Closed mbookman closed 8 years ago
Looks nice, but just wondering, why have have in the Readme files a persistent disk (PERSISTENT_HDD
) and not a scratch one (LOCAL_SSD
) if it will be deleted?
LGTM
Hi Paul - Local SSD is much more expensive than HDD-backed persistent disks. See: https://cloud.google.com/compute/docs/disks/ for an overview of block storage on Compute Engine. See: https://cloud.google.com/compute/pricing#disk for pricing details.
Hi Matt,
Wow, didn't realize Local SSD was so expensive! So compared to persistent, then the maximum local monthly throughput benefit per GB/sec would be greater by 604.4x reads (453.3/0.75), and 160x writes (240/1.5) for an additional $0.178 ($0.218-$0.04).
I guess based on the first link, when it is mentioned that locally is the most economical option in terms of throughput would be because of this:
Local Provisioned Space (NVMe) | Persistent Provisioned Space (HDD) | Local/Persistent (rounded to the tens) |
---|---|---|
Reads/$ = 0.75/$0.04 = 18.75 | 453.3/$0.218 = 2079.358 | 2079.358/18.75 ≈ 110x |
Writes/$ = 1.5/$0.04 = 37.5 | 240/$0.218 = 1100.917 | 1100.917/37.5 ≈ 30x |
Thanks, Paul
pipeline create or run but not both. Updated compress, fastqc, and set_vcf_sample_id examples.