Skip to main content

Overview

The LDB (Linked-list Database) is the database engine at the core of SCANOSS. It is headless, which means it has no user interface or client layer. It is built for single-key, read-optimised access across large volumes of data. It keeps memory and disk use low and makes few kernel-level system calls. It stores data in linked lists. SCANOSS SaaS runs on a distributed infrastructure built around the LDB. Load balancers spread software analysis requests across distributed LDB data nodes, so the service handles high volumes and keeps running when a node fails.

How load balancing works

The load balancer distributes incoming requests across several processing nodes. Each processing node runs the SCANOSS Engine and queries a shared LDB Knowledge Base, a read-optimised store of open source component data. The Engine matches file fingerprints against known open source components. A fingerprint is a cryptographic hash of a source code snippet. This architecture gives you:
  • Even workload across processing nodes.
  • High availability through redundancy.
  • Horizontal scaling to handle more requests.

Example deployment scenario

A typical large enterprise deployment has these layers:
  1. An external request endpoint receives incoming scan requests.
  2. A load balancer distributes requests across processing nodes.
  3. Processing nodes each run the SCANOSS Engine.
  4. The LDB Knowledge Base is the shared, read-optimised data layer that all nodes query.

Architecture diagram

Reliability and scalability properties

The architecture is built for:
  • Fault tolerance. Redundant processing nodes remove single points of failure.
  • Consistent query performance. The LDB layer is read-optimised.
  • High availability. The load balancer keeps the service running when a node fails.
  • Horizontal scaling. You can add processing nodes without downtime.

LDB and Engine version compatibility

LDB and the SCANOSS Engine ship as two parallel release lines. You must upgrade them together.
  • The traditional line is LDB 4.x and Engine 5.x (for example, tag v5.5.1).
  • The CRC64-compatible line is LDB 5.x and Engine 6.x and later. Its tags carry a mandatory -crc64 suffix (for example, v5.0.0-crc64, v6.0.0-crc64).
A CRC64-line LDB requires an Engine from the CRC64 line, and the reverse is also true. An Engine from before 6.0.0-crc64 cannot read tables built by LDB 5.x-crc64 and doesn’t link against its library API. The split covers the on-disk table layout and the library API. It doesn’t decide which hash a table uses, so a CRC64-line deployment can still run MD5-hashed tables. At startup, Engine compares the LDB version it was built against with the LDB version it runs against. On a mismatch it refuses to start, so it can’t fail in unpredictable ways later. ldb -v, scanoss -v, and the JSON server.version field all report which line a binary belongs to. Check them when you plan an upgrade.