Mobile Computing Databases

Mobiles operates in environments with wireless communication, limited bandwidth, and intermittent connectivity.

Consists of:

• Mobile hosts: laptops, phones, or tablets.
• Wired hosts: servers or fixed systems.
• Communication network: Wi-Fi, 4G, 5G, Bluetooth.

Communication can either be:

• Through base stations
• Direct host-to-host

Database Issues

Applications which rely on a database must handle:

• Query optimization: minimize data transferred and network cost.
• Energy consumption: transmission uses more energy than receiving.
• Location services: queries depend on device location.
• Data broadcast: base stations broadcast shared data.
• Disconnection: must maintain cached local data and sync later.
• Consistency: local copies can become outdated.

Routing and Query Processing

Optimization balances:

• User time: query response delay.
• Communication cost: data transfer charges.
• Energy: minimize transmission usage. Reception requires much less energy than transmission.
• Timing: off-peak vs peak network costs.

Data Broadcast

• Base station broadcasts frequently accessed data.
• Clients listen instead of sending requests → saves power.
• Supports devices without transmit capability.
• Broadcasts can be scheduled (fixed or adaptive).
• Noise or signal loss handled by redundancy (parity bits).

Disconnectivity and Consistency

When devices are offline:

• Local updates may conflict when reconnecting.
• Cached data may become stale.

Challenges:

• Recoverability: updates may be lost.
• Consistency: local data may not match server state.

Mobile Updates

• If only one host updates → simple sync.
• If multiple hosts update → conflicts occur.
• Version numbering used to track document versions.
• Sometimes manual reconciliation is required.

Detecting Inconsistent Updates — Version Vector Scheme

Each host maintains a version vector per document.

For host i and document d:

• Vd,i[k] = version number of d at host k known to i.

Rules:

If all vectors equal → copies are identical. If all Vd,i[k] ≤ Vd,j[k] and not equal → i’s copy is older. If some elements are higher/lower on both sides → inconsistent copies (conflict).

Used in distributed and mobile systems to detect divergence.

Handling Inconsistent Updates

• Automatic merge is complex.
• Manual resolution often needed.
• Applied in mobile DBs and distributed file systems.
• Works if inconsistencies are rare or predictable.

Commercial Embedded Databases (Oracle)

Oracle offers several embedded database products for different scenarios.

TimesTen

• In-memory RDBMS for real-time, high-throughput applications.
• Used in telecom, finance, and CRM.
• Can serve as standalone DB or cache for Oracle DB.
• Features replication and high availability.

Feature	Oracle 10g	TimesTen
Optimization	Disk-centric	Memory-centric
Deployment	Database tier	Application tier
Response time	ms	µs
Data capacity	TBs	GBs

Berkeley DB

• Open-source, high-performance embedded key–value store.
• No SQL layer; accessed via API.
• Runs in-process with the application.
• Used by Amazon as a high-speed cache in front of Oracle Database.

Key features:

• ACID transactions.
• Low latency and high throughput.
• Minimal hardware requirements.

Oracle Database Lite

• Designed for mobile and field-force environments.
• Provides secure offline access and syncs when connected.
• Synchronization: async, event-driven, conflict-aware.
• Supports automatic reconciliation and device management.

Architecture:

• Embedded client DB + Mobile server for sync and provisioning.
• Sync uses In-Queue (upload) and Out-Queue (download) model.

Summary

• Embedded databases run inside applications, require no DBA, and have small footprints.
• Two architectures: lightly embedded (external engine) and deeply embedded (in-process).
• Mobile databases face challenges with energy, connectivity, and consistency.
• Version vectors help detect conflicts after disconnections.
• Oracle’s embedded solutions—TimesTen, Berkeley DB, and Database Lite—illustrate different deployment and synchronization models.