Ref the question: ‘Your application has to process a very volatile inconsistent flow of data inbound in order. Which of the following options would be most reliable and cost-effective?’
Reserved instance is suggested as the answer, citing ‘If we solely use a spot fleet, we might be outbid and not have available instances.’.
However, the question does not specify that we have to process all orders within an SLA. Can we not process all orders ‘most reliably and cost effectively’ with a spot fleet?
Ack an order to SQS once processed?
2 Answers
Hi Varun,
Remember that in the exam you can only choose from the offered answers, so practice the process of elimination and matching the answers to that facts presented in the question.
I get what you are saying about no SLA being stated. That is tricky. You will need to look for clues and assume some sort of ‘normal’ behaviour. It may be perfectly acceptable to make customers or clients wait hours or days for a result, but that is not ‘normal’ in this era.
With this question, this is how I would process it :-
The phrase ‘ very volatile inconsistent flow of data inbound in order’ is important. Mentally tick off the AWS recommendedsolutions for volatility. Don’t choose, but do have a list in your head.
The kicker is "most reliable and cost-effective". So you need to grade any viable answers against these criteria.
- SNS to receive inbound and use a single instance. SNS is a notificationservice and not suitable for managing work flow. and is not an AWS recommended solution for this problem (Reject)
– SQS and a dedicated host (singular). SQS is the correct tool, a dedicatedhost will provide control and potentially save money. However it does nothing for responsiveness to volatility, and no resilience with a single host. Viable but poor.
SQS with Spot fleet. SQS is the correct tool. Spot is good for bulk and low priority work. Spot has a high risk of losing all compute. Viable but poor.
SQS with single RI instance. SQS is the correct tool, the RI host saves money and provide some assurance. However it does nothing for responsiveness to volatility, and no resilience with a single host. Viable but poor.
So I have three to choose from.
SQS provides buffering for all.
Spot is not an elastic solution, It is a cost saving mechanism (tick), but it adds risk because you can predict that you will get bumped (negative reliability)
Dedicated hosts does nothing for reliability and the depending on how you do it Dedicated Hosts can be significantly more expensive if you pay for a whole host and only run one VM on it. At best you break even on cost if you fill it to capacity with VMs. (on avg. more expensive)
RIs provide both cost saving and assurance of havingresource availability (a basis of reliability if you are considering a fail and replace strategy for resilient design ).
So I would go tend to go with the RI answer. Like may AWS exam questions, AWS will not give you the ideal green field answer (Autoscaling or Lambda worker nodes across multiple AZs). Instead they will ask you to demonstrate you understanding by choosing the best of a group of sub-optimal answers.
Rusty
Moderator & Coach
Hi Varun,
Remember that in the exam you can only choose from the offered answers, so practice the process of elimination and matching the answers to that facts presented in the question.
I get what you are saying about no SLA being stated. That is tricky. You will need to look for clues and assume some sort of ‘normal’ behaviour. It may be perfectly acceptable to make customers or clients wait hours or days for a result, but that is not ‘normal’ in this era.
With this question, this is how I would process it :-
The phrase ‘ very volatile inconsistent flow of data inbound in order’ is important. Mentally tick off the AWS recommendedsolutions for volatility. Don’t choose, but do have a list in your head.
The kicker is "most reliable and cost-effective". So you need to grade any viable answers against these criteria.
- SNS to receive inbound and use a single instance. SNS is a notificationservice and not suitable for managing work flow. and is not an AWS recommended solution for this problem (Reject)
– SQS and a dedicated host (singular). SQS is the correct tool, a dedicatedhost will provide control and potentially save money. However it does nothing for responsiveness to volatility, and no resilience with a single host. Viable but poor.
SQS with Spot fleet. SQS is the correct tool. Spot is good for bulk and low priority work. Spot has a high risk of losing all compute. Viable but poor.
SQS with single RI instance. SQS is the correct tool, the RI host saves money and provide some assurance. However it does nothing for responsiveness to volatility, and no resilience with a single host. Viable but poor.
So I have three to choose from.
SQS provides buffering for all.
Spot is not an elastic solution, It is a cost saving mechanism (tick), but it adds risk because you can predict that you will get bumped (negative reliability)
Dedicated hosts does nothing for reliability and the depending on how you do it Dedicated Hosts can be significantly more expensive if you pay for a whole host and only run one VM on it. At best you break even on cost if you fill it to capacity with VMs. (on avg. more expensive)
RIs provide both cost saving and assurance of havingresource availability (a basis of reliability if you are considering a fail and replace strategy for resilient design ).
So I would go tend to go with the RI answer. Like may AWS exam questions, AWS will not give you the ideal green field answer (Autoscaling or Lambda worker nodes across multiple AZs). Instead they will ask you to demonstrate you understanding by choosing the best of a group of sub-optimal answers.
Rusty
Moderator & Coach