a number of IN services under development (Freep- hone, Split Charge, and Virtual Private Network), and future services (mobile services, UPT, directory ser-.
Overview
of Databases
Requirements
for Intelligent
Networks
Willem Jonker and Lambert J.M. Nieuwenhuis KPN, PTT Research, Groningen, The Netherlands e-mail: { w.jonker 1 l.j.m.nieuwenhuis } @ research.ptt.nl
Abstract
an SCP processes a large number of service requests in parallel. Obviously, the performance of the SCP and SDP is directly related to the Quality of Service experienced by the service end-users. Consequently, the SDP can be characterised as a high capacity database which has to support simultaneous, real-time transactions. The SDP (and SCP) have to meet high reliability and availability requirements, as SDP faults may cause network wide service failure affecting all subscribers. In order to meet these requirements, the SDP implementations will be based on replication strategies to support fault tolerance. Hence, the requirements for the SDPs in IN are challenging: highly reliable database systems must provide simultaneous real-time access to a large numbers of users. These challenges make databases for telecommunications systems an interesting subject for academic and industrial research[$, 5, 6, 91. The objective of this paper is to give an overview of the required database functionality in IN networks (Section 2). We then extend the requirements analysis for future telecommunications services, e.g., mobile communication (Section 3). The requirements analysis justify future research on distribution, partitioning and replication techniques for databases in various applications areas of telecommunications (Section 4).
We discuss requirements and techniques for the Service Data Point (SDP) in Intelligent Network (IN) architectures in telecommunications systems. The SDP must provide real-time, simultaneous access to high volume databases. We give a short overview of SDP requirements derived from currently developed and future IN telecommunications services.
1
Introduction
The Intelligent Network (IN) architecture [2] has been developed to improve the slow and costly process of introducing new telecommunications services in today’s telecommunications systems requiring software updates in all switching systems of the public network. The basic idea of the IN architecture is to shift the service control out of the switches to a small number of Service Control Points (SCPs) somewhere located in the public network. In an IN architecture, switching systems provide a set of basic switching functions. New services can be introduced by adding a new service control program in the SCP rather than updating the software of the switching systems across the entire network. An example of an IN service is short number dialing in a Virtual Private Network (VPN). The IN service request of a subscriber is detected by the switch, which then routes the request, including the ‘short number’, to the SCP. For each subscriber, the SCP maintains a tabel in which the ‘short numbers’ are mapped on the ‘long numbers’. The SCP sends the ‘long numbers’ back to the switch, which proceeds with a normal call setup, using the received information. Obviously, the SCP needs a repository function to store the large amounts of data needed to provide various telecommunications services to all subscribers of the public network. The storage function of the SCP is provided by the Service Data Point (SDP), the physical entity supporting the SCP. If we assume that in future each subscriber will use at least a few IN services, an SDP has to store enormous amounts of data. In practice,
2
IN database
requirements
Most of the requirements presented here, are from an EURESCOM foresight study carried out in 1992[1]. Some observations are the results of an American study by Nicholas Roussopoulos[7], presented at a workshop in Germany in April 1994[5]. The requirements reported here are derived from a number of IN services under development (Freephone, Split Charge, and Virtual Private Network), and future services (mobile services, UPT, directory services) as well as a number of specific network management services (accounting and billing): Storage
requirements
simple and tabular
The data involved is rather representation suffices. The
160 O-8186-7099-1/95 $4.0001995
IEEE
Proceedings of the 3rd Workshop on Parallel and Distributed Real-Time Systems (WPDRTS '95) 0-8186-7099-1/95 $10.00 © 1995 IEEE
7T
several SCPs by replicating data in their corresponding SDPs. There is doubt whether pure SSP-SCP techniques will (can) be used for these services.
amount of data to be stored is expressed in storage capacity per million lines (since one SCP database per million lines seems achievable) and estimated to range from 1 to 100 Mbytes for current services.
4
requirements For current services, simple but fast retrieval is the main mode of operation. The amount of data involved is relatively small, for example for credit card calling 25 to 50 bytes per access, while for freephone numbers a service profile of about 2 Kbytes is retrieved.
Access
requirements Most IN services will use small, real-time, read-only transactions. As far as consistency is concerned, requirements cover the whole spectrum from degree 0 (e.g., for updates on certain sensor data) till degree 3 (e.g., in billing services)[4]. A variety of transaction types should be supported, including noninterruptible, prioritised and non-wait transactions.
Transactions
References
PI Final
Deliverable S23. Fact Finding Study on Requirements on Databases for Telecom Services. Technical Report , EURESCOM, 1993.
requirements For current services, there are at most 20 operations per second per million lines. This results on average in less than 60 simple transactions per second.
Performance
3
Future
database
Conclusions
Although the requirements listed above give a good impression of the kind of database functionality that is needed, they still are too vague to start implementing operational systems. What is needed is hands-on experiments in a real-world environment. The INDIA experiments[8] are a good example, and similar experiments are needed to judge the suitability of the variety of distribution, partitioning, and replication techniques proposed. We are currently starting such an experiment focusing on validated performance and reliability models, i.e., performability models[3], of systems based on the proposed replication techniques, when supporting CS-1 IN services.
PI ITU-T.
ITU-T
Recommendations
912x2.
ITU-T,
April 1992.
[31 J.F. Meyer. On Evaluating the Performability
requirements
Degradable Computer
Current support for GSM mobile communication is based on a centralised database per managed domain, i.e., when two German GSM subscribers both are in Portugal, a call between them will cause an access of databases located in Germany. For future mobile services, the estimated storage requirements are between 1 Gbyte and 1 Tbyte. Hence, alternatives based on distributed database technology must be considered. Strong requirements on concurrency control and real-time response is needed to support frequent hand-overs en location updates needed for broadband mobile services. Other future services, e.g., charging, require at least one write for each call. During rush hour, this will result in 2000 writes per second per million lines. For UPT, requirements go up to 10,000 reads and 10,000 writes per second per million lines. The main strategy to cope with the increasing demands is load distribution both in space and in time. Partioning can often successfully be used in cases where data is specific for a well-defined group of subscribers. However in some services, like numbertranslation, this is not the case. For these cases, replication techniques are investigated, e.g., Atomic Delayed Replication [8] or Relay Race [6]. In an IN context, replication can be used to share the load among
tions on Computers,
of Systems. IEEE Transac7(31):361 - 372, 1980.
PI J.N.
Gray, R.A. Lorie, G.R. Putzolu, I.L. Traiger. Granularity of Locks and Degrees of Consistency in a Shared Database. In G.M. Nijssen, editor, Modelling in Data Base Management Systems. NorthHolland, 1976.
[51 K. Klabunde, uted Databases
M. Jarke. for
Workshop on DistribPhilips Telecommunications.
Research Laboratories, Aachen, 1994.
PI M.
Cannataro, A. Limongiello, T. Mostardi, A. Pelaggi, C. Siciliano. Reliable Management of Replicated Data with Relay Race. In TINA ‘93: Proceedings of the Fourth TINA Workshop. TINA Consortium, 1993.
[71 N. Roussopoulos. Personal versity of Maryland,
Communication.
Uni-
1994.
P31K.
Klabunde R. Gallersdorfer, M. Jarke. Intelligent Networks as a Data Intensive Application (INDIA). Technical Report 94-20, Aachener Informatik - Berichte, 1994.
PI TINA-C. TINA
TINA’93: Proceedings of the Fourth workshop. TINA Consortium, 1993.
161
Proceedings of the 3rd Workshop on Parallel and Distributed Real-Time Systems (WPDRTS '95) 0-8186-7099-1/95 $10.00 © 1995 IEEE
7T
