Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


GCP has this so bad IO?
New on LowEndTalk? Please Register and read our Community Rules.

GCP has this so bad IO?

I tried to boot up a GCP instance for testing, but its IO seems very bad. and it's price is not cheap. how do you think of it?

Basic System Information:

Uptime : 0 days, 0 hours, 54 minutes
Processor : Intel(R) Xeon(R) CPU @ 2.20GHz
CPU cores : 2 @ 2199.998 MHz
AES-NI : ✔ Enabled
VM-x/AMD-V : ❌ Disabled
RAM : 975.7 MiB
Swap : 1024.0 MiB
Disk : 9.7 GiB
Distro : Debian GNU/Linux 11 (bullseye)
Kernel : 5.10.0-18-cloud-amd64

Generating fio test file...

fio Disk Speed Tests (Mixed R/W 50/50):

Block Size 4k (IOPS) 64k (IOPS)
Read 6.03 MB/s (1.5k) 96.26 MB/s (1.5k)
Write 6.02 MB/s (1.5k) 96.77 MB/s (1.5k)
Total 12.06 MB/s (3.0k) 193.03 MB/s (3.0k)
Block Size 512k (IOPS) 1m (IOPS)
------ --- ---- ---- ----
Read 98.67 MB/s (192) 98.18 MB/s (95)
Write 103.91 MB/s (202) 104.72 MB/s (102)
Total 202.59 MB/s (394) 202.91 MB/s (197)

Comments

  • AXYZEAXYZE Member
    edited September 23

    Its slow because you created instance with slow disk, is it even SSD tier? Because for sure it isnt local SSD to which you likely comparing it, but some sort of highly redundant network-attached block storage option. Local SSD has 375GB minimum and you have just 10GB.
    https://cloud.google.com/compute/docs/disks

    Looking at the spect, isnt it always free instance?

  • jarjar Member, Patron Provider
    edited September 23

    The lowend market has set unreasonable expectations of average disk performance. When you use a higher end provider you are not expected to be able to benchmark their cloud instances to the same result as the first day on a new node running an Epyc CPU with NVMe, from an LET provider. They are unnecessary for most jobs, and these clouds expect you to scale your application rather than expect one instance to handle a large job by itself without utilizing other services (ex. DBaaS, etc).

    Generally if you find yourself running a full stack in a single server instance, these clouds are not an economically sound choice.

  • @AXYZE said:
    Its slow because you created instance with slow disk, is it even SSD tier? Because for sure it isnt local SSD to which you likely comparing it, but some sort of highly redundant network-attached block storage option. Local SSD has 375GB minimum and you have just 10GB.
    https://cloud.google.com/compute/docs/disks

    Looking at the spect, isnt it always free instance?

    not free. it's 9.7 USD/month.

  • I confirm the basic VM of gcp has worse performance.

    I installed these packages which take few hours to finish.

    sudo apt-get update && sudo apt-get --only-upgrade install google-cloud-sdk-cbt google-cloud-sdk-kpt google-cloud-sdk-minikube google-cloud-sdk-terraform-tools google-cloud-sdk-app-engine-java google-cloud-sdk google-cloud-sdk-config-connector google-cloud-sdk-datalab google-cloud-sdk-cloud-run-proxy google-cloud-sdk-app-engine-grpc google-cloud-sdk-nomos google-cloud-sdk-gke-gcloud-auth-plugin google-cloud-sdk-spanner-emulator google-cloud-sdk-anthos-auth google-cloud-sdk-app-engine-go google-cloud-sdk-cloud-build-local google-cloud-sdk-bigtable-emulator google-cloud-sdk-package-go-module google-cloud-sdk-firestore-emulator google-cloud-sdk-local-extract google-cloud-sdk-app-engine-python google-cloud-sdk-datastore-emulator google-cloud-sdk-pubsub-emulator kubectl google-cloud-sdk-harbourbridge google-cloud-sdk-kubectl-oidc google-cloud-sdk-skaffold google-cloud-sdk-app-engine-python-extras

  • What kind of VM and which disk? Balanced? Standard? There are several types of VMs and disks you have to choose during configuration.
    If you chose the cheapest one (maybe also Spot), then do not expect high IO.

  • NeoonNeoon Member, Community Contributor

    @frakass said:
    I confirm the basic VM of gcp has worse performance.

    I installed these packages which take few hours to finish.

    sudo apt-get update && sudo apt-get --only-upgrade install google-cloud-sdk-cbt google-cloud-sdk-kpt google-cloud-sdk-minikube google-cloud-sdk-terraform-tools google-cloud-sdk-app-engine-java google-cloud-sdk google-cloud-sdk-config-connector google-cloud-sdk-datalab google-cloud-sdk-cloud-run-proxy google-cloud-sdk-app-engine-grpc google-cloud-sdk-nomos google-cloud-sdk-gke-gcloud-auth-plugin google-cloud-sdk-spanner-emulator google-cloud-sdk-anthos-auth google-cloud-sdk-app-engine-go google-cloud-sdk-cloud-build-local google-cloud-sdk-bigtable-emulator google-cloud-sdk-package-go-module google-cloud-sdk-firestore-emulator google-cloud-sdk-local-extract google-cloud-sdk-app-engine-python google-cloud-sdk-datastore-emulator google-cloud-sdk-pubsub-emulator kubectl google-cloud-sdk-harbourbridge google-cloud-sdk-kubectl-oidc google-cloud-sdk-skaffold google-cloud-sdk-app-engine-python-extras

    not as bad when I got a LXC container and it took 50 minutes for mariadb-server, the support told me this is normal.

Sign In or Register to comment.