Patent Application
20110196837
The present invention relates to information systems, and more particularly to information systems comprising a plurality of data sources, database instances and data consumers.
In large multi-user information systems data is typically generated and accessed by a plurality of entities in a plurality of places. In order to facilitate efficient utilization of information, data is generally distributed to several database instances located in various places in the network. The requirement is, however, that the information needs to be available to processes and users applying it quickly and efficiently. In addition, it is important that the information is configured simply such that cost-effective implementations are possible.
Conventional systems apply many methods to ensure availability of information. One option is to maintain multiple copies of data, called replicas, stored in multiple storage instances, and the access to a replica is made uniform with access to a single, non-replicated entity. In high-availability clusters, for each piece of information there typically exists one or several replication masters. A replication master is in charge of consistence of data between separate database instances. In case of a single master, all the updates from information sources go through the single master and it controls the replication of the data to other instances. If several masters are involved (so called multimaster replication) a strict mechanism is needed to solve the conflicts between the simultaneous data updates. Different approaches exist for the conflict resolution: transaction timestamps, hierarchy of the origin nodes, etc.
These configurations typically apply pessimistic replication where replication mechanisms make sure that all database instances with the same data are updated in a proper way. Pessimistic algorithms synchronously coordinate replicas during accesses and block other users during an update. These algorithms fulfil strictest single-copy requirements, often referred as ACID properties (atomicity, consistency, isolation, durability). However, fulfilling all the replication requirements ends in complex replication mechanisms. This applies especially to systems with numerous database instances, several replication masters and unreliable connections between the instances.
In optimistic replication data may be accessed without a priori synchronization. Replicas are allowed to diverge and are guaranteed to converge when the system is idle. Optimistic replication faces the challenges of controlling diverging replicas and conflicts between concurrent operations. It has thus been considered applicable only for systems that can tolerate occasional conflicts and inconsistent data (“Optimistic Replication”, YASUSHI SAITO, Hewlett-Packard Laboratories, Palo Alto, Calif., USA and MARC SHAPIRO, Microsoft Research Ltd., Cambridge, UK).
In telecommunications, subscriber data is typically distributed to at least two database instances. Dynamic subscriber data is stored in one database that changes according to the mobility of the user. Static subscriber data is stored in a fixed database, which also maintains a pointer to the present database for dynamic subscriber data. However, in new advanced system, there are several aspects that drive towards other type of subscriber data configurations.
The demand to have several geographically distributed database instances is motivated at least partly by optimization of the transmission costs, partly by resilience requirements.
In a communications network there can be several nodes, which generate status information. In cellular networks for example mobile phone and base stations are sources of status information. Also many telecommunications services, as such, generate large volumes of status information, which needs to be distributed over the network. For example, mobile subscribers generate location updates when they move in a cellular network from the area of one base station to the area of another base station, or transmit updates of presence information to a presence server. The lifetime of such status information in communications systems is quite short. All this results in a large number of status information, the transmission of which needs to be carefully optimized.
In wide regional and nationwide networks it is possible that due to breaks in the transmission services, the network will be split into two or several isolated sub-networks. It would, however, be important that even in such an abnormal situation the isolated parts of the network are able to provide services in a proper way.
An object of the present invention is thus to provide a solution for implementing the method so as to meet at least some of the above demands and thus facilitate simpler and more cost-effective provision of dynamic data in an information system. The objects of the invention are achieved by an information system, an apparatus, a method and a computer program product, which are characterized by what is stated in the independent claims. The preferred embodiments of the invention are disclosed in the dependent claims.
The invention is based on the idea of using a less stringent process for data sourcing, and in the consuming end applying a validity check to recover from the situations where the lightened process fails. Contrary to the conventional understanding, there are numerous applications, especially in field of communications systems, where the increase of processes and traffic incurred by the validity checks is generously compensated by the reduction of traffic and processes of transactional database replication. This and further advantages of the invention are discussed in more detail with the following embodiments of the invention.
In the following the invention will be described in greater detail by means of preferred embodiments with reference to the attached drawings, in which
It is appreciated that the following embodiments are exemplary. Furthermore, although the specification may in various places refer to “an”, “one”, or “some” embodiment(s), reference is not necessarily made to the same embodiment(s), or the feature in question does not only apply to a single embodiment. Single features of different embodiments may be combined to provide further embodiments.
The present invention relates to an information system where data provided from a plurality of sources is maintained in a plurality of database instances accessed by a plurality of consumers. A variety of apparatus and system configurations applying a variety of communication technologies may be used separately or in combinations to implement the embodiments of the invention. Information systems and technologies evolve continuously, and embodiments of the invention may require a number of modifications that are basically obvious for a person skilled in the art. Therefore all words and expressions of this specification should be interpreted broadly, as they are intended merely to illustrate, not to restrict, the embodiments.
Event source (ES) 10 illustrates an element that provides event data items (ek) for utilization in the processes of the information system. An event ek refers here to an occurrence of significance to a task of an event consumer (EC) 12. An event data item ek is preferably implemented as a block of information that is coded in electronic form to allow computer devices and computer software of the information system to convert, store, protect process, transmit, and securely retrieve it in the processes of the information system. The information system is able to automatically detect and process the event data item ek that typically comprises one or more separable information elements iek, one of which carries the actual data on the occurrence. The task is associated with an entity recognized in the information system and therefore the event data item typically comprises an information element identifying the entity associated to the task. For example, if the entity is a user of the information system, the event data item typically comprises an information element indicating the identity of the user. In some cases distributed data items may relate to several tasks, and the event data item typically comprises and information element identifying the task the occurrence is significant to.
Depending on the implementation, indications by the information elements may be explicit or implicit. For example, event sources may be configured to provide data elements only on occurrences of one entity (e.g. their own) and therefore the origin of the data element may be determined from the source address of the message comprising the data item, and a separate information element to identify the entity is not needed. On the other hand, if the data items are only applied for one task, for example location information of the user, the task itself does not need to be separately indicated with an information element. The format of the event data items provided by different event sources is preferably the same such that further operations for enabling the task to access the information in the event data items are not needed. However, any intermediate element of the information system between the event source and the event consumer may be adapted to receive data items in various formats of various event sources and process them into various formats applied by various event consumers.
The event data items provided by the one or more event sources are delivered to a database (DB) 14. The database represents a systematically arranged collection of data, structured so that it can be automatically retrieved or manipulated. Input of event data items is preferably separate from output of event data so that transactions with event sources may be performed independently, without necessarily regarding the operations in the event consumer side. The event source may transmit information on its own initiative and/or independent from any other event source. Alternatively, the database may prompt or query event data items from event sources. In many embodiments one occurrence is notified by one event source only, so basically there is no evident need to detect and settle conflicts between event data items on same occurrence, the event data items received from two or more independent event sources. However, such configurations are not excluded from the scope, so when necessary, any conventional or new mechanism, known to a person skilled in the art, e.g. from the field of optimistic replication, may be used for the purpose.
Event consumer (EC) 12 represents here a logical element that applies event information. EC is operatively connected to the database and typically comprises a task, for example, an application, a service, a procedure, a process, a function, or the like that inputs event data and provides an output with the content and in the format of the task at hand. DB may send event data (Ek) to EC of its own initiative, or EC may provide to DB a query q(Ek), to which DB responds with event data (Ek) derived on the basis of one or more stored event data items (ek). DB may, for example, forward the event data items (ek) to EC in the format it receives them or may process them to another format (ek′) before providing them to EC.
In case the information system comprises only one database, all data items are delivered to it and inconsistencies in the event data stored in the database result from unsuccessful communications between the event source and the data base. If the amount of event sources increases, there are more probable links to fail, and the consistency of event data increases. Moreover, if there is more than one database in the system, the situation gets more complicated. The task maintained in EC typically requires that database actions are processed reliably and EC can apply any of the available databases such that the event data received from them is ubiquitously consistent. In order to achieve this, one database is conventionally elected as a primary replica that is responsible for handling all data items from ESs. After an update in the primary, it synchronously inputs the update to other secondary replicas. Access to any replica is blocked unless it is provably up to date. With a limited number of replicas such single-copy acknowledged mechanism works well. However, this approach does not allow building large systems with frequent updates, because its throughput and availability significantly suffer as the number of event sources and database instances increases.
The invented solution applies stochastic database replication where data items are provided from a plurality of event sources to a plurality of database instances, and use of present event data from database instances is allowed without mandatory requirement of its consistency. The event consumers, on the other hand, are configured to determine the validity of the received (applied or applicable) event data, and in case of invalid data to initiate another query for the event data. The solution assumes that event data in database instances replicate as stochastic processes that vary statistically according to transmission probabilities in links between the originating event sources and the receiving database instances. As a result, the event data in one database instance may deviate from event data in another database instance and from the actual situation; it is accepted that at least some event data at least at some time is incorrect. The event consumers are, however, able to detect invalid event data and trigger an appropriate procedure to retrieve valid event data from another place, for example from another database instance or from the original event source.
Implementation of the triggered procedure naturally causes additional operations and communications in the information system. However, the probability for a transmission of event data to succeed in various links may be quite easily estimated and improved. Typically throughput probabilities in communication links of information systems are relatively good, so the likelihood of new queries being initiated is actually very small. The increase of traffic from new queries is thus well compensated by the significant reduction in the number of exchanged acknowledgement messages within the information system. Due to the reduced amount of traffic and control operations associated to single event data deliveries, the distribution and application of event data items is faster and simpler. In addition, the information system according to the invention scales easily to larger configurations and very dynamic event data. This enables embodiments even in processes and services of mobility management and call control in wide area communications systems where, due to scalability and congestion problems, only conventional single-copy acknowledged database arrangements have been applicable so far.
The invention may be applied to any information system where event data is accessible from a plurality of parallel database instances. In the following, an embodiment of the invention applying a communications system of
The embodied communications system provides wide area access to a plurality of users that have a user terminal and a subscription. With the combination of the user terminal and the subscriber identity the users have access to the services of the communications system. The network infrastructure of
The simplified radio access network configuration shown in
The radio access network may comprise a separate controlling network element, which manages the use and integrity of the radio resources of a group of one or more base stations. However, the radio network control functions may also be implemented in individual base stations. The present embodiment applies the latter configuration. The cellular system of
Connections between network elements in the radio access network and in the core network may be circuit switched or packet switched. A circuit switched type of connection is a connection for which dedicated network resources are allocated upon connection establishment and released upon connection release. A packet switched type of connection transports user information in packets so that each packet can be routed independently of a previous one. Transmissions over packet switched connections may be acknowledged or unacknowledged. An acknowledged transmission is repeated automatically until the entity in the addressed destination confirms that it has received it. Acknowledged transmission enables ensuring that planned information is duly exchanged and actions based on the information may be reliably performed. On the other hand, acknowledgement messages and repeated retransmissions increase traffic and may cause considerable delays to the overall communications. Unacknowledged transmissions, where the fact that an acknowledgement is not received from the destination, does not automatically cause retransmission, are well suited for dynamic multi-user packet switched communications.
The core network of the communications system comprises several subscriber registers for static and dynamic subscriber data. The static data is created when subscribers are provisioned and changes very rarely. Static data can thus be replicated between subscriber registers using conventional replication mechanisms. Dynamic subscriber data comprises subscriber data that may vary frequently according to events occurring to the subscriber. An event may result from an activity taken by the subscriber or from an activity or occurrence caused or happened to the subscriber.
In the present embodiment, referring to
It should be noted that the present embodiment deals with dynamic data provided by the user terminals and/or base stations of the radio access network. Conventionally communications networks have comprised a home location register for essentially static subscriber information and a visitor location register where all subscriber parameters for call set-up are stored as long as the mobile subscriber is in a location area controlled by this register. The home location register has, in addition to the static subscriber information, maintained a dynamic element that points to the present visitor location register where the user terminal presently resides. This division has been required to minimize the replication of dynamic data and thus to cope with the update resources available in the network. Due to the lightened update procedure, the invention also allows new distributed configurations for static and/or dynamic data, and thereby a while range of new query procedures optimized in view of load, geographic distance, or other considerations. Some new subscriber register configurations are discussed in more detail in this description.
From the plurality of base stations BS1, BS2, . . . , BSm, a group of base stations BS1, BS2, BS3 are in area where a subscriber A with user terminal UT1 presently resides and another group of base stations BS4, BS5, . . . , BSm in an area where another subscriber B with user terminal UT2 presently resides. When a base station BS4 receives a location update message from a user terminal UT2, it sends the message <lu1> as an event data item to all subscriber registers SR1, SR2, SR3, . . . , SRn using a defined packet switched protocol that does not apply acknowledged transmissions. According to the earlier definition, this means that when a base station BS4 sends information on a location data update of UT2 of subscriber B to a database instance SR1, retransmission of the location update to SR1 is not dependent on acknowledgement from SR1, i.e. BS3 does not necessarily need to retransmit the location update to SR1, even if no acknowledgement is received from SR1.
BS3 may be configured to retransmit the location update at least once to SR2, notwithstanding whether it receives the acknowledgement or not. Such arrangement is actually preferred because it may significantly increase the probability of the location update messages getting though without, however, excessively increasing the traffic in the communications system. BS3 may also be configured to require, for some other purpose, acknowledgements from one or more databases or the databases may be configured to acknowledge the transmissions without the sender actually requiring it. In any case, in BS3 the decision whether to retransmit the event data item to a database instance is not mandatory dependent on reception of an acknowledgement from the database instance.
When subscriber A that resides presently in the cell of BS3 wishes to call subscriber B, it initiates a call setup procedure in UT1. The call setup procedure is a task that applies event data of mobility management. In order to access appropriate call set-up parameters, a relevant subscriber register needs to be accessed. In the present embodiment, BS3 of subscriber A may select the database instance to which it sends the query requesting information on the present cell of subscriber B freely, i.e. randomly or according to any defined unique or ubiquitous selection criterion or criteria. In the present application, let us assume that BS3 is configured to send the query <q1> to SR2 that is geographically closest to it. SR2 responds with a message <r1> that includes UT1 as the present cell of BS4.
Having this information, BS3 sends the call setup signaling message to BS4 and if BS4 responds accordingly, the call setup continues normally from there on. The probability that the location information in a subscriber register is correct depends by far on the transmission mechanism used to deliver the location update message. The present embodiment applies IP multicast technology that uses user datagram protocol (UDP) as an underneath protocol, and does not provide mechanisms to ensure successful data transmission. Other corresponding packet-switched protocols that fulfil the claimed features may be applied without deviating from the scope of protection. For example the location update message may be multiplied and sent as a unicast message, one message being transmitted per one database instance.
In UDP, reliability is based on the properties of the protocol stack's lower layer, e.g. Ethernet. In IP networks the measure of the packet delivery probability is measured by packet loss, i.e. the probability that packets are not appropriately delivered. For example, in well-designed and maintained IP networks a typical figure of 0.1% for packet loss can be easily achieved:
Pr(“IP packet lost”)=0.1%.
If it is assumed that base stations repeat the transmission of location update messages to subscriber registers two times to ensure the update delivery, and further that the packet loss probability of two successively repeated messages is independent and identically distributed the probability of the loss of location updates can be computed from:
Pr(“location update message lost”)=Pr(“IP packet lost”)2=(1*106)−1. In such circumstances every millionth location information reply from any subscriber register would be incorrect.
In order to control the number of unsuccessful call attempts due to incorrect data, the event consumer, here the querying base station, is configured to detect whether the information queried from the database instance is valid or not. For a person skilled in the art it is clear that such checks may be implemented in various ways. In this embodiment, the base station BS3 sends the call setup signalling message to BS4 and waits for a response for a defined period. In case BS4 responds positively during the period, the call setup procedure may continue normally. If the response is negative, for example indicates that UT2 does not reside in cell of BS4, or no response is received within the defined period, BS3 determines the received location information being invalid and initiates a new query to another destination.
Let us assume that before the call setup, subscriber has already moved to cell of BS5, but the location update message by BS5 has, for some reason, not reached SR2. Accordingly, the query proceeds as described earlier, but when BS3 sends the call setup message to BS4, BS4 does not find UT2 in its cell and acknowledges the message negatively. This causes a new query, the format and destination of which may vary considerably within the scope of protection.
Several other approaches for performing the new query are possible within the scope of protection. For example, the base stations may have some prior knowledge on probable locations of subscribers and use that to restrict the number of messages transmitted in the second query. For example, the subscriber identity may be associated to a particular domain in the network, and the querying base station may restrict the second query to those base stations, or subscriber registers closest to those base stations, and broaden the query to cover all base stations only on the third attempt.
The validity detection described in
In the embodiments of
It is also possible to integrate the static and dynamic event data to same database instances, but perform the updates according to the event data type in question. For example, in the embodiment of
The stochastic database replication embodiments of the invention provide several advantages. In a network there can be several geographically distributed subscriber registers, which means that transmission costs from database queries can be optimized. Several essentially parallel subscriber registers provide good resilience in case one of the registers is not able to provide service. In addition, if a large network gets split into several sub-networks due to breaks in the transmission services, the isolated parts of the network can provide services in a proper way. After the sub-networks are again reconnected, the subscriber registers will automatically converge towards the state of having the correct information about the network as a whole. Some additional synchronization mechanism can be used to quickly replicate the location data between the subscriber registers. It is further noted that database instances with less stringent update procedures may be implemented with relatively modest hardware and software implementations, no carrier grade and high availability platforms are necessarily needed. The simpler configuration is highly cost-efficient: less expensive and easier to maintain.
When discussing the pros and cons of the invention, it was noted that the operations to compensate the relaxed update procedures were dependent on the probability of the unacknowledged transmissions to succeed. It is clear that the invention may be further enhanced by actions that increase the likelihood of successful transmissions from event sources to databases. In the embodiments of
The solution further allows the time period between two successively repeated update messages to be adjusted optimally according to the implementation. It may, for example, be defined that the event source sends first two transmissions with a short interval in between and then repeats transmissions with a considerably longer interval. For example, in the case of location updates, the base stations may be configured to transmit the location update messages to all database instances two times with some seconds interval and then retransmit the message with an increasing interval, (if acknowledged sending used). The time period between two successively repeated update messages is preferably longer than the convergence time of the network, for example with OSPF typically from seconds to tens of seconds. Otherwise it would be likely that during a service break both the first transmission and the retransmission would be lost. If the period between the retransmissions is long enough, it may be assumed that the network recovers from the fault during the period and at least one of the messages is delivered to the database instance.
In the embodiments of
In conventional communications systems parallel subscriber database instances are not applied and dynamic information is available in one entity, the visitor location register, only. The embodied mechanism proposes an improved mechanism that enables distribution of subscriber data to several parallel database instances, which may be selected as the accessed entity according to, for example, their associated transmission distance, quality of the transmission link, or present availability.
The method may be further improved by a step where the base station at least once retransmits (step 44) the information element, notwithstanding whether an acknowledgement is received from any database or not. This increases the probability of the information in the databases being correct and thus reduces the associated need for new queries in fault situations.
The method may be also further improved by a procedure where the base station checks (step 45) whether at least one response from some database instance is received. If at least one response is received, it considers the throughput adequate and continues to step 41 to standby for further information. If not, the base station adjusts (step 46) a period T to a next retransmission and waits (step 47) this period T before proceeding to step 44 to retransmit the data item. The period T may be constant or it may increase or decrease between consecutive retransmissions, according to the application. This allows adjusting the amount of retransmissions to match the other configuration of the applied system.
It is noted that the steps/points, signaling messages and related functions described in any of the flow charts are in no absolute chronological order, and some of the steps/points may be performed simultaneously or in an order differing from the given one. Other functions can also be executed between the steps/points or within the steps/points and other signaling messages sent between the illustrated messages. Some of the steps/points or part of the steps/points can also be left out or replaced by a corresponding step/point or part of the step/point.
Whenever such request is detected (step 506), the base station selects a database DB(Ek) from which event data for the mobility management procedure is queried. Due to the invention, this decision may be made freely according to, for example, transmission costs, quality of transmission links or present availability of the parallel databases. The base station performs a query (step 510) with the selected database DB(Ek) and checks (step 512) the validity of the information in the response. This may be implemented by, for example, utilizing the event data in the process P and, on the basis of success or failure of the outcome, determine the validity of the used data. If (step 514) the result is valid, for example if the call setup message is duly acknowledged, the procedure continues to step 504 to standby for further information needs of the process P. If (step 514) the result is invalid, for example the call setup messages goes unanswered, and there are still unqueried databases (step 516), a new database is selected (step 518) and the procedure continues to step 510 to implement the query in the new selected database. Otherwise (step 516) the call setup fails (step 520) and the procedure moves to step 504 to standby for further information needs of the process P.
The block diagram in
The apparatus comprises an interface unit 61 with at least one input unit for inputting data to the internal processes of the apparatus and at least one output unit for outputting data from the internal processes of the apparatus. In a user terminal apparatus, the interface unit typically comprises a user interface with a keypad, a touch screen, a microphone, and equals for inputting data and a screen, a touch screen, a loudspeaker, and equals for outputting data. In a network element apparatus the interface unit typically comprises plug-in units acting as a gateway for information delivered to its external connection points and for information fed to the lines connected to its external connection points.
The interface unit 61 is electrically connected to a processor unit 62 for performing systematic execution of operations upon data. The processor unit 62 is a central element that essentially comprises an arithmetic logic unit, a number of special registers and control circuits. Memory unit 63, data medium where computer-readable data or programs, or user data can be stored, is connected to the processor unit 62. The memory unit 23 typically comprises volatile or non-volatile memory, for example EEPROM, ROM, PROM, RAM, DRAM, SRAM, firmware, programmable logic, etc.
A user terminal and a base station apparatus comprise a radio transceiver unit 64, which includes a transmitter 65 and a receiver 66, and is also electrically connected to the processor unit 62. The transmitter 65 receives a bitstream from the processor unit 62, and converts it to a radio signal for transmission by the antenna 67. Correspondingly, the radio signals received by the antenna 67 are led to the receiver 66, which converts the radio signal into a bitstream that is forwarded for further processor to the processor unit 62. The functions implemented by the processor unit 622 in transmission typically comprise encoding, reordering, interleaving, scrambling, channel multiplexing, and burst building.
The reference hardware configuration of some other network element apparatus corresponds with the configuration of the base station, but the typically does not comprise a radio transceiver unit.
The processor unit 62, memory unit 63, interface unit 62 and radio transceiver unit 64 are electrically interconnected to provide functional entities for performing systematic execution of operations on the received and/or stored data according to the predefined, essentially programmed processes of the apparatus. In solutions according to the invention, the functional entities of an event source apparatus comprise at least a database record for storing information on a group of database instances associated to a communication unit comprising a communication device and a module for identifying a subscriber using the communication device, an event manager for generating an event data item associated with the communication unit and a database interface a database interface for transmitting the event data item to the group of database instances in the database record. In solutions according to the invention, the functional entities of an event consumer apparatus comprise at least a database record for storing information on a group of database instances storing event data applied by the process, an event data provider for determining, from the group of database instances, a database instance and querying event data for the process from the database instance, and a validity detector for detecting whether information queried from the database instance is valid or not. These operations are described in more detail with
The invention may also be embodied in a computer program product, readable by a computer and encoding a computer program of instructions for executing a computer process for controlling functions in an apparatus of an information system.
It will be obvious to a person skilled in the art that, as the technology advances, the inventive concept can be implemented in various ways. The invention and its embodiments are not limited to the examples described above but may vary within the scope of the claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 08165999.7 | Oct 2008 | EP | regional |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/FI2009/050780 | 9/30/2009 | WO | 00 | 4/6/2011 |
