A company is deploying a machine learning (ML) research project that will require a lot of compute power over several months. The ML processing jobs do not need to run at specific times.
Which Amazon EC2 instance purchasing option will meet these requirements at the lowest cost?
For machine learning processing jobs that do not need to run at specific times and require a lot of compute power over several months, Spot Instances provide the lowest cost option. Spot Instances allow you to use spare EC2 capacity at a significantly reduced price compared to On-Demand Instances. Since the jobs can be interrupted and do not have specific timing requirements, the inherent interruptions with Spot Instances are not an issue, making them the most cost-effective solution.
Alright guys, I found this. Pretty sure it settles things. https://naterar.medium.com/a-deep-dive-into-aws-instance-types-solving-high-cpu-workload-and-memory-spikes-57e8b5623777 Spot: Spot instances can be valuable for handling sporadic high CPU workloads and memory spikes. >>>However, the inherent uncertainty of availability makes them less reliable for consistent, high-demand workloads.<<< On-Demand: provide the necessary computational horsepower exactly when needed, making them ideal for dealing with unexpected CPU and memory demands. So, yeah, I'd say 'A'.
wrong B.
Correct answer should be B. On-Demand Instances: These instances allow you to pay for compute capacity by the hour or second, with no long-term commitments. They are suitable for short-term, spiky, or unpredictable workloads that cannot be interrupted. However, they are not the lowest cost option for long-term, predictable workloads. Spot Instances: This is the correct choice. Spot Instances allow you to bid on spare Amazon EC2 computing capacity. Spot Instances are available at up to a 90% discount compared to On-Demand prices. They are ideal for workloads that are not time-sensitive and can tolerate interruptions. Since the company's ML processing jobs do not need to run at specific times, Spot Instances offer a cost-effective solution for long-term, predictable workloads.
A. On-Demand Instances: On-Demand Instances provide the highest level of flexibility and do not require any upfront commitment. On-Demand Instances are suitable for workloads with variable usage patterns, but they are generally more expensive than other purchasing options. B. Spot Instances: Spot Instances are the most cost-effective option for scenarios where the workload is flexible and can be interrupted. Spot Instances allow you to use spare EC2 capacity at a significantly lower cost than On-Demand Instances. Spot Instances are suitable for workloads like batch processing, data analysis, and machine learning jobs that do not require continuous, uninterrupted operation C. Reserved Instances: Reserved Instances involve a commitment to a specific instance type in a particular region for a term of 1 or 3 years. In exchange for this commitment, users receive a significant discount compared to On-Demand Instances. D. Dedicated Instances: Dedicated Instances are instances that run on hardware dedicated to a single customer. They are typically used in scenarios where compliance or regulatory requirements mandate dedicated hardware.
B. Spot Instances Spot Instances allow you to bid for unused EC2 capacity, often resulting in significantly lower costs compared to On-Demand Instances. Since the ML processing jobs do not need to run at specific times, Spot Instances can be a suitable choice as they provide access to spare EC2 capacity at reduced prices.
https://aws.amazon.com/ec2/spot/?refid=32f4fbd0-ffda-4695-a60c-8857fab7d0dd Use cases : Deploy containers Run containerized workloads of any scale using Amazon EC2, Amazon ECS, Amazon EKS, or self-managed Kubernetes clusters on Spot Instances. Run big data and ML workloads : Fast-track big data and machine learning workloads by using Spot Instances with Amazon EMR, Hadoop, Spark, or Amazon Sagemaker to process massive amounts of data.
B is the best solution for lowest costs
B Spot instances https://aws.amazon.com/ec2/instance-types/p3/
B Spot instances https://aws.amazon.com/ec2/instance-types/p3/
B. - On-demand is not suit for run over the long period because it's not cost optimize - any reserved instance is not suite for this use case too, because it reserved model is need at least 1 year commitment.
Fast-track big data and machine learning workloads by using Spot Instances with Amazon EMR, Hadoop, Spark, or Amazon Sagemaker to process massive amounts of data. See use cases: https://aws.amazon.com/ec2/spot/?gclid=EAIaIQobChMIq8X2me3ogwMV6qtmAh26FgjUEAAYASAAEgKWyfD_BwE&cards.sort-by=item.additionalFields.startDateTime&cards.sort-order=asc&trk=a9b30b20-d23f-4d61-9452-c51a7e407fcd&sc_channel=ps&ef_id=EAIaIQobChMIq8X2me3ogwMV6qtmAh26FgjUEAAYASAAEgKWyfD_BwE:G:s&s_kwcid=AL!4422!3!517651795606!p!!g!!amazon%20ec2%20spot!12876354690!124026471080
B = CORRECT
answer is B
The AWS machine Learning service's public website said that ML frameworks based on "on-demand" in EC2 part, which is very wired that No where said ML must use on-demand to run or whether it can use spot insdance. But I think A is more likely to be right, because ML public side do not mention "spot" in Ec2 parts.
Will require a lot of compute power over several months.The ML processing jobs do not need to run at specific times. Can spot handle large scale computing?
the question doesn't explicitly state whether the workload can be interrupted or not. In such cases, it's always better to err on the side of caution and consider options that are less likely to cause disruptions. While Spot Instances can be incredibly cost-effective, if the workload cannot tolerate any interruptions whatsoever, then choosing a more reliable option like On-Demand
it SPECIFICALLY states that it does not need to run at any specific times. that means it can be interrupted at any point.
aaaaaaaaaaaaaaaaaaaaaaa
bbbbbb