Mobile Computing and Databases (modified from ICDE98) Margaret H. Dunham Southern Methodist University Dept of Computer Science and Engineering Dallas,

Mobile Computing and Databases (modified from ICDE98)

Margaret H. Dunham

Southern Methodist University

Dept of Computer Science and Engineering

Dallas, Texas 75275

[email protected]

http://www.seas.smu.edu/~mhd

2/24/98 ICDE/SMU - Dunham 2

Outline

Introduction & Data Management IssuesQuery ProcessingData BroadcastingTransaction ProcessingProjects & ProductsConclusion


Mobile Computing Architecture


Terminology

Fixed Network (FN)Base Station (BS) (Mobile Support Station -

(MSS))Fixed Hosts (FH)Cell - Area covered by BS (1-2 miles)Handoff - Changing BS by intercell moveMobile Host (MH) (Mobile Unit (MU))


Wireless Networks

Cellular High Cost Scalability Issue Limited Bandwidth: 10 Kbps

Wireless LAN Traditional LANs with wireless interface Low Cost Limited range: 10-100 meters Bandwidth: 10Mbps NCR Wavelan, Motorola ALTAIR


Wireless Networks (cont’d)

Satellite Services Wide Coverage Very Expensive Low Bandwidth: 1-2Mbps

Paging Networks Wide Coverage Sky Tel, Motorola

Slow: (Ethernet: 10Mbps; FDDI or switched Ethernet: 100Mbps; ATM: 155Mbps)

Ad Hoc Networks


Handoff

Changing BS due to movement between cells

State information transferred

Current handoffs in cellular phones may take up to a few seconds with breaks in conversation of 100-300 ms.

Soft - Temporarily connected to two BSsHard - Only connected to one BS


Location Management

Tracking mobile userUser associated with home location

server (Home Agent)May augment by searching in local

area firstMay augment with user profilesMobile IP [11,14]

Triangle Routing Route Optimization Location Control (Routing Agent)

M

S

A

A

f

h

AfAh

S

M


Location Management (cont’d)

Active Badge (Cambridge,[2]) Track employees and route telephone calls Unique code emitted every 15 seconds Sensors placed in offices and corridors

Location Information Replications No HLR Hierarchy of Location Servers Each server maintains information about its subtree


Mobile Applications

Information Services (Yellow Pages)Law Enforcement and Medical EmergenciesSales and Mobile OfficesWeather, Traffic, Sports, EntertainmentTruckingCellular Subscribers in the United States:

90,000 in 1984;4.4 million in 1990;13 million in 1994

Handheld computer market will grow to $1.77 billion by 2002


Technology Push

Internet: ftp, telnet, email, http,htmlAdvancing Wireless Communication

TechnologiesLaptop, Notebook, and Palmtop Computers


Classification of Mobile Database Systems


Data Management Issues

Speed of wireless linkScalabilityMobility Location dependent data; Location specific queriesLimited by battery powerDisconnection (Voluntary, Involuntary)Replication/CachingHandoff


Insurance Example


Medical Example

911 CallAmbulance arrives/departsClosest hospitalAccess patient recordsSend vital signsUpdate patient recordsPage hospital personnelOrder medical supplies


MC/DB Research

Transaction ProcessingCaching - ReplicationBroadcast DisksAgentsMobilityLocation Dependent DataRecoveryACID (?)


Outline

Introduction & Data Management Issues Query Processing

Location Dependent Queries and DataNew Query TypesQuery Optimization

Data BroadcastingTransaction ProcessingProjects & ProductsConclusion


Location Dependent Data

Value of data depends on locationTemporal Replication - One consistent value at one

timeSpatial Replication - Multiple different correct data

values at one timeTemporal Consistency - All data objects satisfy a

given set of integrity constraints.Spatial Consistency - Consistency constraints

satisfied within Data Region.SMU/University of Missouri at Kansas City, [17]


Location Dependent Queries

Result depends on locationDifferent from traditional distributed goal of

location independenceEx: Yellow Pages, Directions, MapPredicates based on location: “Find the

cheapest hotel in Dallas.”Location constraints: “Find the nearest hotel (to

me).”


Similarity to Spatial Queries

Spatial Data: Data associated with space occupied by object.

Types of spatial queries: contains, contained in, intersects, neighboring, east of, etc.

Spatial data structuresSpatial operators Spatial selects and joinsPSQL - extend SQL, [18,20]


Differences from Spatial Queries

Client is actually movingLocation of client may be

part of the query itselfMay depend on direction of movementData may not directly contain location

information Includes temporal features as well

Spatial data is dynamic


Querying Moving Objects

Moving Objects Spatio-Temporal (MOST) data model Dynamic Attributes - Change over time Queries over temporal history:

Instantaneous - Ex: “Find all restaurants I’ll reach in the next half hour. ”

Continuous - Ex: “Find all restaurants within 5 miles.” The answer continuously changes as the MU moves.

Persistent - Ex: “Find the cars that travel greater than 10 miles in the next half hour.”

Future Temporal Logic (FTL) language University of Illinois, [20]


Query Optimization

How best to satisfy the information request made by the client?

Different Cost Factors: I/O, networkDifferent Access Options: cache, FN, broadcastDynamic and Adaptable - environment changesAlternative plans include deciding (based on state

of MH and environment) whether to access in the cache at the MH, to request a mobile transaction, or to obtain from a broadcast disk.


Outline

Introduction & Data Management IssuesQuery Processing Data Broadcasting

OverviewIndexingResearch

Transaction ProcessingProjects & ProductsConclusion


Data Broadcasting

Server continually broadcasts data to MUs. Scalability: Cost does not depend on number of users

listening.Mobile Unit may/may not have cache.Facilitates data access during disconnected periods.Allows location dependent data access.No need to predict with 100% accuracy the future data needs.Broadcast based on probability of access.Periodic broadcasting of all data.


Data Broadcasting (cont’d)

Classification: Coverage - Everything, Subset Content - Static, Dynamic Indices - Index, Self Descriptive Data Stream - Flat, Skewed, Multiple Disks Client - Passive, Active

For uniform page access, flat disk has best expected performance.

With skewed page access, nonflat disks are better.Push based.


Broadcast Disks

Simulate multiple disks of varying sizes and speeds. Data of higher interest on smaller faster disks (broadcast more frequently).

Each “disk” contains data with similar access behavior.

Combination of caching and broadcast disks.

Figure 4.1 from [15]


Broadcast Disks (cont’d)

Don’t want to store hottest pages. They may be broadcast frequently.

Store in cache if probability of access (P) is greater than the frequency of broadcast (X).

Cost based page replacement.Replace cache page with smallest P/X - PIX. Too

expensive to implement.LIX - PIX approximation. Works well particularly with

noise.Brown, MITL, Maryland, [37,38,39]


Air-Cache

Dynamic - Adapts to system workload.Define temperature of data:

Vapor (Steamy) Hot - Accessed frequently and broadcast.

Liquid Warm - Accessed often, not broadcast, but kept in server’s main memory.

Frigid (Icy) Cold - Accessed infrequently and stored on secondary storage.


Air-Cache (cont’d)

Three level memory hierarchy based on temperature.

Sparks (access) to data can increase temperature. No sparks, results in a reduction of temperature.

Simulation results predict very good performance.

Maryland, [43]


Adaptive Protocols

Dynamically modify broadcast contents.Constant Broadcast Size (CBS) Server Protocol:

Limited size and periodic Priority Popularity Factor (PF) Ignore Factore (IF)

Variable Broadcst Size (VBS) Server Protocol: Aperiodic All data above threshold PF included.

Arizona and UMKC, [40]


Outline

Introduction & Data Management Issues Query Processing Data Broadcasting Transaction Processing

OverviewTransaction ModelConcurrencyRecoveryResearch

Projects & Products Conclusion


Mobile Transaction (MT)

Database transaction requested from a MU. May execute in FN or MU

IssuesDisconnect/HandoffMobilityLocation Dependent DataError ProneMU Resources/ PowerRecovery/Restart Management


MT Requirements

Keep autonomy of local DBMSLLT InteractiveAdvanced transaction models

Nested Multidatabase

Request from MUExecute anywhereCapture movementACID (?)


MT Approaches

No consensus on accepted approachMU may not have primary copy of data [45]:

Transaction Proxy: MU does no transaction processing Read Only Transaction: MU only reads data Weak Transaction: Read and update cached data;

Must synchronize updates with primary copy on FN.

MU may have primary copy of dataMU may access data on other MUsFirst class and second class transactions


MT RecoveryTransaction, site, media, network failure - More frequent

than in wired network.Different types of failures (partial)

Handoff Voluntary disconnection Battery problems Lose computer??

Checkpoint data at MU to BSCheckpoint at handoffDatabase log plus transaction logMay need compensating transactions


Atomicity for MT

Weaken or provide different types of atomicityMay decompose transaction into

subtransactionsMay require atomicity at lower than transaction

levelAtomic commitment difficult (expensive)Global commit/Local Commit


Consistency for MT

Weakening isolation and atomicity may weaken this as well.

May divide data into clusters with consistency within clusters.

Reintegration of updates after reconnect may cause many conflicts.

May use bounded inconsistency. Impacted by location dependent data


Isolation for MT

May be too restrictiveCan’t always do at MU (disconnection) Isolation at lower levels in transactionCommitment at different levels of transactionCooperating transactions


Durability for MT

Durability for partial resultsMay want durability for parts of transactions.Due to conflicts at reconnect, even durability of

subtransactions may not be guaranteed.Local commit vs.. Global commit


MT Concurrency Control

Mobility of MUs may increase message traffic for lock management

MU failure may leave some data locked /unlocked

6) T1: Unlock(Xa); Commit;

Server ACell A

XaYa

Server BCell B

XbYb

Server CCell C

XcYcZc

1) T1: Lock(Xa); Read(Xa)

2) T1 moves to B

3) T1: Lock(Yb); Read(Yb)

4) T1 moves to C

5) T1: Lock(Zc); Write(Zc);

Unlock(Zc); Commit

6) T1: Unlock(Yb); Commit;

Fig 2 from [48]


Revised Optimistic Locking

O2PL-MT Read locks may be

executed at multiple servers.

Read unlock can be executed at any site

Benefit shown using analytic model

Purdue, [48]

LOCK HELD

LOCKREQUEST

W_INTEND R_LOCK W_LOCK

W_Intend No Yes No

R_Lock No Yes No

W_Lock No No No

Figure 3 from [48]


Kangaroo Transaction (KT)

Built on top of global transactionsCaptures data and movement behaviorDAA as BS - Maintains logging and transaction

statusLogging at BSFlexible atomicityRestart after disconnectManagement moves


Kangaroo Transaction (cont’d)

Local Transaction - Sequence of read and write operations ending in commit or abort

Global Transaction - Sequence of global or local transactions

Joey Transaction - Sequence of global and local transactions ending in commit, abort, or split

Kangaroo Transaction - Sequence of one or more Joeys with last one ending in commit or abort. All earlier end in split

SMU, [47]


KT and Movement


Reporting and Co-Transactions

Mobile transaction is a special type of multidatabase transactions.

GDMS exists at each base station.Subtransactions of the mobile transaction will

commit or abort independently.Atomic and non-compensatable transactions.Reporting and co-transactions.Pittsburgh, [46]


Clustering Model

Views mobile transaction as beginning on mobile and nonmobile hosts.

Transaction migrationTransaction model is designed to maintain

consistency of the database.Database is divided into clusters.Data is divided into core and quasi copies.Mobile transactions and operations are decomposed

into a set of weak and strict transactions.


Clustering Model (cont’d)

Weak operations access only data in the same cluster. Strict operations allowed database wide access. Two copies of data can be maintained (strict and weak).

Clusters defined based on location and user profile.Transaction Proxy: dual transaction of one executed

at mobile host which includes only the updates. Purdue, [51,52]


Mobile Transactions and Ambulatory Care

Medical Personal Digital Assistant (MPDA)Battlefield - Cache copy of soldiers’ medical records in

MPDADistributed Medical Database - EMT obtains patient’s

medical record and updates.BSA (Base Station Agent) is responsible for logging

and recovery.Recovery based on sagas with save-points.Mailboxes used to save information.Purdue, [49,50]


Semantics-Based Mobile Transaction Processing

Views mobile transaction processing as a concurrency and cache coherency problem.

A stationary database server dishes out the fragments of an object on a request from a Mobile Unit.

On completion of the transaction, the Mobile Units return the fragments to the server.

These fragments are put together again by the merge operation at the server.

Pittsburgh, [54]


Multidatabase Transaction Processing Manager

Mobile transactions built on top of multidatabase global transactions.

Timestamps used to enforce orderingAllows voluntary disconnections.MU part of MDSMessage Queuing Facility (MQF)MU sends request to designated coordinating

node on FN.Monash, [56]


PRO-MOTION

MC/Database Transaction Processing approachMultiple transaction types

Controlled divergence ACID Update cache and later DB at FH

Compact - Compact Agent at MU, Mobility Manager at BS, Compact Manager at Server

Pittsburgh, [55]


MT Research Limitations

Architectural AssumptionsNo support for location dependent dataFew Implementations


MT Management Options

MUBSCombinationFixed/Relocatable/Moving

Agent


Outline

Introduction & Data Management IssuesQuery ProcessingData BroadcastingTransaction Processing Projects & Products

Conclusion


Some DB/MC Projects URLs MobiDick - Monash Univ. (Australia);

http://www.ct.monash.edu.au/~mobidick Mobisaic - Univ. of Washington;

http://www.cs.washington.edu/homes/voelker/mobisaic Purdue; http://www.cs.purdue.edu/research/cse/mobile SMU; http://www.seas.smu.edu/~mhd/mobile.html MCC - Collaboration Managment Infrastructure;

http://www.mcc.com/projects/transaction University of Ioanina; http://zeus.cs.uoi.gr/ Michigan - CITI; http://www.citi.umich.edu/projects/mobile.html UCLA - Ficus; http://ficus-www.cs.ucla.edu/ficus Columbia; http://www.mcl.cs.columbia.edu


Rover

Figure 6.1 from [15]


Oracle Mobile Agent

Commercial ProductApplication, Static,

MultipleMessage Manager - MUMessage Gateway - BSAgent - FN (Server) [67,69]

Message Manager

Gateway

Corporate Network

Agent

Database Server


Sybase - SQL Anywhere

Designed for Windows, (95, 3.x, NT), OS/2, DOS

Limited memory requirements

Full TP capabilities Includes SQL RemoteCompatible with Sybase

SQL Server [68]

Remote DatabaseSQL Anywhere standalone engine

Message agent

Consolidated DatabaseSQL Anywhere network server

Message agent


Sybase (cont’d) - SQL RemoteTwo way replication based on

message passing.Remote database are synchronized

with consolidated DBMessage Agent required at DB serverReplication of subscribed fragmentsPeriodic changes sent from consolidated

DB to remote DBsUpdates from committed transactions at remote submitted to

consolidated database. Conflicts: Consolidated is master; Triggers used.

ConsolidatedDB

Remote Databases


Informix

I-Mobile 1.0 discontinued: No replication Three tier approach appropriate for long term, but in the

short term users wanted to be able to use existing client-server applications (not rewrite).

Small DBMS server to run on mobile client Only dial up needed for now

Informix Dynamic Server/Personal Edition (IDS/PE) for Windows 95/NT. Mobiles and desktop clients

[64,66]


Outline

Introduction & Data Management IssuesQuery ProcessingData BroadcastingTransaction ProcessingProducts Conclusion


Future

Combine different approachesSemantic cachingQuery OptimizationAdaptive Data BroadcastingPerformance BenchmarksSecurityLocation Dependent Queries


Acknowledgements and URL Bibliographies

Earlier version of this tutorial presented at the 1996 Brazilian Database Symposium.

We particularly want to thank Evaggelia Pitoura for providing several tables and figures from her recent book [15].

Some slide information obtained from slides presented at a database class at the University of Massachusetts, http://www-ccs.cs.umass.edu/mobile.

Online bibliographies http://www.seas.smu.edu/~mhd/mobile.html http://www.ct.monash.edu.au/~mobidick

Mobile Computing and Databases (modified from ICDE98) Margaret H. Dunham Southern Methodist University Dept of Computer Science and Engineering Dallas,

Documents

bs slide

subtree slide

icdesmu dunham2

motorola altair slide

mobile offices zweather

mbps fddi

mbps atm

mbps zad