Capacity Planning and Performance Tuning in Java Draw Code 3 of 9 in Java Capacity Planning and Performance Tuning

7. generate, create barcode 3/9 none for java projects iOS Capacity Planning and Performance Tuning CHAPTER Deploying Highly Available LDAP Data Services LDAP direct ory servers provide their own high availability features by performing directory data replication. With replication, the content of a directory tree, or subtree on one server, is duplicated on other servers. If one of the servers fails, the content is still available on the other servers.

However, the current iPlanet Directory Server supports only a single-master replication model. This means that only one server at any one time is granted the access rights to update the content in the directory. If that server fails, updates cannot take place until the server is back online or a replica server is granted update permission.

This chapter looks at a method for automatically switching over to a backup master directory server if the primary one fails. The software to do this is Sun Cluster 2.2 in conjunction with the Highly Available Data Services for the LDAP module.

Even with cluster deployments, LDAP replication plays an important role. Most companies are geographically dispersed with Wide Area Network (WAN) connections between sites. By positioning replicated LDAP directory servers at end points of WAN connections to service local users, you conserve precious bandwidth.

The second part of this chapter presents LDAP replication architecture examples and deployment strategies for making LDAP data services available with the efficient use of network bandwidth.. iPlanet Directory Services 4.12 HA Architecture Models This sectio n provides working examples of the High Availability (HA) architecture that will enable you to understand what iPlanet Directory Server HA models are available and the concepts behind them. Since no single architecture fits every environment, we provide material about model availability to inform your decision about an architecture..

High Availability Strategy When design ing the architecture of a highly available directory strategy, you must be aware that availability comes at a price. It is generally thought that the more highly available a system is, the more its design and operations cost will be. One of the major reasons for designing the architecture of highly available services is to prevent against the loss of business due to application service outages and downtime.

In the case of directory services, unavailability of an application service can lead to loss of income, potential Internet subscribers, and even future revenues. The value of HA to directory services customers is directly related to the costs of downtime. This means that the higher the cost of downtime, the easier it will be to justify the additional expense of implementing HA.

Take the case of Internet Service Providers (ISPs) who are core users of the directory services. They probably have service-level agreements with subscribers whereby they guarantee a certain level of availability. If ISPs do not meet this service-level agreement, they could incur financial penalties.

To provide HA solutions for your organization s directory services you must identify the goals of providing HA services. Possible goals might include performance, load balancing, and management. Once you understand your organization s goals, then you can look at some of the available options.

Understanding the design goals will influence the way you design an architecture of an HA solution. Some of the more common approaches to designing high availability directory services include:. s s s Replication Models Referral Models Asymmetric HA (hot standby model). Replication Models Replication models support a single writable Supplier, and multiple read-only Consumer servers. The mechanism used is a simple replay-based replication scheme, in which a single-master server records the changes made to it and at some later time replays those changes to Consumer servers, which hold a read-only copy of the replicated subtree. Consumer servers may service bind, unbind, search, and compare operations on the replicated subtree, but they refer add, delete, modify, and modifyRDN/modifyDN operations to the master server, unless those operations come from the master server itself.

The Supplier server stores state information in a special attribute (the copiedFrom attribute) in the Consumer s copy of the entry at the top of the replicated subtree. Before commencing replay of changes, the Supplier retrieves the value of this attribute and sends only changes that have occurred since the last replication run..

Copyright © . All rights reserved.