Introduction

 

Global server load balancing (GSLB) is regarded as an important internet infrastructure that supports business on the internet. The scalability and the availability of the Web can be provided by distributing the Web servers where the client requests must be balanced among these Web servers in order to improve the performance. Network latency has been identified as an important metric to improve the quality of service (QoS). Given the limitations to response times because of the geographic distance between servers and end-users, it becomes important to leverage advanced architectures and functionalities for GSLB to realize fast responses.

 

In this article we will see how Citrix App Delivery and Security (CADS) service provide a SaaS based fully managed cloud load balancing solution. We will also discuss functionalities that contribute towards improving QoS.

 

Need for advanced GSLB for applications

 

GSLB is generally implemented to achieve:

 

• Disaster recovery
• Improved application performance
• Reliability
• Fulfil compliance requirements
• Efficient Connections
• Reduced latency

 

Among this, disaster recovery or high availability is one of the primary reason that a lot of businesses deploy resources redundantly at multiple places. In most typical configurations, application data is served  in one location (active) where most of its user accesses from and one or more standby (passive) locations that functions only when the active site fails. This mode of switching is also called as failover.

 

The other configuration is active-active deployment, where multiple sites are active at the same time, which demands more advanced implementation of the GSLB. Some of which include constantly monitoring the health and latency of the sites from the actual end user’s location. The service needs to actively have this data and co-relate the end user experience to make better decisions in facilitating the users to access an optimal site depending on various factors like user’s ISP to public cloud server link performance, real user monitoring data of clients accessing various public cloud endpoints like web application, APIs, CDNs, factoring in usage data in the form of time of the day, server performance, etc.

 

CADS service advanced GSLB capabilities

 

CADS service leverages real-user performance monitoring (RUM) data functionality and perform data-driven DNS or API-based global load balancing. This platform is unique in that it employs end-user-based probes for collecting real-time information from clients, along with synthetic monitoring from multiple PoP locations spread globally as shown in the Figure 1. Here, based on the type of configuration specified on CADS service, users receive the optimal site information for their DNS queries. These sites can be on-prem, public or private cloud.

 

 

Figure 1: CADS service with advanced GSLB capabilities

 

CADS service offers a simplified workflow to quickly deploy GSLB service for globally distributed application sites. Following are the steps to onboard yourself and use CADS service for your advanced GLSB solutions. For detailed steps click here.

 

 

In the following sections, let us see various use-cases that CADS service supports for Global server load balancing.

 

Disaster recovery site for on-prem datacentre in cloud - Active-Passive Mode

 

Use-case: Create a Disaster Recovery site in Public cloud for your existing On-Prem datacentre.

 

Create a multi-site application with CADS service in GSLB Active-Passive mode.

 

Note: Detailed procedure available

 

Specify Main and Backup Site

 

Add two sites (Existing On-Prem datacentre and new public cloud datacentre) for Active-Passive deployment as show in Figure 2.  

 

 

 

 

Specify GSLB parameters

 

Select Fail-Over as the algorithm (Figure 3). By default a priority of 1 is assigned to the primary site (site1) and an increasing priority of 2 to the standby site (site2) as shown in Figure 4.

 

 

Figure 3: Select desired GSLB configuration

 

 

Figure 4: Option to change the priority of the site

 

Note: Application data replication and sync between the primary and backup sites are not managed by CADS.

 

 

Figure 5: CADS service Active-Passive Deployment scenario

 

If site with priority 1 is down, client requests are directed to site with priority 2. If both of these sites are UP, traffic is directed to site1 since it has higher priority as shown in the Figure 5.

 

 

 

Even distribution of users across datacentres with persistence.

 

Use-case: Round-robin distribution of users across sites spread globally and  ensure consecutive requests should go to the same datacentre.

 

 

 

Create a multi-site application with CADS service in GSLB Active-Active mode, evenly distributing traffic across sites. Also configure stickiness.

 

Note: Detailed procedure available . Following are steps specific to the use-case

 

Specify the application site details

 

Add two or more sites for Active-Active deployment

 

Step3: Specify GSLB parameters

 

Select Round-Robin as the algorithm (Figure 6), Enable Stickiness with a time to live value of 120 as show in Figure 7.  This value controls the time duration in seconds within which subsequent client requests to the Multi-Site Application will be sent to the same site.

 

 

Figure 6: Select GSLB configuration

 

 

Figure 7: Enable site persistence with CADS stickiness settings

 

 

 

 

Figure 8: CADS service Active-Active Deployment scenario with Round Robin algorithm with site stickiness

 

As shown in the Figure 8, individual client DNS requests are served with the addresses of sites in a round robin way across all the datacentres. When the first site (On-Prem datacentre in this case) goes down, clients are redirected to the next datacentre in the round robin queue. Once the site is back up, client traffic continues to the current selected site until the client stickiness time of 120secs are not expired. Over time clients gets distributed across all the healthy datacentres.

 

 

 

Note: You control the distribution of traffic across your datacenters using weights to each site. For example, assign a weight 90 to Site 1 and weight 10 to Site 2. Weights are proportional, i.e. 90 % of the traffic is received by Site 1 and 10% by Site 2. You can alter this to control the traffic proportions to your datacentres

 

 

 

Optimal site selection with advanced GSLB capabilities

 

Use-case: Route users to the optimally performing site irrespective of their location proximity.

 

 

 

Create a multi-site application with CADS service in GSLB Active-Active mode with advance GSLB capabilities to optimally route users to the Site.

 

Note: Detailed procedure available . Following are steps specific to the use-case

 

Specify the application site details

 

Add two or more sites for Active-Active deployment.

 

 

 

Specify GSLB parameters

 

Select Optimal RTT as the algorithm as show in Figure 9. This option is only available for managed site and user-defined sites with radar enabled (more details).

 

 

Figure 9: Select GSLB configuration

 

 

 

Figure 10:  Penalty settings for sites with Optimal RTT GSLB configuration

 

You can penalize a site by adding an additional latency (Figure 10). When you add penalty to a site, its additional latency is added to the one calculated by Real User Measurements.

 

  

 

 

Figure 11: CADS service Active-Active Deployment scenario with Optimal RTT algorithm

 

When a Site goes down, as shown in Figure 11, users are redirected to the next available optimal site. Once the Site recovers users are redistribute again to their respective optimal site which is dynamically determined.

 

 

Figure 12: Optimal site selection

 

As shown in Figure 12, even if user’s physical location is closer to Site2 user will be directed to Site1 if it has lower application latency (response time in ms) when compared with Site2.

 

Geo based traffic steering

 

Create a multi-site application with CADS service and configure Geo based site preference

 

Specify the application site details

 

Add two or more sites for deployment. While adding a site, configure the geo fencing for the site based on your preference.

 

 

Figure 13: Configure geo fencing for site

 

As shown in the Figure 13, all users accessing the application from North America region will be served from site located in San Francisco, CA, USA as per this configuration. This works for all the GSLB algorithms.

Benefits to customers
 

ITM PoPs spread across the globe monitoring internet service provider, public cloud provider, content delivery service provider performance and ensuring end user experience is improved and get optimal global server load balancing service.

 

Call to Action

 

 Try out Citrix App Delivery and Security Service here.