from accessing the cache.
As is evident from
In the new implementation, the session cache 420 issues an RPC to generate one or more pointers 422 to point directly to the corresponding type objects in global cache when the session places those type objects in use. After each RPC, the pointers 422 in session cache 420 will be set back to NULL (or destroyed if only a session copy) if there is no open transaction involving those type objects.
5. Object Reference Count
In current implementations of the type class, the variable reference count [ref_count] is used to keep track of how many times the type object is referenced, including the following:
(i) sub-type reference count—the number of sub-types pointing to the type object;
(ii) cache reference count—anytime a type object is added to the cache or removed from the cache, the reference count is increased by one or decreased by one, respectively; and
(iii) object reference count—the number of persistent objects of the type pointing to the type object.
In particular, the cache reference count and the object reference count cause the type object to be updated frequently, and such operations unnecessarily consume memory creating a possible negative impact on performance. For example, the reference count may trigger the copy-on-write mechanism on the Unix platform.
In an effort to reduce the memory overhead requirements for type objects and type cache in sessions, in one embodiment, the existing variable reference count is changed from an infer-session counter to an intra-session counter by referring only to subtypes. During type creation, the variable [subtype_ref_count] will be incremented as needed.
Thus, the reference count will not be increased when a type object is inserted into session cache or global cache, but only when a subtype is used. Likewise, there is no need to decrease the reference count when a type object is detached, but only when a subtype is detached. However, a type object will still be marked as invalid when it is detached from the cache.
In one thread embodiment for a Windows environment shown in
In step 502, a session is active. In step 504, a type object is requested by the session. In step 506, the session looks for the type object in its session cache. Step 508 poses the question “is the type object in the session cache?” If so, then proceed to step 510, which poses the question “is the type object in the session cache valid?” If so, then in step 512, the session uses the type object is the session cache. If the type object is not valid in step 510, then the type object is removed from the session cache in step 514, and the session reference count in the global context map is decremented in step 516.
If the type object is not in the session cache in step 508, then in step 518, the session looks it up in the global cache. If the type object is in the global cache (step 520), then a pointer is placed in the session cache that points directly to the corresponding type object in the global cache in step 522. Further, the session reference count in the global context map is increased by one in step 524. If the type object is not found in the global cache in step 520, then the global cache fetches the type object from the database in step 526, then proceeds to step 522.
At the end of each session RPC, all entries in the session cache will be removed (except for local copies). Further, the session reference count for each type object in the global context map will be decreased by one if there are no open transactions.
If the subtype reference count and the session reference count are both equal to 0 and the type object is marked as invalid, then it will be removed from global cache and destroyed. When a type object is destroyed, the reference count to its supertype subtype will be decreased by one.
6. Updating Object Types
Under some conditions, the object type may need to be updated in the global cache. However, since the global cache is shared among threads/processes, any concurrency must be managed in order to avoid read or write errors. A prior solution was to simply lock the object in the global cache, then make and save changes to the object, and then unlock the changed object. Of course, no one has access to the object when it is locked except the user making the changes, and this can cause unacceptable delays.
For example, changes to the type definition may be triggered by an SQL statement. Referring to
This application is a continuation of, and claims a benefit of priority under 35 U.S.C. 120 of the filing date of U.S. patent application Ser. No. 13/524,487, filed Jun. 15, 2012, entitled “METHODS FOR UPDATING REFERENCE COUNT AND SHARED OBJECTS IN A CONCURRENT SYSTEM,” the entire contents of all of which are hereby expressly incorporated by reference for all purposes.
Number | Date | Country | |
---|---|---|---|
Parent | 13524487 | Jun 2012 | US |
Child | 16883681 | US |