История изменений

Для этого должна быть БД для начала. БД нет. Тупо нечего синхронизировать этими всеми недо-HA.

Можно примеры приложений, которые ты крутишь на своём ремусе?

Этот термин появился позже, в середине-конце 1990-х годов, как развитие концепции высокой доступности. Потому что всякие говнюки стали использовать термин HA в маркетинговых целях там, где availability оказывалась не такой уж high.

Early high availability systems from the 1980s and 1990s typically had failover times measured in minutes rather than seconds.

Here are some typical ranges from that era:

Mainframe Systems (1980s-early 1990s):

• IBM mainframes with clustering: 2-10 minutes
• Tandem NonStop systems: 30 seconds to 2 minutes (these were among the fastest)

Unix Clusters (late 1980s-1990s):

• Early Sun Cluster: 3-8 minutes
• HP ServiceGuard: 2-5 minutes
• IBM HACMP: 3-10 minutes

Database Systems:

• Oracle Parallel Server failover: 5-15 minutes
• Early Sybase clusters: 3-8 minutes

The long failover times were due to several factors:

• Detection mechanisms relied on heartbeat timeouts (often 30-60 seconds)
• Resource cleanup and state verification took time
• Storage systems needed to be unmounted/remounted safely
• Network reconfiguration was slower
• Applications had to restart completely rather than just reconnecting

Tandem’s NonStop systems were notable exceptions, achieving sub-minute failovers because they were designed from the ground up for continuous operation with process pairs and shared-nothing architecture.

Modern systems typically achieve failover in seconds or even milliseconds, a dramatic improvement from these early implementations.

Ну примерно как сейчас используется термин «ИИ» для тупорылых нейросеточек где никаким интеллектом вообще не пахнет.

PS: Сравни себя и LLM :)

Для этого должна быть БД для начала. БД нет. Тупо нечего синхронизировать этими всеми недо-HA.

Можно примеры приложений, которые ты крутишь на своём ремусе?

Этот термин появился позже, в середине-конце 1990-х годов, как развитие концепции высокой доступности. Потому что всякие говнюки стали использовать термин HA в маркетинговых целях там, где availability оказывалась не такой уж high.

Ну примерно как сейчас используется термин «ИИ» для тупорылых нейросеточек где никаким интеллектом вообще не пахнет.

Early high availability systems from the 1980s and 1990s typically had failover times measured in minutes rather than seconds.

Here are some typical ranges from that era:

Mainframe Systems (1980s-early 1990s):

• IBM mainframes with clustering: 2-10 minutes
• Tandem NonStop systems: 30 seconds to 2 minutes (these were among the fastest)

Unix Clusters (late 1980s-1990s):

• Early Sun Cluster: 3-8 minutes
• HP ServiceGuard: 2-5 minutes
• IBM HACMP: 3-10 minutes

Database Systems:

• Oracle Parallel Server failover: 5-15 minutes
• Early Sybase clusters: 3-8 minutes

The long failover times were due to several factors:

• Detection mechanisms relied on heartbeat timeouts (often 30-60 seconds)
• Resource cleanup and state verification took time
• Storage systems needed to be unmounted/remounted safely
• Network reconfiguration was slower
• Applications had to restart completely rather than just reconnecting

Tandem’s NonStop systems were notable exceptions, achieving sub-minute failovers because they were designed from the ground up for continuous operation with process pairs and shared-nothing architecture.

Modern systems typically achieve failover in seconds or even milliseconds, a dramatic improvement from these early implementations.

PS: Сравни себя и LLM :)