Location: Europe (Tallinn)
SLA: Standard Service Level Agreement is applied for this cluster.

Compute Resources

To request computing resources you need to create a Pod. Pod is a group of one or more containers, which share the same resources, data storage and network. Each time you create a Pod it will be put under your namespace. All of your Pods share local network while being isolated from other users. Read more about namespaces in the official Kubernetes documentation.

Pod can be created in 3 ways:

  • By using Puzl Dashboard. Dashboard provides Docker Images prebuilt by Puzl to simplify your workflow and make it faster to launch your code. Or you can use any Docker image from Docker Hub, GitLab, and GitHub.

    You can configure a fully custom Pod according to your needs or use one of Cloud Marketplace Apps — easy to use predefined development solutions.

  • By using Kubernetes Dashboard. You can securely access your personal Kubernetes Dashboard from your Puzl Dashboard. This way allows you to either manually specify application details, or upload a YAML or JSON configuration. Check out a manual on how to deploy an application in the official Kubernetes documentation.

  • By using Kubernetes API and the Kubernetes command-line tool — kubectl. It's a powerful tool that allows you to deploy applications, manage and inspect resources of your namespace. See how to access your Pod with kubectl and general kubectl overview.

Pricing

Pod price doesn't cover pricing for data storages like NVMe®-based data storage or Shared File System Data Storage, which should be created separately. Volume creation is optional in most cases.

Usage is billed on per-second basis with zero cluster management fee.

Each Pod comes with a minimal amount of RAM and CPU, and 50GB of . Optionally, you can request GPU and a data storage.

As a default Pod comes with Guaranteed Quality of Service, which means that price will be the same for each hour of operation. For RAM or CPU you can specify limits higher than request. In this case a Burstable Quality of Service will be assigned to this Pod and cost can vary depends on resources actually used by Pod. Check for both QoS classes.

Estimate your application costs with with Pod Configurator.

GPU

NVIDIA A100, 40GB VRAM, PCIE

 
$1.60

per hour

GPU

Quarter of NVIDIA A100, 10GB VRAM, PCIE

 
$0.40

per hour

GPU

Half of NVIDIA A100, 20GB VRAM, PCIE

 
$0.80

per hour

RAM1

ECC RAM 2933 MHz or higher

128MB or 8GB per 1 GPU

minimum

$0.002

per GB-hour

vCPU2, 3

1 vCPU is equivalent to one thread of AMD EPYC 2nd Gen core, 2.5GHz or higher

0.1vCPU or 2vCPU per 1 GPU

minimum

$0.015

per vCPU-hour

Container File System

up to 50GB

per container

$0.007

per usedGB-day

Docker Image4
up to 10GB

per container

$0.007

per GB-day

LAN Bandwidth
up to 10Gbit/s
Free

for incoming and outgoing traffic

Internet Bandwidth
up to 1Gbit/s
Free

for incoming and outgoing traffic

1. Requested RAM amount must be a multiple of 1 MB and will be rounded to 1 MB up (when requested via API).

2. Requested vCPU amount must be a multiple of 1 mCPU (0.001 vCPU) and will be rounded to 1 mCPU up (when requested via API).

3. For integer amount of vCPU in combination with Guaranteed QoS exclusive CPU cores will be allocated. Shared CPU cores will be allocated otherwise.

4. You will be billed based on a size of uncompressed Docker image.

If you require resources for a regular usage or a specific hardware configuration, we are happy to offer you an individual long-term contract with the best prices in the EU. Please describe your requirements and contact our team by sales@puzl.cloud.

Pod Quality of Service Classes

Guaranteed

Pod have the same memory limit and a memory request, and Pod have the same vCPU limit and a vCPU request. Learn more about requests and limits.

In this case price is the same for every hour of Pod operation.

For instance, user requests a Pod with the following configuration: 1GPU, 11 GB RAM, 2 vCPU. Pod works for 12 hours, and consumes 2.5 GB of container file system, Docker image takes 8 GB uncompressed.

PriceRequestWork duration, hCost of request
GPU$1.60
per GPU-hour
112$19.20
RAM$0.002
per GB-hour
11 GB12$0.26
vCPU$0.015
per vCPU-hour
2 vCPU12$0.36
PriceUseWork duration, hCost of use
Container File System$0.007
per GB-day
2.5 GB12$0.0088
Docker Image$0.007
per GB-day
8 GB12$0.028
Total: $19.86

Burstable

Pod has memory limit higher than memory request, or Pod has vCPU limit is higher than vCPU request. Limit represents the maximum amount of resources that can be allocated, while request represents the amount of resources that is guaranteed to be allocated in any case. Learn more about requests and limits.

In this case price consists of 2 parts: constant - calculated according to the amount of requested resources, plus flexible - all what exceeds request.

For instance, user requests a Pod with the following configuration: 1 GPU, 11 GB RAM request and 20 GB limit, 2 vCPU request and 5 vCPU limit. Pod works for 12 hours and uses 2.5 GB of container file system, Docker image takes 8 GB uncompressed. For 3 hours of 12 Pod consumes 17 GB RAM and 4 vCPU in average. For the rest 9 hours Pod consumes resources below the request.

PriceLimitRequestWork duration, hCost of requestAvg use over requestDuration of use over request, hCost over request
GPU$1.60
per GPU-hour
112$19.20
RAM$0.002
per GB-hour
20 GB11 GB12$0.266 GB3$0.036
vCPU$0.015
per vCPU-hour
5 vCPU2 vCPU12$0.362 vCPU3$0.09
PriceLimit per containerUseWork duration, hCost of use
Container File System$0.007
per GB-day
5 GB2.5 GB12$0.0088
Docker Image$0.007
per GB-day
10 GB8 GB12$0.028
   Total per request: $19.86Total over request: $0.13
Total: $19.99

Container File System

Container File System is a fast temporary data storage, which can be used for the data needed in your application runtime. Container File System is being created automatically with a Pod and is being killed when its Pod finishes work. Container File System size is limited to 5 GB per each Pod’s container and cannot be increased. You pay only for used capacity, not for the whole 50GB. Check how cost is calculated in the .

We highly recommend against using Container File System for data storage purposes because its lifecycle depends on the Pod: if something goes wrong with the Pod, the Container File System will be terminated with the Pod, and you will lose all your data. You should use Volumes, which are based on persistent data storages, if you need your data to persist after a Pod is deleted.
BenchmarkIOPS1Bandwidth2
random readup to 32kup to 4030MB/s
random writeup to 81kup to 1070MB/s

1. 4KB block size.

2. 4MB block size.

All prices do not include EU VAT. If you're located in the EU:

  1. All prices above are listed in USD, but for EU residents, the prices are the same numbers in EUR.
  2. Puzl Cloud is required to charge VAT on sales of electronic services to non-business customers (B2C) in the EU.
  3. Puzl Cloud is required to charge VAT on sales of electronic services to business customers (B2B) in Estonia.
  4. Puzl Cloud charges local VAT, unless you have a valid Registration Number associated with VAT.
  5. All VAT collected from customers will be reported to the local tax authorities under the EU simplified VAT registration regime known as the Mini One Stop Shop (“MOSS”).