Application Scenarios

When deploying online and offline services together, you can use the MPAM feature to limit resources for offline services while ensuring the performance of online services.

• MPAM addresses the requirement for prioritizing response to online services.
  • Container cloud customers generally deploy online services for real-time loads and offline services for non-real-time loads (for example offline settlement), and they want that response to online services must be ensured at the same time offline services are running.
  • In most scenarios, offline service containers are created periodically and destroyed after tasks are executed. They do not have fixed locations and may be deployed together with any service.
• Offline services compete with online services for L3 cache and memory bandwidth. As a result, the response to online services may be delayed.

  Customer's container scheduling platforms try to ensure that both online and offline services are allocated sufficient CPU and memory resources. However, offline and online services deployed on the same node contend for L3 cache and memory bandwidth, and the performance online services may deteriorate.

  

  In hybrid deployment environments, online service performance may be impacted by multiple factors. Since MPAM isolation specifically addresses the memory subsystem, it proves particularly effective for online services that are sensitive to memory bandwidth. However, it is crucial to maintain offline service CPU utilization at moderate levels, ideally between 10% to 20%, as excessive utilization could significantly disrupt online service performance.