| There are several algorithms that you should know to help you determine the best configuration of TED components. The following sections shed some light on how the internals of TED components function. Construction of a Routing Hierarchy The distributor is responsible for constructing a routing hierarchy when it sends each distribution. The following describes the steps that the distributor follows to construct the hierarchy: The distributor discovers all the subscribers (that is, leaf subscribers) that are to receive this distribution. The distributor looks into the routing hierarchy that has been administered in its eDirectory object. For each subscriber, it looks to see whether that subscriber is in its hierarchy list. If it is, it sends the distribution to that subscriber following the routing hierarchy defined in the distributor object. If the subscriber is not found in the administered list, the distributor looks in the subscriber's object to see whether it has identified a parent subscriber. If a parent subscriber is identified, the distributor then looks for the parent subscriber in its routing hierarchy list. If the parent subscriber is in the routing hierarchy list, the distributor follows the hierarchy described to get to the parent subscriber and then tack the end-node subscriber onto the route to get the distribution to the subscriber. If the parent subscriber is not in the distributor's routing hierarchy list, the distributor sends the distribution directly to the identified parent subscriber (identified in the subscriber's object) or directly to the subscriber (when no parent subscriber is identified).
You should note that in the algorithm the distributor does not go to the subscriber and get its parent and then go to the parent and get its parent, and so forth, trying to construct the hierarchy of subscribers. The distributor goes up only one level in your TED distribution hierarchy. Therefore, you must, if you have more than one level of subscribers, describe n 1 levels in the routing hierarchy in each distributor object. You can then leave the last level undefined in the distributor, defining it in each subscriber object (or in a TED policy that affects each subscriber) instead. The following is a list of guidelines you should use when constructing your routing hierarchy: Include at least one parent subscriber on each LAN segment. This significantly reduces the amount of traffic across a WAN link. Include multiple parent subscribers on each LAN that has 40 or more subscribers. The performance of distributors and parent subscribers is significantly reduced when they are assigned more that 40 subscribers. Make certain that any subscribers that are not included in a distributor's distribution round are assigned to a parent subscriber on the same LAN.
Designing Distribution Routes Now that you understand how to construct a route hierarchy, you should consider the following when designing your Distributor's routing hierarchy: Distribution Route. A distribution route should represent the most efficient path to any given segment of your LAN or WAN. A distribution route is made up of a list of parent subscribers that are responsible for transferring distributions through the channel. The distribution route should use parent subscribers to minimize the workload for a distributor by spreading the workload of distributions. End-Node Subscribers. An end-node subscriber is the last subscriber in a distribution route to receive a distribution. End-node subscribers do not need to be added to the routing hierarchy, because they receive only the distributions. Distribution Topology. The distribution topology is the LAN/WAN topology that the distributions must follow on their assigned routes. You should consider your WAN topology design and the number of subscribers on each LAN and then build the routing hierarchy to reflect your network topology. Multiple Distributors. By using multiple distributors you can place the same subscribers under each of them, effectively creating the same distribution route for each distributor. Reuse Subscribers. A subscriber can be used more than once, as long as you have multiple distributors to assign it to. However, you should be careful not to overload the subscriber.
Scheduler Interactions As you introduce yourself to the various components of the TED system (distributor, subscriber, distribution, channel), you will discover that each component has its own independent scheduler to determine when its work is done. That seems clear enough at first, but it becomes confusing as you try to jumble how the schedules of these various components can keep or enable a successful distribution. This section briefly examines the various schedulers for each component and how they can play together to create a functioning TED system in your network. The following list looks at each of the components and the responsibility the scheduler plays: Distributor schedule. Determines when the distributor reads eDirectory for the configuration information for itself, its distributions, and any channels and subscribers it interacts with. It discovers the channels and subscribers by looking at any distributions that it has in any channels and the subscribers that are hooked to these channels. The distributor also looks for the eDirectory objects for any TED policies that may affect the configuration of itself, channels, or subscribers, and includes them in its configuration information. The information for a subscriber is collected from eDirectory when the distributor schedule fires, and it reads the information from eDirectory. Any changes to the subscriber objects or policies that affect the subscribers need to be made in eDirectory prior to the distributor refresh schedule firing. This information is then sent on to the subscribers when a distribution is given to them. Distribution schedule. This schedule determines when the agents on the distributor server will be activated to perform any gathering of the files and compact them into the single distribution file. This process must be completed before the channel begins its processing or the distribution needs to wait until the next channel cycle. The gathering does not occur if the distribution is inactive or does not belong to a channel with active subscribers. ZENworks for Servers 3 has added downstream control that forces all downstream distributors to stop distributions when the send schedule ends. The distributions start up again the next time the send schedule starts. Channel schedule. Determines when the distributor transmits any previously constructed distribution files to any destination subscribers. Any distribution files that have not been constructed when the channel is activated (based on the schedule) have to wait until the next firing of the channel schedule to be transmitted. The transmission is terminated if the time has expired for the channel, and the transmission picks up where it left off on the next scheduled time for the channel. Subscriber schedule. Determines when the subscriber activates the agents on the subscriber server to extract any files from distributions that have been received and have not been processed. The extraction continues after it has started on a particular distribution even if the time has expired for the extraction schedule.
Looking at the network, you would want the time frames to occur in your systems as represented in Figure 6.7. Figure 6.7. Schedule coordination time frame. 
For example, if you schedule your distributor to refresh the configuration each day at the close of the day (say 5:00 p.m.), and then schedule the distribution to gather the files at 6:00 p.m., and then the channels to begin their distributions at 10:00 p.m., and the subscribers to extract at 4:00 a.m., the cycle could be completed in the night. This obviously gets more complicated when you cross time zone boundaries. You could simplify this some by having, possibly, channels for each different area of the world, which would have a different local time to begin the distribution so that the remote site and the local site will be busy at appropriate levels. NOTE The channel schedule is "translated" into the distributor's time zone. For example, if the channel schedule says 3:00 p.m. MST and the distributor is in California, the distribution starts at 2:00 p.m. PST. If you have a distributor in New York that also has distributions in this channel, the distribution starts at 5:00 p.m. EST. This means that all distributors with distributions in a channel start sending at the same time, regardless of time zone.
|
