NeuroJSON infrastructure

We have build dedicated resources and infrastructure for promoting FAIR neuroimaging datasets at NeuroJSON, and provide these to the broad research community at no cost. These key resources include

Dedicated server system

Hosted in the Northeastern University server room, our team maintains a dedicated rack system with high-performance servers for data sharing, storage, processing/conversion, running database servers among other support. Our server system provides high bandwidth Internet connection, data storage and high-performance compute capabilities.

Each server is equipped with 64-core CPU (128 threads), 265 GB memory, NVIDIA GPUs, and 10GBE Ethernet connection. The entire server system has 20 GBE fiber-optics uplink, permitting high-speed unlimited web and database user connections across the globe.

Between 2024 and 2025, our server has achieved five-nine high-availability (99.999% up-time), our servers are stable and dependable.

Highly scalable NoSQL database with automatic replication

Apache CouchDB is the world's second most popular document-store (NoSQL) database, similar to MongoDB, but is fully open-source. We host dedicated CouchDB databases over our high-performance system, offering fast, and fine-grained non-structured data query, transformation and manipulation functionalities.

A notable feature of CouchDB is the automatic database replication. We maintain multiple CouchDB servers within our network, and automatically duplicate selected datasets (or a selected subset) between these mirrored sites. The replication feature servers not only can serve as automatic data backup, but also can offer locality for fast data access (for example, adding new replication servers in Europe or Asia to speed up data access in these areas). Dataset replication can be individually configured, permitting fine-grained and locality-dependent data access control (such as federated database access).

Commercially hosted high-availability web servers

NeuroJSON.org, NeuroJSON Wiki, cross-database search, among other core functionalities provided by NeuroJSON platform, are hosted on commercial high-availability web service.

Large and redundant storage space

NeuroJSON project transforms complex neuroimaging datasets into lightweight, searchable metadata, stored inside the CouchDB database, and keeps all non-searchable, large sized data files as externally linked data files or data attachments stored on our dedicated storage space.

Our system currently supports around 200 TB data storage space, with redundancy (that can tolerate break-down of data drives), but it can easily grow to accommodate larger space if demands arise.