Product Documentation

Configuring the Dynamic Method (RTT)

May 22, 2015

Dynamic round trip time (RTT) is a measure of time or delay in the network between the client’s local DNS server and a data resource. To measure dynamic RTT, the NetScaler appliance probes the client’s local DNS server and gathers RTT metric information. The appliance then uses this metric to make its load balancing decision. Global server load balancing monitors the real-time status of the network and dynamically directs the client request to the data center with the lowest RTT value.

When a client’s DNS request for a domain comes to the NetScaler appliance configured as the authoritative DNS for that domain, the appliance uses the RTT value to select the IP address of the best performing site to send it as a response to the DNS request.

The NetScaler appliance uses different mechanisms, such as ICMP echo request / reply (PING), UDP, and TCP to gather the RTT metrics for connections between the local DNS server and participating sites. The appliance first sends a ping probe to determine the RTT. If the ping probe fails, a DNS UDP probe is used. If that probe also fails, the appliance uses a DNS TCP probe.

These mechanisms are represented on the Netscaler appliance as Load Balancing Monitors and are easily identified due to their use of the "ldns" prefix. The three monitors, in their default order, are:
  • ldns-ping
  • ldns-dns
  • ldns-tcp

These monitors are built in to the appliance and are set to safe defaults, but may be customized just like any other monitor on the appliance.

The default order may also be changed by setting it explicitly as a GSLB parameter. For example, to set the order to be the DNS UDP query followed by the PING and then TCP, type the following command:

set gslb parameter -ldnsprobeOrder DNS PING TCP

Unless they have been customized, the NetScaler appliance performs UDP and TCP probing on port 53, however unlike regular load balancing monitors the probes need not be successful in order to provide valid RTT information. ICMP port unavailable messages, TCP Resets and DNS error responses, which would usually constitute a failure are all acceptable for calculating the RTT value.

Once the RTT data has been compiled, the Netscaler uses the proprietary metrics exchange protocol (MEP) to exchange RTT values between participating sites. After calculating RTT metrics, the appliance sorts the RTT values to identify the data center with the best (smallest) RTT metric."

If RTT information is not available (for example, when a client’s local DNS server accesses the site for the first time), the NetScaler appliance selects a site by using the round robin method and directs the client to the site.

To configure the dynamic method, you configure the site’s GSLB virtual server for dynamic RTT. You can also set the interval at which local DNS servers are probed to a value other than the default.

Configuring a GSLB Virtual Server for Dynamic RTT

Updated: 2014-11-24

To configure a GSLB virtual server for dynamic RTT, you specify the RTT load balancing method.

The NetScaler appliance regularly validates the timing information for a given local server. If a change in latency exceeds the configured tolerance factor, the appliance updates its database with the new timing information and sends the new value to other GSLB sites by performing a MEP exchange. The default tolerance factor is 5 milliseconds (ms).

The RTT tolerance factor must be the same throughout the GSLB domain. If you change it for a site, you must configure identical RTT tolerance factors on all NetScaler appliances deployed in the GSLB domain.

To configure a GSLB virtual server for dynamic RTT by using the command line interface

At the command prompt, type:

set gslb vserver <name> -lbMethod RTT -tolerance <value>

Example

 
set gslb vserver Vserver-GSLB-1 -lbMethod RTT -tolerance 10 

To configure a GSLB virtual server for dynamic RTT by using the configuration utility

  1. Navigate to Traffic Management > GSLB > Virtual Servers and double-click the virtual server.

Setting the Probing Interval of Local DNS Servers

Updated: 2014-11-24

The NetScaler appliance uses different mechanisms, such as ICMP echo request / reply (PING), TCP, and UDP to obtain RTT metrics for connections between the local DNS server and participating GSLB sites. By default, the appliance uses a ping monitor and probes the local DNS server every 5 seconds. The appliance then waits 2 seconds for the response and, if a response is not received in that time, it uses the TCP DNS monitor for probing.

However, you can modify the time interval for probing the local DNS server to accommodate your configuration.

To modify the probing interval by using the command line interface

At the command prompt, type:

set lb monitor <monitorName> <type> -interval <integer> <units> -resptimeout <integer> <units>

Example

 
set lb monitor monitor-HTTP-1 HTTP -interval 10 sec -resptimeout 5 sec 

To modify the probing interval by using the configuration utility

Navigate to Traffic Management > Load Balancing > Monitors, and double-click the monitor that you want to modify (for example, ping).