Last week we discussed how the dynamic scale factor matters in Cloud and Enterprise deployments (/blogs/2012/05/31/scaling-your-services-dynamically-in-an-enterprise-or-cloud-environment/ ). For such environments the administrator has hardly any control on ensuring that the new entities are not impacted by traffic surges. Also no control on the impact such changes have on core load balancing logic within existing entities.
For the metric based Load Balancing methods once a new service is added to the LB farm, the load balancing logic moves to a slow start phase to ensure that we bring up the service slowly. This is good for the static use cases but in more dynamic environments you would want to control how the service should be ramped up than taking the whole logic into slow start mode. Getting into slow start mode certainly impacts the existing services and how the traffic was handled before.
NetScaler 10 introduces a new feature which allows you to specify key metrics as “Request Rate” and “Increment Interval” while NetScaler brings up a new service or the state has changed for an existing service. With these values NetScaler takes care of bringing up the new services while not impacting actual load balancing logic for the existing set of services.
For example, if we set:
New Service Startup request rate: 10 per second
Increment Interval: 2 minutes
Following is how the new services will be brought up till the point they fall under the common LB algorithm.
This is how you can scale your deployment dynamically while ensuring that your existing and new services are not impacted in any ways…