Traffic processed by your 128T router for configured services may be distributed among viable next-hops using the 128T Networking Platform’s rich load balancing techniques. Configured in the service-policy, the load balancing strategy may be either hunt or proportional. In the above eLearning, we show you how to configure load balancing, define what we mean by proportional and hunt, and then show you how you apply load balancing to each service you want to load balance.
Load Balancing - Service Policy
As mentioned previously, load balancing is configured in the service policy, which is Global Data. That means, we can define just 1 load balancing policy and then apply it to as many services as we like.
When you go to configure your service policy, you need give it a name, and then select a strategy; either hunt or proportional.
The hunt strategy will order the set of potential next-hop targets from highest capacity to lowest and send traffic to the first one in the list until it approaches its capacity, then distribute traffic to the next agent in line, and so on. For example, if there are 3 possible next-hops and 1 can handle 500 sessions, 1 can handle 300 sessions, and 1 can handle 200 sessions, then the first next-hop will be selected until it is handling close to 500 sessions, and then the second next hop will be selected.
The proportional strategy distributes traffic among all of the possible targets in proportion to their relative capacities. For example, if we look at the same next hops as before: 1 that can handle 500 sessions, 1 that can handle 300 sessions, and that 1 can handle 200 sessions, then the first next-hop will get 50% of the traffic, the second will get 30% and the last will get 20%.
For the sake of a demo, let's give our service policy the name “lb-servicepolicy” and the strategy “proportional.”
Load Balancing - Service
Now that we have our service-policy configured, we just need to associate the service-policy to the service. Select the service you want to apply load balancing to, and then find the section titled “Policies” and select your service-policy where it says “Service Policy.”
For our example, we’re going to create a new service called lb-testservice. For our service address, we’ll give it 128.128.128.128/32. And then under service policy, we’re going to select lb-servicepolicy. So any traffic that enters this 128T and is headed for 128.128.128.128 will be meant for the lb-testservice service and will be load balanced proportionally to the next-hops as per lb-servicepolicy.
Load Balancing - Service Route Policies
Now we just need to define the routes that we will load balance between and the capacity of our next-hops.
Let’s do the capacities first. We configure these under service route policies within the context of a router. This means they are local data and only apply to a single router. When you click on service-route-policy, you are presented with four configurable options here:
-
Maximum Session Rate
-
Maximum Active Sessions
-
Session High Water Mark
-
Session Low Water Mark
The Maximum Session Rate is the rate limit threshold in sessions per second that mark the limit of this service route. The Maximum Active Sessions is the maximum number of active concurrent sessions a particular service route can handle. The Session High Water Mark is the percentage of maximum sessions this route will take before it is no longer selected for load balancing, and the Session Low Water Mark is the percentage of maximum sessions this service route needs to drop down to in order to be an option for load balancing again.
So let’s say I set my maximum active sessions to 100 for this policy and then I leave the watermarks at their defaults of 95 for High and 90 for Low. That means that once this route has it’s 95th active session, it will not be considered for load balancing. Only once it drops down to 90 can it be considered again.
For the demo, we'll configure 3 service route policies, one with a Maximum Active Sessions set to 500, another set to 300, and the last set to 200.
Load Balancing - Service Routes
Last thing we need to do is configure our service routes and then we commit. We configure service routes at the router level again. Just give them a name, select the service they apply to under “service name” and select the service route policy we want to apply for each service route under “service route policy.” Lastly, put in information about your next-hop and then you are done.
For the demo, we’re going to create 3 service routes, one for each of the service route policies we configured.
Once we validate and commit our config, we are all set to start load balancing our traffic. With the config we created here, we will be load balancing between 3 paths for our service lb-testservice. Service-route dnat-testservice1 will receive 50% of the sessions because we picked a strategy of proportional and set the max sessions as 500. Service route dnat-testservice2 will receive 30% of the sessions because its max session is 300 and service route dnat-testservice3 will receive 20% because its max session is 200.
Lastly, we're just going to add a route to our datacenter from our branch office router for traffic destined to 128.128.128.128.
Now, if we look at our FIB from my datacenter router, I can see an entry that says that any traffic with a destination of 128.128.128.128 coming from the customerx tenant will be allowed access to the lb-testservice and it’s next-hop is either 172.16.128.2, 172.26.128.2, or 172.36.128.2.
show load-balancer
Let’s send some traffic through our 128Ts and then look at the “show load-balancer” command which you can use to troubleshoot your load balancing.
From a client sitting behind our branchoffice router, we're going to send some http traffic to 128.128.128.128 as well as send some pings.
Now, log into the PCLI of your conductor, and issue the command “show load-balancer router bostonsite1” (that’s the datacenter router where we set up the load balancing). Show load-balancer will show us all service routes, so let’s scroll down to the load balancing ones. Here, we can see The capacity of each path, the max we configured for each path, the % of utilization and the rate per second.
#LoadBalancing