Back to home
Platform architecture

Engineering principles of the platform architecture

Fault-tolerant architecture

Continuous operation with no single point of failure

Fault tolerance

  • Active node redundancy
  • Active channel redundancy
  • Coherent in-memory cache
  • Graceful degradation

Reliability

  • Configuration backup
  • Recovery on schedule and on event
  • Graceful degradation under load
  • Isolation of failed components

Availability

  • Symmetric Active-Active cluster
  • Hot standby (0-RTO)
  • Coherent in-memory cache
  • Up to 10,000,000 tags per node

Node architecture

Real-time core, I/O load

RT coreInternal brokersCoupling brokersNon-volatile DBEvent brokerService brokerDriver brokerClient brokerCluster brokerReal-time DBMetadataArchive

Node and channel redundancy

Redundant interconnect. Each client is connected to the I/O broker of both nodes

Client interconnectCluster interconnectCoreCluster brokerCore real-time DBCoupling brokersCoreCluster brokerCore real-time DBCoupling brokersC1C2C3C4C5C6

Graceful degradation

When the I/O load exceeds the interconnect limit, the system degrades by dropping the least important messages. If needed, messages can be stored and re-synchronized once the load decreases.

I/O (IO)
limit · 100150 GbE0
Broker queue
Interconnect (IC)
limit · 100150 GbE0
Interconnect queue
queued synchronized dropped (low priority)priority: 1 — highest · 5 — lowest