amazon web services - Does it make sense to oversize ebs gp2 volumes for better iops? -


i have server with:

12 x (io1) volumes. each 1024gb capacity , 4000 provisioned iops.

this server ends costing on $5000 month.

i cut costs bit , replace io1's regular gp2's, gp2's have

3 iops/gb,

so 1024gb gp2 has 3072 iops. keep above 4000.

i not want more capacity, if increase gp2's size 1.5 terabytes, leave half terabyte un-allocated, can iops on each volume 4608.

i pay less, , have overall better performance.

assuming dropping down 99% yearly-performance gp2 %99.9 io1 acceptable:

is there reason not work? there other drawbacks?

yes makes sense oversize volume increase iops. common thing do. leaving space unallocated in os have no bearing on iops allocated ebs. although allocate in os , have space available future growth.

the drawback have lower max throughput per volume gp2 (160 mib/s) vs io1 (320 mib/s) seen in the table here.


Comments