Oracle · Software · YoctoIT tech page

RAC

The active-active cluster: multiple nodes, a single database — continuous availability for workloads that allow no pauses.

FOCUS · TRUE ACTIVE-ACTIVEThe node falls, the sessions continue: high availability as architecture
Request the Free Assessment Learn more
YoctoIT material for clients and partners · Oracle, Exadata, GoldenGate and the other products mentioned are trademarks of Oracle Corporation and/or its affiliates.
01 · What it is

Real Application Clusters, made clear.

RAC runs the same database on multiple nodes at once: a server falls, the others keep serving — with Application Continuity in-flight transactions are stitched back and the user doesn't notice. It's also scalability: more nodes, more power, same data.

Attivo-attivo
all nodes work: no standby sleeping
AC
Application Continuity: the transaction survives the node failure
+Nodi
power grows by adding servers, without migrating anything
Real Application Clusters
OFFICIAL ORACLE BRANDING · RAC
ARCHITETTURA UFFICIALE · ORACLE RAC · FONTE: ORACLE DOCS
OFFICIAL ARCHITECTURE · ORACLE RAC · SOURCE: ORACLE DOCS
02 · How to use it well

The things that make the difference.

L'architettura

Applications (via SCAN)a single name, N nodes behind it
Node 1
Node 2
Node N
all active on the same DB
Grid Infrastructure & ASMcluster and shared storage
Exadata or a certified clusterla base fisica
One database, many hearts

Quorum design

Private networks, voting disks and fencing: the solid cluster is designed in the boring details.

RAC-aware apps

Connection pools, FAN and retries: the application must be trained to exploit the continuity.

Patch rolling

One node at a time, service always up: maintenance without a window.

With Data Guard

RAC for the local failure + standby for the disaster: the complete pair.

03 · In depth

Cache Fusion and services: how it really scales

RAC runs the same database on multiple active nodes: Cache Fusion transfers blocks between instances over a private interconnect (today RDMA on Exadata), Grid Infrastructure manages cluster and voting, services route connections (per app, per workload) with transparent failover (TAF/Application Continuity). Patches become rolling: one node at a time, the database doesn't stop.

  • Cache Fusion — blocks shared via memory across nodes: OLTP's horizontal scale
  • Servizi — each app on its own node set: the load governed, the blast radius reduced
  • Rolling patch — Grid and DB updated node by node: maintenance without downtime
  • TAF/AC — sessions migrating on their own at node failover
  • SCAN — a single name for the cluster: clients don't know the nodes
  • Interconnect — the critical private network: on Exadata it's RDMA, elsewhere it must be designed
04 · Numbers and lifecycle

The numbers that matter.

0
downtime for rolling patches and node loss
2-8
the typical nodes (beyond that, you think in services)
<5 s
session recovery with Application Continuity
24/7
the workloads that justify it: mission-critical OLTP
RAC is powerful and stern: interconnect, services and rolling patching must be done well — we do them for a living.
05 · Use cases

Where it really pays off.

Core banking, ticketing, logistics

Where every minute down makes the front page.

Demanding consolidation

Many critical services on a governed cluster.

Vertical scalability exhausted

When the biggest server isn't enough anymore.

Slide-deck HA and production HA are different sports: we play the second one, and have for years.