Ninety-eight percent of Google search on ‘AWS reserved instance (RI) benefits’ shows that you can get great discounts and save tremendously compared to on-demand pricing.The fact is, this discounted pricing can be reaped provided you know what RIs are, how to use them, when to buy them, how to optimize them, how to plan them, etc.
Many organizations have successfully put RIs to its best use and have the optimal RI planning and management in place due to the complete knowledge they have.
This overarching, in-depth blog post is a beginner’s guide that helps you leverage RIs completely and correctly, so that you can make that perfect RI planning and management. It also provides information on how to save smartly on AWS cloud.
Upon completely reading this post, you will know the basic 5Ws of AWS RIs, how to bring RIs into practice, types of AWS Reserved Instances, payment attributes associated with instance reservations, attributes to look for while buying/configuring an RI, facts to be taken into account while committing RIs, top RI best practices, top RI governance tactics that help reduce AWS bill shock, and common misconceptions attached to RIs.
The Essence: Get Your RI Basics Right to Reduce AWS Bill Shock
Today, RIs are one of the most effective cost saving services offered by AWS. Irrespective of whether the reserved instances are used or unused, they will be charged. And AWS offers discounted usage pricing for as long as organizations own the RIs. So, opting for reserved instances over on-demand instances might waste several instances. However, a solid RI planning will provide the requisite ROI, optimal savings, and efficient AWS spend management for a long term.
AWS RIs are purchased for several reasons, like savings, capacity reservation, and disaster recovery.
Some of them are listed below here:
Reserved instances provide the highest savings approach in AWS cloud ecosystem. You can lower costs of the resources you are already using with a lower effective rate in comparison to on-demand pricing. Generally, EC2 and RDS RIs are contenders of projecting highest figures in your AWS bill. Hence, it’s advisable to go for EC2 and RDS reservations.
A Case-in-point: Consider an e-commerce website running on AWS on-demand instances.Unexpectedly, it started gaining popularity among customers. As a result, the IT manager sees a huge spike in his AWS bill due to unplanned sporadic activity in the workload. Now, he is under pressure to control both his budget and efficiently run the infrastructure.
A swift solution to this problem is opting for instance reservation against on-demand resources. By reserving instances, he can not only balance capacity distribution and availability according to work demands, but it can also reap substantial savings due to reservation.
P.S: Just reserving the instances will not suffice. Smart RI Planning is the way forward to reap optimal cost savings.
[mk_blockquote style=”quote-style” font_family=”none” text_size=”12″ align=”left”]Botmetric helps you make wise decisions while reserving AWS instances. It also provides great insights to manage and utilize your RIs that ultimately lead to break-even costs. Get a comprehensive free snapshot of your RI utilization with Botmetric’s free trial of Cost & Governance.[/mk_blockquote]
2. Capacity Reservation
With capacity reservation, there is a guarantee that you will be able to launch an instance at any time during the term of the reservation. Plus, with AWS’ auto-scaling feature, you will be assured that all your workloads are running smoothly irrespective of the spikes. However, with capacity reservation, there will be a lot of underutilized resources, which will be charged irrespective of whether they are used or unused.
A Case-in-Point: Consider you’re running a social network app in your US-West-1a AZ. One day you observe some spike in the workload, as your app goes viral. In such a scenario, reserved capacity and auto-scaling together ensure that the app will work seamlessly. However, during off season, when the demand is less, there will be a lot of underutilized resources that will be charged. A regular health check of the resource utilization and managing them to that end will provide both resource optimization and cost optimization.
[mk_blockquote style=”quote-style” font_family=”none” text_size=”12″ align=”left”]Botmetric performs regular health check of your usage of reservations and presents them in beautiful graphical representation for you to analyze usage optimally. Further, with the metrics, you can identify underutilization, cost-saving modification recommendations, upcoming RI expirations, and more from a single pane.[/mk_blockquote]
3. Always DR Ready
AWS supports many popular disaster recovery (DR) architectures. They could be smaller environments ideal for small customer workload data center failures or massive environments that enable rapid failover at scale. And with AWS already having data centers in several Regions across the globe, it is well-equipped to provide nimble DR services that enable rapid recovery of your IT infrastructure and data.
The Point-in-Case: Suppose, East coast in the U.S. is hit by a hurricane and everybody lines up to move their infrastructure to US-West regions of AWS. If you have reservation in place beforehand in US-West then your reservation guarantees prevention from exhaustion. Thus, your critical resources will run on US-West without waiting in the queue.
[mk_blockquote style=”quote-style” font_family=”none” text_size=”12″ align=”left”]Botmetric scans your AWS infra like a pro with DR automation tasks to check if you have configured backup on RDS and EBS properly[/mk_blockquote]
How to Bring RI into Practice
The rationale behind RI is simple: getting AWS customers like you to commit to the usage of specific infrastructure. By doing so, Amazon can better manage their capacity and then pass on their savings onto you.
Here are the basic information on RI types, options, pricing, and terms (1 and 3 year) for you to leverage RIs to the fullest. These will help you bring RI into practice.
Types of AWS Reserved Instances
1. Standard RIs: These can be purchased with one year or three-year commitment. These are the best suited for steady state usage, and when you have good understanding of your long-term requirements. It provides up to 75% in savings compared to on-demand instances.
2. Convertible RIs: These can be purchased only with three year commitment. Unlike Standard RIs, Convertible RIs provide more flexibility and allows you to change the instance family and other parameters associated with a RI at any time. These RIs also provide savings but up to 45% savings compared to on-demand instances. Know more about it in detail.
3. Scheduled RIs: These can be launched within the time-span you have selected to reserve. This allows you to reserve capacity in a stipulated amount of time.
Types of AWS Reserved Instances and their characteristics
AWS RIs can be bought using any of the three payment options:
1. No-Upfront: The name says it all. You need not pay any upfront amount for the reservation. Plus, you are billed at discounted hourly-rate within the term regardless of the usage. These are only available for one year commitment if you buy Standard RI and for three years commitment if you opt for Convertible RI.
2. Partial Upfront: You pay a partial amount in advance and the remaining amount is paid at discounted hourly rate.
3. All Upfront: You make the full payment at the beginning of the term regardless of the number of hours of utilized. This option provides the maximum percentage of discount.
From our experience, a lot of stakeholders take a step back while committing towards reservation, because it’s an important investment that needs lot of deliberation. The fact is: Once you understand the key attributes, then it gives you all the confidence to commit on RIs.
Realize: How to?
- Identify the instances, which are running constantly or having a higher utilization rate (above 60%)
- Estimate your future instance usage and identify the usage pattern
- Spot the instance classes that are the possible contenders for reservation
Evaluate: How to?
Once you know how to realize the RIs, you can identify possibilities and evaluate the alternatives with the following actions:
- Look for suitable payment plans
- Monitor On-Demand Vs. Reserved expenditure over time
- Identify break-even point and payback period
- Look for requirements of Standard or Convertible RIs
Select: How to?
Once you know how to evaluate, you can analyze and choose the best option that fits your planning, and further empower your infrastructure for greater efficiency with greater savings.
Implement: How to?
Once you know what your requirements are to commit for a RI purchase, implementation is the next stage. It is very crucial you do it right. For the reason that: Discounts might not apply in all cases. For instance, if you happen to choose the incorrect attributes or performs incorrect analysis. At the end of the line, your planned savings might not reflect in your spreadsheets (*.XLS) as calculated.
How to Implement the Chosen RI Like a Pro
The key parameter to reserve an EC2 instance is the instance class. To apply reservation, you can either modify or go for a new RI purchase by selecting platform, region, and instance type to match the reservation.
Consider a company XYZ LLC, where it has an on-demand portfolio of
- 2*m3.large Linux in AZ us-east-1b
- 4*c4.large Linux in AZ us-east-1c
- 2*t2.large Linux in AZ us-east-1a
And XYZ LLC now purchases standards RIs as below:
- 4*c4.large Linux in AZ us-east-1c
- 2*t2.large Linux in AZ us-east-1b
- 4*x1.large Linux in AZ us-east-1a
Based on the above on-demand portfolio and purchases, the following reservations are applied for XYZ LLC:
- 4*c4.large Linux in AZ us-east-1c. Here’s how: This matches exactly the instance class the reservation was made, so offers discounts
- 2*t2.large Linux in AZ us-east-1b. Here’s how: The existing instances class is in a different AZ but in the same region, so no discount is applied. However, if you change the scope of RI to region then the reservation will be applied but there is no guarantee of capacity
- 4*x1.large Linux in AZ us-east-1a. Here’s how: The instance family, region, and OS don’t match. In this case, reservation will not be applied for these instances. However, if XYZ LLC had purchased Convertible RIs, modifying reservation will never be a problem but they have to commit for 3 years with a lesser discount.
Making Sense of the RIs for Payer and Linked Accounts
AWS bills, evidently, includes charges only on payer account for all utilization. However, in larger organizations, where the linked accounts are divided into business units, reservation purchases are made by these individual units. No matter who makes the purchase, the benefits of RI will float across the whole account (payer + its own linked accounts).
For Instance: Let’s assume X is the payer account and Y and Z are its two linked accounts. Then in an ideal situation:
U-Can be applied
X($) then Y (U) or Z (U)
If Z($) then Y(U) or X(U) or Z(U)
Hence, in a group, reservation can be applied in any instance class available.
How to Govern RIs with Ease
Monitoring just a bunch of RIs are easy when the portfolio is small. However, in case of mid-sized and large sized businesses, RIs generally don’t get proper attention due to the dynamic environment and the plethora of AWS services to manage. This causes a dip in efficiency, unexpected minimal savings, and many more such issues. Nevertheless, this dip in efficiency and bill shock can be assuaged with few tweaks:
Make a regular note of unused and underutilized RIs:
[mk_blockquote style=”quote-style” font_family=”none” text_size=”12″ align=”left”]Unused and underutilized states of RIs are key issues that lead to inefficiency.[/mk_blockquote]
In case of unused RIs: The reservations were intended to be bought for determined constant utilization but somehow the utility ended just after few months of purchase and the reservation is now in dormant state or unused. If you modify and eliminate them, then they will add to cost savings.
In case of underutilized RIs: Few RIs are bought with the intention to use them for continuous workload but somewhere in the timeline the utility reduced and the reservation is not clocking to its ideal utilization. If you start reusing them, then they will add to cost savings. Read this blog post by Botmetric Director of Product Engineering Amarkant Singh’s post on how to go about unused and underutilized RI modifications and save cloud cost intelligently.
Finding the root cause of unused and underutilized RIs:
1. Incorrect analysis: While performing the analysis to determine RI purchase some miscalculations or lack of understanding of environment can be cause of trouble in management of RIs.
a. Wrong estimation of time (1 year/ 3 years): If you couldn’t understand your projected workload time then purchasing reservation for a longer interval e.g.: 3 years may bring RI into unused state
b. Wrong estimation of count: This could be due to overestimation/underestimation of the number of reservations required. If it’s too many, then you may modify them for DR capability. But if it’s too less, then you may still not satisfy your savings
c. Wrong estimation of projected workload: If you have not understood your workload, then chances are that you could have bought RIs with incorrect attributes like time, number of instances bought, etc. In such cases, RIs either go unused or underutilized
2. Improper Management: RIs, irrespective of the service, can offer optimal savings only when they are modified, tuned, managed, and governed continuously according to your changing infrastructure environment in AWS cloud.
You should never stop at reservation. For instance, if you have bought the recent Convertible RIs, then modifying them for the desired class. And, if you have older RIs, then get them to work for you either by breaking them into smaller instances or by combining them for a larger instance as per the requirement.
[mk_blockquote style=”quote-style” font_family=”none” text_size=”12″ align=”left”]FACT: Almost 92% of AWS users fail to manage Reserved Instances (RIs) properly, thus failing to optimize their AWS spend.[/mk_blockquote]
If you find all this overwhelming, then try Botmetric Smart RI Planner, which helps with apt RI management, right sizes your RIs and helps save cost with automation.
Top RI Best Practices to Live By
There are few best practices you should follow to ensure your RIs work for you and not the other way around.
Get Unused RIs back to work
If you have bought the recent Convertible RIs, then modifying them for desired class is now a child’s play. However, if you have older RIs, then getting them back to work is not so easy as Convertible RIs. But with just few modifications like breaking them into smaller instances /combining them for a larger instance according to your needs will do the trick.
Keep an eye on expired and near-expiry RIs in your portfolio
Always list your RIs in three ways to keep a constant check on them:
Active: RIs that are either new or will expire in 90 days
Near-expiry: RIs that are nearing to 90 days of expiration. Analyze these RIs and plan accordingly for re-purchase
Expired RIs: RIs that are expired. If there is an opportunity for renewal go ahead with it
Be sure of your workload demands and what suits your profile the best. Standard RIs work like a charm, in regards to cost saving and offering plasticity, only when you have a good understanding of your long-term requirements.
And if you have no idea of your long-term demand, then Convertible RIs are perfect, because you can have a new instance type, or operating system at your disposal in minutes without resetting the term.
[mk_blockquote style=”quote-style” font_family=”none” text_size=”12″ align=”left”]Botmetric has found a smart way to perform the above. It uses the NoOps and AIOps concept to put the RIs back to work. Read this blog to know how.[/mk_blockquote]
Compare on-demand Vs. reserved instances to improve utilization
If you want to improve your utilization of reservation, the game plan is to track on-demand Vs. reserved instances utilization. It is evident from our experience that RI cost over a period of time offers the greatest discounted prices. Read this blog post to know the five tradeoff factors to consider when choosing AWS RIs over on-demand resources.
For further benefits, a report on RI utilization that can throw the below insights will help:
1. Future requirement of reservation
2. Unused or underutilized RIs
3. Opportunities to re-use existing RIs
Here is a sample Botmetric RI Utilization Graph for your reference:
Before wrapping-up, here are few common RI misconceptions that you must know.
Common RI Misconceptions You Must Know
- If you buy one EC2 instance and reserve an RI for that type of EC2, then you don’t own two instances but you own only one
- RIs are not only available in EC2 and RDS but in five other services as well that can be reserved
- Purchasing RIs alone and not monitoring and managing them may not give you any savings.
- Managing and optimizing them is the key
- Never purchase instance for an instance ID, but for an instance class
- Buying a lot of RIs will not bring down the AWS bill
- Managing RIs is very complex. It’s a continuous ongoing process. Few key best practices — if followed — can give desired savings and greater efficiency
- Older RIs cannot have Region benefit
- RIs can’t be re-utilized, if you fail to understand your workload distribution RIs can’t be returned, instead
- AWS RI Marketplace facilitates you to sell your RIs to others
RIs, as quoted earlier, are the highest saving option in your dynamic cloud environment. Buying RIs is not sufficient. A proper road map and management coupled with intelligent insights can get you the desired savings.
AWS is always coming up with new changes. Hence, understanding its services and knowing how to use them for greater benefit will always prove beneficial for your cloud strategy irrespective of your business size, above all for the startup world. And if you find all this overwhelming, then just try Botmetric’s Cost and Governance.