how to make real-time data ingestion better for tons of events tho

hey everyone im workin on a system that needs to ingest a massive amount of real time event data think millions of events per second were hitting some performance bottlenecks with our current approach im wondering what strategies or best practices others have found successful for optimizing real time data ingestion pipelines under such high load are there specific architectural patterns technologies or tuning techniques that are particularly effective lookin forward to hearin your experiences