We recently hosted our third Ask the Cloud Experts (ACE) Meetup, a monthly series focused on giving you an additional channel to connect with Citrix experts and get answers to your Citrix Cloud questions. At our June 19 meetup, we discussed Citrix ADC Autoscale with ADM. This blog will cover some of the highlights of the session as well as address the questions attendees asked.
The term Autoscale is commonly discussed when leveraging public clouds like AWS or Azure, where the primary objective is to tap into the virtually unlimited ability of cloud to automatically adjust your capacity to maintain steady, predictable performance at the lowest cost.
Autoscale is often used in AWS and Azure where one would create an Autoscaling group / VM Scale Set and define scaling policies (when to scale out and when to scale in) for your image. If done correctly, this would allow AWS or Azure to automatically spin up more copies of your desired instance and reduce capacity as needed without guessing peak-load capacity.
In the on-prem ADC world, Citrix supports clustering, as well, where you can create a cluster of multiple appliances working together as one logical appliance. Internal logic would automatically ensure that traffic is distributed correctly to all the nodes in the cluster. Refer to the Citrix Clustering guide for more details.
Autoscale with ADM (aka Clustering) in the cloud leverages clustering technology similar to on-prem. Citrix ADC Autoscale with ADM simply allows for dynamic creation of ADC instances to handle unpredictable application traffic patterns.
The following apply to Autoscale with ADM today:
- Autoscale is supported only with ADM (previously MAS) to allow for new VPX instances of ADC to be spun up automatically or shut down based on metrics like CPU, memory, and throughput. This blog and the associated on-demand recording reference the Autoscale configuration as it pertains to scaling of the ADC instances.
- Citrix supports Autoscale for backend traffic on AWS and Azure, as well. This refers to a configuration where the ADC can automatically determine the status of backend services that use Autoscale and add/remove them from the service group intelligently.
- Autoscale leveraging ADM is supported for both AWS and Azure. There are several prerequisites in place both on the AWS/Azure side and ADM. The vast majority of those prerequisites can be handled using automation (CFT, ARM, StyleBooks, etc.).
- ADM with Autoscale generally should be leveraged for green field deployment for standardized configuration that would benefit from StyleBooks and scale horizontally to accommodate for traffic pattern spikes.
- Any features outside for LB/CS provided by Citrix Autoscale StyleBooks can be implemented with changes by personalizing your StyleBook with the appropriate logic.
- Migration from on-prem to cloud (AWS/Azure) will not be able to seamlessly and automatically inherit Autoscale functionally. Application configurations will need to be converted to individual StyleBooks that include Autoscale support.
Here are some of the questions we got from attendees during our session:
It looks like we are not using dynamic routing but an IP Set across the cluster. How does this work with NLB /Azure LB for distribution?
Each IP added to the IP set is mapped with one of the nodes in the cluster. Each of the above-mentioned IPs are added to the backend pool of the NLB/Azure LB.
How do you configure persistence to ensure stickiness in this deployment?
There are two levels of persistency. The decision to stick to a specific cluster is maintained by an ELB or Azure LB using IP based persistence. The decision to reach the backend server is maintained by the cluster.
Is there a backplane? If so, which interfaces is that occurring on? Do we need dedicated interfaces for the backplane, or does it only happen over the primary interface that is usually the management plane?
There are no dedicated interfaces used for backplane traffic in this deployment. The primary interface is used as the backplane.
Are we then limited to 1Gb of backplane traffic if steering has to occur?
There is no 1Gb limit for server-side traffic. Backplane steering is eliminated by using SNAT, thereby ensuring symmetric traffic flows between the cluster nodes and backend services.
Does Autoscale functionality work without ADM or ADM Service?
No. ADM or ADM Service is required. The POC done previously for this exercise was done using ADM Service. A 30-day trial is offered for ADM Service at cloud.citrix.com.
What type of StyleBooks do are supported using AutoScale?
We only support LB and CS for now. There are plans to tentatively support more StyleBooks for other features in the near future. We actively welcome any feedback or interests you have. Please connect through your Citrix Sales team.
How do I integrate AutoScale to an existing ADC environment in the cloud?
This is currently not possible. Connect with your Citrix sales team and let us know if this is an important step for your business.
Are there any limitations that we need to be aware of when using AutoScale?
Some of the ones discussed in the presentation included:
- Only AWS and Azure Marketplace licenses are supported today. BYOL is not supported today. Note: BYOL support is being planned for the future.
- UDP applications are not supported with the ADC Autoscale feature in Azure specifically.
- StyleBooks and ADM are crucial parts of this deployment.
- Load Balancing and CS are the currently supported StyleBooks for this deployment. Additional StyleBooks can be personalized using the existing StyleBooks for AutoScale support.
Sign Up Today for the Next ACE Meetup
Thank you for reading and be sure to check out our next ACE Meetup on July 17 at 9 a.m. ET. There will be a brief presentation on Director and Studio, followed by live Q&A to address any and all cloud questions with our panel of experts.
— Jay Chandrasekar, Senior Architect; Dave Potter, Senior Solutions Architect; Hariharan Subramanian, Product Manager; Arvind Kandula, Principal Product Manager