Un collègue m’a posé une question aujourd’hui à laquelle, je n’ai pas su précisément répondre…
« Sur quel mécanisme se base Citrix pour déterminer qu’un serveur est plus chargé qu’un autre ? »
Naturellement, je lui ai répondu qu’il se basait sur des compteurs du perfmon… Mais je ne savais pas lui en dire plus…
Du coup, voici la réponse à la question, extrait de l’article suivant http://docs.citrix.com/de-de/xenapp-and-xendesktop/xenapp-6/lm-topics-all.html
List of Load Management Rules
Application User Load
Limits the number of users allowed to connect to a selected published application. This rule monitors the number of active ICA sessions using the published application. The default value to report full load is 100.
Context Switches
Defines a range of context switches per second for a selected server. A context switch occurs when the operating system switches from one process to another. The default value to report full load is 16000. The default value to report no load is 900—at that value this rule is ignored.
This rule uses the System: Context Switches/sec performance counter to determine load.
CPU Utilization
Defines a range of processor utilization, as a percentage, for a selected server. The default value to report full load is 90 percent. The default value to report no load is 10 percent—at that value this rule is ignored.
This rule uses the Processor: % Processor Time performance counter to determine load.
Disk Data I/O
Defines a range of data throughput, in kilobytes per second, for a selected server. The default full load value is 32767 kilobytes per second. The default no load value is 0 kilobytes per second—at that value this rule is ignored.
This rule uses the PhysicalDisk: Disk Bytes/sec performance counter to determine load.
Disk Operations
Defines a range of disk operation, in read/write cycles per second, for a selected server. The default full load value is 100 operations per second. The default no load value is 0—at that value this rule is ignored.
This rule uses the PhysicalDisk: Disk Writes/sec and Disk Reads/sec performance counters to determine load.
IP Range
Defines a range of allowed or denied client IP addresses for a published application. It controls access to a published application based on the IP addresses of the clients. You can define ranges of IP addresses, then select to allow or deny access if the client IP addresses are within the defined ranges.
This rule must be used in conjunction with another.
Load Throttling
Limits the number of concurrent connection attempts that a server handles. This prevents the server from failing when many users try to connect to it simultaneously. The default setting (High impact) assumes that logons affect server load significantly. This rule affects only the initial logon period, not the main part of a session.
The Load Throttling rule can be applied only to a server, not to an individual application.
Memory Usage
Defines a range of memory usage by a server. The default full load value is 90. The default no load value is 10—at that value this rule is ignored.
This rule uses the Memory: % Committed Bytes in Use performance counter to determine load.
Page Fault
Defines a range of page faults per second for a selected server. A page fault occurs when the operating system tries to access data that was moved from physical memory to disk. The default full load value is 2000. The default no load value is 0—at that value this rule is ignored.
This rule uses the Memory: Page Faults/sec performance counter to determine load.
Page Swaps
Defines a range of page swaps per second for a selected server. A page swap occurs when the operating system moves data between physical memory and the swap file. The default full load value is 100. The default no load value is 0—at that value this rule is ignored.
This rule uses the Memory: Pages/sec performance counter to determine load.
Scheduling
Schedules the availability of selected servers or published applications. This rule sets the weekly days and hours during which the server or published application is available to users and can be load managed.
Server User Load
Limits the number of users allowed to connect to a selected server. The default full load value is 100 and represents the maximum number of users the system can support on a server. Load Manager user loads are calculated using active ICA sessions only.
Pour ceux qui serait encore sur d’ancienne version de XenApp 4.0 ou plus ancienne,voir l’article CTX111965
Enfin pour compléter cet article et rester dans les perfmon, je vous invite à aller lire les articles ci-dessous :