Caché Distributed Data Management Guide

	Contents
Preface :	Introduction


Chapter 1:	ECP Overview
	1.1 ECP Features 1.2 Uses for ECP 1.3 ECP Architecture 1.3.1 Databases and Namespaces 1.3.2 ECP Application Servers and Data Servers 1.3.3 ECP Connections and Recovery
Chapter 2:	Configuring Distributed Systems
	2.1 Configuring an ECP Data Server 2.1.1 Restricting ECP Application Server Access 2.2 Configuring an ECP Application Server 2.3 Configuring ECP Remote Data Access
Chapter 3:	Monitoring Distributed Applications
	3.1 ECP Connection Information 3.2 ECP Connection States 3.2.1 Application Server Connection States 3.2.2 Data Server Connection States 3.3 ECP Connection Operations
Chapter 4:	Developing Distributed Applications
	4.1 ECP Recovery 4.2 Forced Disconnects 4.3 Performance Considerations 4.3.1 Memory Use on Large ECP Systems 4.3.2 Temporary Globals 4.3.3 Multiple ECP Channels 4.3.4 Load-balanced Application Servers 4.3.5 Repeated References to Undefined Globals 4.3.6 Big String Nodes 4.3.7 The $Increment Function and Application Counters 4.4 ECP-related Errors 4.4.1 <NETWORK> Errors 4.4.2 Rollback Only Condition
Chapter 5:	Distributed Cache Protocol (DCP)
	5.1 DCP Configuration

Appendix A:	ECP Recovery Guarantees and Limitations
	A.1 ECP Recovery Guarantees A.1.1 In-order Updates Guarantee A.1.2 ECP Lock Guarantee A.1.3 Clusters Lock Guarantee A.1.4 Rollback Guarantee A.1.5 Commit Guarantee A.1.6 Transactions and Locks Guarantee A.1.7 ECP Rollback Only Guarantee A.1.8 ECP Transaction Recovery Guarantee A.1.9 ECP Lock Recovery Guarantee A.1.10 $Increment Ordering Guarantee A.2 ECP Recovery Limitations A.2.1 ECP and Clusters $Increment Limitation A.2.2 ECP Cache Liveness Limitation A.2.3 ECP Routine Revalidation Limitation A.2.4 Conflicting, Non-Locked Change Breaks Rollback A.2.5 Kill of Large Global in Transaction Breaks Rollback A.2.6 Journal Discontinuity Breaks Rollback A.2.7 ECP Can Miss Error After Recovery A.2.8 Partial Set or Kill Leads to Journal Mismatch A.2.9 Loose Ordering in Cluster Failover or Restore A.2.10 Dirty Data Reads When Cluster Slave Crashes A.2.11 Dirty Data Reads in ECP Without Locking A.2.12 Asynchronous TCommit Converts to Rollback