Adobe is modernizing its Edge Compute infrastructure with Aerospike to help enterprises deliver the best experience to their consumers.
– Sandeep Nawathe, Senior Director of Engineering, Adobe Cloud Platform, Adobe
I lead engineering groups for what is called the platform and core services for Adobe. This is the group that actually is building the common platform that will become the foundation for all Adobe solutions in the Experience Cloud.
Aerospike is becoming a key component of our edge computation story today, which is becoming the backbone of the data storage replication and distribution for the entire edge computation for all Adobe infrastructure.
Today we see that the number of requests that come in for which we need to respond are in the range of several million requests coming every single second. These requests, many of them actually require look up times in a couple of milliseconds. And we need a stable characteristic, so we always look at 95th percentile, where do we land up.
So we have deployed Aerospike today in various different configurations, different clusters have been deployed. They are deployed on hard metal. We have deployed on, actually, the virtual clouds. All these locations today, we are loading the profile data into Aerospike as a database.
Aerospike features we are using today are the first and foremost it’s the database on the edges which is doing the key value store. The second most important thing that we actually look at is that it’s a feature which is the high performance, low latency, high throughput and providing stable characteristics – extremely important. The third one that gets in would be the replication ability across different edges, which is XDR.
With Aerospike, the first thing that started happening was that we can now meet the performance characteristics with others coming from the programmatic stack: all the way to personalization, real time segmentation. Entire spectrum use cases can be met with one single system.
Without Aerospike being in-picture, what we had a situation, we had to continue with that situation was that there were multiple different solutions, which were addressing very specific needs of different business use cases. Managing these many different infrastructures operationally was extremely challenging as well as cost prohibitive.
Common infrastructures really start benefiting in multiple different ways. We can provide a common pragmatic API for everybody, which is developers experience start actually improving. Operationally being one single system, actually, becomes a lot easier to manage. The operational team sizes starts actually dropping. The provisioning and deployment starts becoming easy because it’s just one system to manage.
So all three vectors if I were to look at it starts becoming easier for us with one single system.