Справочник Пользователя для Macromedia flash media server 2-using flash media server edge servers
Introducing edge servers
17
When the load on an edge starts to scale down, so does its affinity value. Because the client
always looks to connect to the edge with the lowest affinity value, this edge begins to take
back connections that had spilled over into the next edge. The objective is to balance the
number of connection requests to the available resources, and have all connections to the same
application instance again pass through the same edge.
always looks to connect to the edge with the lowest affinity value, this edge begins to take
back connections that had spilled over into the next edge. The objective is to balance the
number of connection requests to the available resources, and have all connections to the same
application instance again pass through the same edge.
The client logic remains simple: to connect to the edge with the lowest affinity value. The
client only needs to know the affinity value for the different edges in the cluster to make the
connection to the right edge.
client only needs to know the affinity value for the different edges in the cluster to make the
connection to the right edge.
Clustering reverse proxies
Reverse proxies are configured to redirect incoming connection requests instead of outgoing
connection requests, as is the case with explicit and anonymous proxies. The client’s request to
connect to the desired application must first pass through the load balancer positioned before
the cluster of reverse proxies before it reaches the origin server. When reverse proxies are
deployed, the client's broadcast for a connection request cannot reach either an edge server or
the origin server and the requested application. The URI for the proxy cluster or the origin
server is not found on the client. Reverse proxying requires a different strategy for connecting
clients and applications.
connection requests, as is the case with explicit and anonymous proxies. The client’s request to
connect to the desired application must first pass through the load balancer positioned before
the cluster of reverse proxies before it reaches the origin server. When reverse proxies are
deployed, the client's broadcast for a connection request cannot reach either an edge server or
the origin server and the requested application. The URI for the proxy cluster or the origin
server is not found on the client. Reverse proxying requires a different strategy for connecting
clients and applications.
Clustering works efficiently when all clients that request the same application instance
connect to the same edge server in the cluster, where the application is cached. To accomplish
this efficiency with reverse proxying requires a multistep procedure.
connect to the same edge server in the cluster, where the application is cached. To accomplish
this efficiency with reverse proxying requires a multistep procedure.
1.
The client sends an HTTP request that includes an
XML.load
call to the origin server:
http://origin[:<port>]/fms/fpad?uri=<uri>
2.
The load balancer on the remote network intercepts the request and routes it to one of the
edge servers in the cluster.
edge servers in the cluster.
3.
This edge server then acts as an FPAD client and broadcasts an FPAD message.
4.
The edge servers enrolled in the cluster respond with their affinity values.
5.
The broadcasting edge server chooses the edge server with the lowest affinity value.
6.
This edge server sends the required information back to the client:
<?xml version=\"1.0\" encoding=\"utf-8\" ?>
<fpad>
<proxy>10.133.192.85:1935</proxy>
<timestamp>627539012</timestamp>
<userdata></userdata>
</fpad>
<fpad>
<proxy>10.133.192.85:1935</proxy>
<timestamp>627539012</timestamp>
<userdata></userdata>
</fpad>
where the
<proxy>
tag contains the
host:ip address
of the edge server to connect to.