I was responsible for investigating m7* machines for the main cluster at my last job, and elected to keep our largest services on m6i and c6i because they fiddled with the prices and for our applications the new machines didn’t perform enough better to offset the price hike. With autoscaling there would be times of the day where we saved a half a machine, and that would cost us less, but it just wasn’t worth the migration effort. The AMD migration shockingly would have cost us more, entirely due to the price hike being out of proportion to other generations.
They used to keep the prices flat, so there was always an incentive to move off the old hardware if there weren’t compatibility problems. But that changed with m7 and seems to have continued with m8.
We went from 0.096 to 0.1008 to 0.10584, for the smallest instance, which is a 5% bump.
I was responsible for investigating m7* machines for the main cluster at my last job, and elected to keep our largest services on m6i and c6i because they fiddled with the prices and for our applications the new machines didn’t perform enough better to offset the price hike. With autoscaling there would be times of the day where we saved a half a machine, and that would cost us less, but it just wasn’t worth the migration effort. The AMD migration shockingly would have cost us more, entirely due to the price hike being out of proportion to other generations.
They used to keep the prices flat, so there was always an incentive to move off the old hardware if there weren’t compatibility problems. But that changed with m7 and seems to have continued with m8.
We went from 0.096 to 0.1008 to 0.10584, for the smallest instance, which is a 5% bump.