Runtimes¶

What are runtimes?¶

In Wherobots, a runtime provides serverless computational horsepower
for your workloads(1).

Workloads are the Wherobots Job Runs, queries, notebooks, sessions, machine learning models, or artificial intelligence capabilities that can run on Wherobots and interact with customer data.

A runtime is a dedicated, distributed, computing cluster that runs on cloud-based virtual machines that you don’t need to manage.

The amount of concurrent horsepower available to your workloads on a runtime, is measured in terms of Spatial Units. Customers pay for their use of Wherobots software and Wherobots-managed hardware through a Spatial Unit.

What runtimes are available?¶

You can choose from a range of runtime sizes. This allows you to achieve the optimal balance of performance, scale, and cost.

As the size of the runtime increases, the computational horsepower increases with it, allowing you to do more work over time. The rate at which Spatial Units are consumed depends on both the runtime's size and type.

Runtime types¶

Wherobots’ runtimes are classified in one of the following ways:

General Purpose
Memory Optimized
GPU Optimized

All of these options are suitable for production workloads.

Can I use any runtime?

Your Organization Edition might not have access to all of the runtimes listed in the Runtime specifications chart. For example, Community Edition Organizations can only use the Tiny runtime. You can see the runtimes available to your Organization within the Start a Notebook dropdown in Wherobots Cloud.

To assist in selecting the appropriate runtime, consider the following guidelines:

Use case	Runtime type	Guideline
Getting started, most use cases	General Purpose	Whether you're Getting Started or running planet-scale workloads, start with the smallest General Purpose runtime and increase its size if you require more performance.
Running memory-intensive workloads	Memory Optimized	If you’re dealing with memory-intensive workloads, like in map matching, for example, consider using a memory optimized runtime.
Performing Raster inference	GPU Optimized	If you're using WherobotsAI Raster Inference, start with the smallest GPU Optimized runtime and then scale up in size if you need additional performance. You must file a Compute Request in order to access a GPU Optimized runtime.

Runtime time-to-live (TTL)¶

The Runtime TTL (Time-to-Live) refers to the duration a runtime instance can remain active in Wherobots Cloud. Professional and Enterprise Edition Organizations have a 24-hour default TTL, while Community Edition Organizations have a 4-hour default TTL.

Find Your Organization's TTL

Professional and Enterprise Edition Organizations have a 24-hour default TTL, while Community Edition Organizations have a 4-hour default TTL.

If Wherobots granted you a compute increase, your TTL may differ. To check your Organization's current TTL, start a runtime in Wherobots Cloud and look at the Expires In section of Start a Notebook.

When should I increase the runtime size?¶

Changing the size of the runtime allows you to control how fast work needs to be done. For example, if a General Purpose Tiny runtime isn’t delivering results as fast as you want them delivered, increase the size to a General Purpose Small or General Purpose Medium, etc. If a workload becomes constricted by memory, increase the runtime size or utilize a memory optimized runtime.

How is runtime usage measured?¶

Runtime usage is measured in terms of Spatial Units.

Note

To get additional information on our pricing, visit Wherobots Pricing.

What are Spatial Units?¶

Spatial Units describe the amount of computational horsepower provisioned to a runtime.

A single Spatial Unit gives you spatial computational performance that's similar to a 32vCPU Apache Sedona cluster running on the latest stable Apache Spark runtime.

Runtimes that consume Spatial Units at a higher rate than others are capable of doing more work, faster.

What factors influence Spatial Unit consumption?¶

Starting a SparkContext requests the maximum compute resources that you've been allocated based on your runtime selection. Since Sedona runs on top of Spark, initializing a SedonaContext simultaneously creates a SparkContext.

To scale down compute usage, you have two options:

Stop the SedonaContext: Call sedona.stop() to release the resources held by Sedona. This will stop the active SparkContext associated with the SedonaContext.
Destroy the notebook instance: If you're done with your analysis and want to release all compute resources, including those used by the SparkContext, you can destroy the notebook instance.

Choose the option that best suits your needs and workflow.

Usage-based billing¶

You're only charged for the compute resources you actually use on Wherobots.

For example, if you request 1000 executors but only 995 are available, you'll only be billed for the 995 executors that were actively used.

This ensures that you only pay for the resources consumed, regardless of initial requests or fluctuations in compute availability.

AWS Marketplace Billing and Spatial Units¶

In the AWS Marketplace, you're billed using "Wherobots Consumption Units" (tenths of a cent), which are based on "Spatial Units." Spatial Units reflect the runtime's work capacity. More powerful runtimes use more Spatial Units, which can result in a higher bill.

How many Spatial Units can a runtime consume per hour?¶

The following table describes the makeup of each runtime and how many Spatial Units are consumed per runtime per hour.

Runtime specifications chart¶

Min Spatial Unit per hour is the base serverless capacity for the Jupyter Notebook environment you can launch after starting a runtime. Max Spatial Unit per hour is the peak runtime usage after running a notebook or job. To change performance tiers, switch to a different sized runtime.

Runtime type	Runtime Size	Min Spatial Units per hour	Max Spatial Units per hour
	General Purpose Ideal for most workloads, from getting started to production.	Tiny	1.0	5
		Small	1.0	9
		Medium	2.0	50
		Large	2.0	98
		XL	4.0	196
		2XL	4.0	388
Memory Optimized For memory intensive workloads. High-memory runtimes provide double the memory compared to General Purpose runtimes of the same size.	Medium	2.8	70
	Large	2.8	138
	XL	5.6	275
	2XL	5.6	544
	GPU Optimized GPU backed, ideal for powering computer vision using WherobotsAI Raster Inference.	Tiny	3.0	15
Small		3.0	27
Medium		3.6	90

Community Edition Spatial Unit calculation¶

As an example, let’s determine the rate limit for a Community Edition Organization.

Community Edition Organizations are rate limited to 20 concurrent Spatial Units. This is calculated by multiplying the Tiny runtime's(1) maximum capacity (5 Spatial Units per hour) by the runtime TTL (4 hours)(2).

Community Edition Organizations can only use the Tiny runtime.
Community Edition Organizations have a 4-hour default TTL.

To increase your Community Organization's concurrent usage quota, submit a Compute Request or upgrade your Organization to one of Wherobots’ paid tiers.

What is my Organization's quota?¶

Your Organization's concurrent computing power is dependent on your Organization Edition of Wherobots as well as the payment history associated with your Organization.

To review your quota in Wherobots Cloud, do the following:

Go to Organization Settings.
Scroll to Usage.

Compute requests¶

If you need additional computing power or would like to access runtimes that are unavailable to your Organization, please email support@wherobots.com with the following:

Warning: Increased Compute May Increase Your Bill

Using more compute resources may result in a higher bill due to increased Spatial Unit consumption.

WherobotsAI Raster Inference requires submitting a compute request

WherobotsAI Raster Inference requires a GPU-Optimized runtime.

To access this runtime category, do the following:

Sign up for a paid Wherobots Organization Edition (Professional or Enterprise).
Submit a Compute Request for a GPU-Optimized runtime.

In your email, use the subject line: Compute Request
Your Organization ID.

To find your Organization ID, go to Organization Settings.
Your request. Specify the following for each quota increase request:
- Runtime type: What type of runtime do you need? (General Purpose, Memory Optimized, GPU Optimized)
- Size: What size runtime do you need? (e.g., Small, Medium, Large)
- Number: How many runtimes of this type and size do you need? (e.g., 1, 2, 5)
- Compute hour increase: How many additional compute hours are you requesting?
  - Review the Runtime TTL to understand the default TTL associated with your Organization Edition.
  - To check your current TTL, start a runtime in Wherobots Cloud and look at the Expires In section of Start a Notebook.

Requests are reviewed on a case-by-case basis and you can expect an answer in 1 business day.

Approval is not guaranteed and is dependent on various factors such as your Organization tier, usage patterns, and available resources.

Upgrade your Organization¶

To upgrade from a Community Edition Organization and access Professional or Enterprise Edition default runtimes, explore these options:

To upgrade your existing Wherobots Organization, see Upgrade your Organization.
To create a new Organization, Wherobots offers the following onboarding options:
- Professional Edition: Sign up through AWS Marketplace.
  
  Unable to sign up through the AWS Marketplace?
  
  If you are unable to create an AWS account and subscribe to the Professional Edition through the AWS Marketplace, contact us at support@wherobots.com.
- Enterprise Edition: Contact us to discuss a custom contract.

Does Wherobots support per second billing?¶

Yes, Wherobots charges for fractional Spatial Unit usage for runtimes in our paid Organization tiers.

Wherobots charges for the exact amount of runtime usage. So, if you only use a runtime for 13.4 minutes, you'll be charged for just those 13.4 minutes, rather than rounded up. You only pay for what you use.

Runtime `enum` values¶

You can use the enum values in this table when interacting Wherobots' APIs to specify the runtime environment for your notebook instances. This allows precise control over resource allocation when automating tasks or integrating with other tools.

Runtime type	Runtime	Enum Value
	General Purpose	Tiny	`Runtime.TINY`
		Small	`Runtime.SMALL`
		Medium	`Runtime.MEDIUM`
		Large	`Runtime.LARGE`
		X-Large	`Runtime.X_LARGE`
		XX-Large	`Runtime.XX_LARGE`
		XXXX-Large	`Runtime.XXXX_LARGE`
Memory Optimized	Medium	`Runtime.MEDIUM_HIMEM`
	Large	`Runtime.LARGE_HIMEM`
	X-Large	`Runtime.X_LARGE_HIMEM`
	XX-Large	`Runtime.XX_LARGE_HIMEM`
	XXXX-Large	`Runtime.XXXX_LARGE_HIMEM`
	GPU Optimized	Tiny	`Runtime.TINY_A10_GPU`
Small		`Runtime.SMALL_A10_GPU`
Medium		`Runtime.MEDIUM_A10_GPU`