Our customers are running year end activity and the memory is thrashing. Rarely used data (or data the Buffer Manager does not predict) is being called into RAM.
After PLE drops to 144 or 86 sometimes, it slowly rises, it can get up to 800 or 900 before in drops again to the lows. I have already removed many indexes that were bad. They were bad because they were speeding up 12k queries, but slowing down 2.7 million queries (see real production numbers below).
Total Writes Total Reads Difference
2,754,287 12,404 2,741,883
My production server has 8.3 Gig of RAM and SQL Server is set to 7.5 Gig. I worry that giving SQL Server more RAM will cause the context switching to rocket above 20k.
Can anyone give me a useful suggestion other than buying more RAM for my 5 year old server?