Fusion Edge Server
Secure, high-performance statistics dissemination at scale
Fusion Edge Server works in tandem with Fusion Registry to deliver secure, high-performance and scalable statistical data dissemination services.
Edge Servers are designed to be deployed in scalable fault-tolerant clusters in the DMZ or on a Content Delivery Network providing data consumers with fast and reliable access to your published datasets.
The Edge Servers provide an SDMX compliant REST data API which can drive a range of dissemination services including web data portals, data explorer applications and BI tools.
Publish a snapshot of your data in seconds
Take a snapshot of either your entire Fusion Registry data warehouse or just selected datasets, and publish in seconds by pushing to your Fusion Edge Server cluster.
To refresh your data - just take another snapshot.
The process is simple so you can choose the refresh frequency to meet your business requirements - weekly, daily, hourly.
Fusion Edge Server is designed to deliver the fast query response times expected of modern data services.
Tests on datasets with 30 million observations show a cluster of four Fusion Edge Servers is able to respond to queries in under 1 second, even with 200 concurrent users.
In practice, end-users can expect typical query response times in the order of 50ms to 100ms.
If supported by a user's client application, the Fusion Edge Servers will also compress data over the wire further reducing the latency between the query being issued and the data being received.
Fusion Edge Servers can be deployed in clusters behind a normal web traffic load-balancer such as BIG-IP, Cloudflare or Amazon's Elastic Load Balancer service.
In practice, the size of a cluster is limited by the infrastructure and load-balancer capabilities, but deployments of two to six nodes are typical.
The Fusion Edge Server application has a small footprint and minimal dependencies making it easy to scale the size of a cluster in response to demand. The only real requirement is a persistent file system which is used to store the published content. Fusion Edge Servers are stateless, and once primed with data operate completely autonomously.
Each Server has its own embedded in-memory database so no shared database service is required which avoids single points of failure and potential performance bottle-necks.
Fusion Edge Servers require no connection back to the master Fusion Registry or any other private services such as databases or authentication services.
Updates are published by using the master Fusion Registry to geneate a ZIP file containing all of the required data and metadata which is 'pushed' to the Edge Server cluster. Even if the cluster is compromised, the approach ensures that the dissemination services can never be used as a gateway to attack private networks, or gain access to restricted services or information.
Edge Servers are also designed to be inherently resistant to attack. In particular, the proprietary in-memory database is read only and does not use SQL eliminating the possibility of SQL-injection - a common attack method.
The Fusion Edge Server dissemination system now supports embargo.
A precise publication date and time can be set when uploading content updates. Each Edge Server in the cluster loads the Embargoed content into a second in-memory database which is instantaneously switched-in at the set time. That means, for very time-sensitive publication use cases, data consumers will be able to gain access to the updated content within milliseconds.