base designersand DBAs required by traditional distributed and multi-database systems [9,10].
This paper introduces the Local Relational Model (LRM) as a data model specifically designed for P2Papplications. LRM assumes that the set of all data in
https://www.51lunwen.org a P2P network consists of local (relational) databases, each
with a set of acquaintances, which define the P2P network topology. For each acquaintance link, domainrelations define translation rules between data items, and coordination formulas define semantic dependencies
between the two databases. The main goals of the data model are to allow for inconsistent databases and tosupport semantic interoperability in the absence of a global schema [13].
1 2002 Bernstein, Giunchiglia, Kementsietsidis, Mylopoulos, Serafini, Zaihrayeu.
2 Microsoft Corporation, One Microsoft Way, Redmond WA, 98052-6399. philbe@microsoft.com
3 University of Trento, 38050 Povo, Trento, Italy. {fausto,ilya}@dit.unitn.it.
4 University of Toronto, Toronto, Canada, M5S 3H5. {jm, tasos}@cs.toronto.edu
5 ITC-IRST, 38050 Povo, Trento, Italy. serafini@itc.it
- 2 -
The main objectives of this paper are to introduce the LRM through examples and to identify a set of openresearch questions on its design and implementation. Section 2 presents a motivating scenario. Section 3
sketches a formalization of LRM. Section 4 offers a preliminary architecture for an LRM-based system andrelates it to past work, while conclusions appear in section 5.
2. A Motivating Scenario
Consider, again, the example of patient databases. Suppose that the Toronto General Hospital owns the TGHDBdatabase with schema:
Patient(TGH#,OHIP#,Name,Sex,Age,FamilyDr,PatRecord) PatientInfo(OHIP#,Record)
Treatment(TreatID,TGH#,Date,TreatDesc,PhysID) Medication(TGH#,Drug#,Dose,StartD,EndD)
Admission(AdmID,OHIP#,AdmDate,ProblemDesc,PhysID,DisDate)
The database identifies patients by their hospital ID and keeps track of admissions, patient information obtainedfrom external sources, and all treatments and medications administered by the hospital staff.
When a new patient is admitted, the hospital may want to establish immediately an acquaintance with her familydoctor. Suppose the view exported by the family doctor DB (say, DavisDB) has schema:
Patient(OHIP#,FName,LName,Phone#,Sex,PatRecord) Visit(OHIP#,Date,Purpose,Outcome)
Prescription(OHIP#,Med#,Dose,Quantity,Date) Event(OHIP#,Date,Description)
Figuring out patient record correspondences (i.e., doing object identification) is achieved by using the patient's
Ontario Health
Insurance # (e.g., OHIP#=1234). Initially, thisacquaintance has exactly one coordination formula
which states that if there is no patient record at the hospital for this patient, then the patient's record from DavisDB
is added to TGHDB in the PatientInfo relation, which can be expressed as:
∀fn ∀ln ∀pn ∀sex ∀pr.(DavisDB : Patient(1234,fn,ln,pn,sex,pr) →
TGHDB : ∃ tghid ∃n ∃a.(Patient(tghid,1234,n,sex,a,Davis,pr) and n = concat(fn,ln)))
When TGHDB imports data from DavisDB, the existentially quantified variables tghid, n and a must beinstantiated with some concrete elements of the TGHDB database. This amounts to generating a new TGH# for
tghid, inserting the Skolem constant
for a (which will be further instantiated as the patient's age) andgenerating name n by concatenating her first name fn and last nam本论文由英语论文网提供整理,提供论文代写,英语论文代写,代写论文,代写英语论文,代写留学生论文,代写英文论文,留学生论文代写相关核心关键词搜索。