Status information provided to the GRM server

Type of status information sent by GRM agent caches

GRM agent caches continuously send GRM servers the information shown in the following table. A GRM server considers this information along with other factors in determining the GRM agent cache IP addresses to return to a client.

Information	Comments
System health	Center and edge caches automatically report their status every 30 seconds. A GRM server considers a center or edge cache down if it has not received a report from the cache for 60 seconds. A GRM server does not return the IP addresses of any edge and center caches that are down.
CPU load	When you configure a GRM server, you set the maximum CPU activity level for each center and edge cache. Each center and edge cache reports its CPU threshold every 30 seconds. The GRM server does not return the IP addresses of center and edge caches that report a CPU load greater than the configured threshold.
Client-to-cache proximity	Edge caches: When you configure a GRM server, you enter CIDR blocks associated with each edge group. This configuration enables the GRM server to determine the set of edge caches to associate with a client request (for a GRM L7 server) or with a local DNS query (for a GRM DNS server). No active proximity measurements are needed. For clients associated with an edge group, the GRM server returns only the IP addresses of edge caches in that edge group. Center caches: The GRM server checks its proximity tables or sends a probe query for proximity measurements to center caches in order to determine the distance between a center cache and a client that is not associated with an edge group. A GRM DNS server measures the distance between the local DNS nameserver that sent the query and the GRM DNS server. A GRM L7 server measures the distance between the client and its GRM agent caches. Measurement intervals increase geometrically (30 seconds, 1 minute, 2 minutes, and so on).

Information

Comments

System health

Center and edge caches automatically report their status every 30 seconds. A GRM server considers a center or edge cache down if it has not received a report from the cache for 60 seconds. A GRM server does not return the IP addresses of any edge and center caches that are down.

CPU load

When you configure a GRM server, you set the maximum CPU activity level for each center and edge cache. Each center and edge cache reports its CPU threshold every 30 seconds. The GRM server does not return the IP addresses of center and edge caches that report a CPU load greater than the configured threshold.

Client-to-cache proximity

Edge caches:

When you configure a GRM server, you enter CIDR blocks associated with each edge group. This configuration enables the GRM server to determine the set of edge caches to associate with a client request (for a GRM L7 server) or with a local DNS query (for a GRM DNS server). No active proximity measurements are needed.

For clients associated with an edge group, the GRM server returns only the IP addresses of edge caches in that edge group.

Center caches:

The GRM server checks its proximity tables or sends a probe query for proximity measurements to center caches in order to determine the distance between a center cache and a client that is not associated with an edge group. A GRM DNS server measures the distance between the local DNS nameserver that sent the query and the GRM DNS server. A GRM L7 server measures the distance between the client and its GRM agent caches.

Measurement intervals increase geometrically (30 seconds, 1 minute, 2 minutes, and so on).

Mechanisms for measuring proximity

The GRM server uses a number of mechanisms to determine proximity, changing from one to another if one does not produce results. The GRM server uses ICMP Echo request and response, DNS query, and traceroute. A GRM server might not be able to measure proximity if traceroute provides inconsistent results, which could result from blocking by firewalls on a WAN. In addition, for security reasons, firewalls and routers sometimes block the GRM proximity measurement traffic.

Additional information about proximity for a GRM DNS server

The assumption is that the round-trip time between the GRM DNS server and the local DNS nameserver that sent the query is similar to the actual round-trip time that the client will experience. In the majority of cases, this assumption is valid. However, an alternative deployment scenario might be required in the following circumstances because the GRM proximity measurements might be inaccurate:

If the local DNS nameserver is configured for recursive queries (that is, the local DNS nameserver does not take a direct path to the authoritative GRM domain)
If the local DNS nameserver is located a great distance from the client
Note
If the local DNS nameserver is located a great distance from the client, Network Appliance recommends that you deploy a GRM L7 server.

Communication protocol used for status updates

The GRM server and the center and edge caches communicate with each other over TCP. Enabling the GRM server or GRM agent caches automatically enables NetCache to listen on TCP port 89 for communication between the GRM server and GRM agent caches.

Viewing the IP addresses that were returned

The GRM server access log shows which center and edge caches the GRM server returned. See the Guide to Messages and Log File Entries and the Appliance Manager online Help for more information about the GRM server access log.