- Datanet enables new genres of use cases (some of which are amazing):
- - Global Distributed Data Cache for CDNs, Real-time user-data at the CDN edge
- - Stateful application-clusters, Stateful serverless: cache data at compute
- - Multi-Master multi-DataCenter pluggable data-store, Multi-DC consistent Memcache-Clusters
- - Fog Computing: compute concurrently on browser/mobile/IoT
In a Content Delivery Network, Datanet acts as an Agent on each server in every CDN Point-of-Presense. Agents can write monitoring data (e.g. server-health-checks, customer-statistics, client-statistics) or even host real-time dynamic client-data at the CDN edge (think [email protected]) and Datanet takes care of replicating the Data to other PoPs. Plus data is replicated between nearby CDN PoPs (e.g. USA1 & USA2) peer-to-peer with optimal latency. Datanet adds a shared distributed brain to CDNs
Datanet makes Highly Available Stateful AppServer Clusters a reality. AppServers use Raft Quorom to form consensus on cluster membership and can then route similar requests to sticky app-servers (for instance requests for user=Bob always go to app-server3). These sticky-requests then use Datanet to cache data locally and are processed w/ 100% local I/O which means they’re really fast. Datanet handles the conflicts that happen in such a setup, as well as the high availability for state on node-addition & node-failure.
Datanet can front consistent memcache clusters in multiple DataCenters. A problem that often occurs in multiple DataCenter ‘homed’ applications is a user flaps between DataCenters on subsequent requests and gets different data from different DataCenter’s memcache-clusters. Datanet solves this issue by insuring consistency between the 2 DataCenters Memcache-clusters ... and Databases (everything is consistent). Datanet's master-less architecture changes the game for MultiDC homed apps.