Apigee Hybrid Model- The hybrid way
Introduction :
Apigee is a very popular full lifecycle API management platform. companies have been using Apigee edge which is a saas platform by google apigee for quite long.
But two things which bother the companies using the saas model of apigee is security and latency, for fixing that apigee then came with an on-prem solution, one could buy a license from google apigee and can do its own apigee on-prem setup but then arises another problem of maintenance of all the components involved in the setup and migration from on-prem to cloud
So In this multi-cloud platform era, google offers a new solution in the form of apigee hybrid model which gives more security, reduced latency, more customizability and somewhat lesser maintenance especially when it comes to managing the analytics and UI components
Let us understand how it works
This model consists of two main components :
- Management plane: It consists of apigee edge UI, apigee management API, Developer portal and, Analytics of all the APIs proxy running in runtime plane. All these set of services are being hosted and maintained by Apigee on Google could platform.
- Runtime plane: It consists of API gateways which are managed by us, All the API traffic flows through these gateways which are in our own cloud or on-premises bringing it closer to our workloads Hence reducing the latency, increasing the security and giving more control over the API flows.
In this offering, it is the Apigee enterprise gateway that works in hybrid mode meaning all the components that are available in saas platform are available here and apigee does manage it.
let’s talk more about the run time plane logical components here:
Message Processor: This is the gateway which processes the API requests. It set up and uses Istio as an ingress gateway for intelligent routing, load balancing and TLS termination.
Synchronizer: This makes sure the replication of data from the management plane to runtime plane. If the developer makes any changes in management plane or creates a proxy then it takes care of the deployment of that proxy on the runtime plane.
Apache Cassandra: It is a runtime datastore, persistent volume which contains API keys, access tokens, KVM etc.
MART(Management APIs for runtime): It is an admin server that help manages/administer certain entities which are required by runtime plane such as API products, API keys, access tokens, cache, KVM.
The whole run time setup runs using an operator model in Kubernetes, All the components uses one-way TLS for all the internal communication.
Analytics
Analytics is an important part of API management. There is an agent called Universal Data Collection Agent (UDCA) which does it all for us.
It collects raw analytics data from each of the gateways and asynchronously sends it to the cloud. We use the same apigee edge UI for seeing these analytics which we use for creating API proxies. The agent is also capable of sending the information to more than one destination one of which is apigee cloud second one could be s3 or any data lake one may choose, and not need to export it manually for data visualization.
Robustness of using the Hybrid model
All the critical components required for the success of an API call are available in the runtime plane i.e. in our cloud or data centre. Even if any failure happens, for example, If the connectivity breaks between the management plane and the run time plane the gateway will still keep processing all the API calls because it has all the information with it for handling the requests so makes it reliable for handling production APIs.
Who should use the Hybrid APIGEE Model
If you want the API calls to be handled within your infrastructure for reducing latency and wants more control over its management and security also don’t want to give any extra effort in the management of the management platform this model is for you.
One may try to use it in their Developer and QA environment, Once feel confident with the load handling and reliability part may roll out in production.
Note: Apigee Hybrid runtime can be set-up on Google Kubernetes engine, on-prem and Azure Kubernetes services as of now.
