12.07.2015 Views

BROCADE IP PRIMER

BROCADE IP PRIMER

BROCADE IP PRIMER

SHOW MORE
SHOW LESS
  • No tags were found...

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Co ns ol ePwrCC U1 PU 2 PU 3TxActRxActActi ve1 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 6J -F16GCJ -F16GCServerIron 850J -F16GCJ -F16GCJ-F16GCJ -F16GCJ -F16GCChapter 5: Load Balancing BasicsA load balancer will want to check on each server frequently, so it will know assoon as a server becomes unavailable. You don't want to check too often, asthis could generate a lot of traffic and load on the servers. You don't want tocheck too little, because the longer you wait to check the health of the server,the longer it may take for the load balancer to realize a server has becomeinactive. A common interval would be about 5 seconds. For some applications,you may want to make the interval 30-60 seconds. This would minimize healthcheck traffic and reduce health check induced load on the server. Rememberthat with a 60-second interval, a server could be dead for 59 seconds beforeyour load balancer knows. That may be acceptable for some applications, butit may not be for others. The interval is adjustable to the needs of the V<strong>IP</strong>.There are three general types of health checks that are typically performed.They are Layer 3, Layer 4, and Layer 7.Layer 3The Layer 3 (or Network Layer) health check verifies the first three layers of theserver's configuration. It makes sure that the <strong>IP</strong> address of the server is reachablefrom the load balancer. To do this, the load balancer uses two Layer 3protocols: ARP and ICMP. In fact, specifically, it uses the ICMP echo and echoreplyprotocol, better known as ping.For the Layer 3 health check, the load balancer first sends an ARP request tothe server's MAC address. This is just to verify Layer 3 connectivity. Regardlessof the response (or lack of response) it gets, it will still try and send an ICMPecho request to the server. The server (assuming all is well on Layers 1, 2, and3) will send an ICMP echo-reply back to the load balancer. The load balancerwill then register the server as being active, and will pass traffic to it when itreceives it. If the load balancer does not receive an ICMP echo-reply after therequest's time-to-live (TTL), it will consider the server inactive, and it will notpass any traffic to it. At the next interval (on the ServerIron, the default is twoseconds), the load balancer will try another health check. After all, it needs toknow when the server has become active again. With a ServerIron, the loadbalancer will try to ping the host four times before it declares it inactive. This isconfigurable.“Hey, are you here?”ICMP Echo Request1 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 61 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 61 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 61 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 61 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 61 2 3 4 5 6 7 8 9 1 0 1 1 12 1 3 14 1 5 1 62 M M S W“Yeah, I’m here!”ICMP Echo Reply104 Brocade <strong>IP</strong> Primer

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!