This is part 6 of a ten part series, "Top 10 Strategies to Manage Cost and Continuously Optimize AWS," which is excerpt from an eBook by the same name, produced in cooperation from AWS by HyperGrid Cloud Management Platform. A link to download the entire eBook with all 10 strategies is located at the bottom of this post.
CLOUD STRATEGY #6 - CENTRALIZE GOVERNANCE OF RESERVED INSTANCES
A common AWS billing model is to have a centralized account with Consolidated Billing, linked to a number of autonomous accounts. With this model, it is common for individual accounts to purchase Reserved Instances based on their individual usage patterns.
Individual accounts may not be linked to each other—the topology is a star, not a mesh. Price breaks due to reservations are isolated in individual accounts, or accounts that are linked to it. This means unused RIs cannot be shared between accounts in this topology.
Wastage—RIs purchased in an individual account but not utilized by this account cannot be reused. Also, since the individual accounts are no longer responsible for billing, they may not have the hygiene in place to discover and repurpose or resell unused RIs.
Reserved Instances should be procured and managed centrally. Purchasing an RI is only the beginning; you should have a process in place to continuously monitor RI utilization and modify unused RIs (split/join or exchange convertible RIs) to maximize their usage.
Use master accounts to manage the lifecycle of Reserved Instances—recommendations, procurement, and monitoring. This way, they can flow to all linked accounts. HyperCloud Analytics is a great place to start for discovering opportunities for Reserved Instances. HyperCloud can easily map all your member and master accounts; plus, it can offer a single consolidated view of all instances across all members, and the analytics to make decision-making across all member accounts easier.
Enroll all stakeholders in RI purchases. While RIs can be procured centrally, the individual account owners should be involved in the process of analyzing and approving the process of RI procurement. Ensuring that all stakeholders are empowered to understand the cost implications of their actions is part of good cloud governance.
There will be exceptions: pricing breaks due to RIs can flow between accounts, but capacity reservations are localized to purchasing account. Therefore, there may be workloads (e.g., mission-critical applications or a DR workflow) that need localized management to ensure capacity reservation.
We also recommend having organizational standards around instance types, increasing the
probability that an unused RI will be picked up and applied to an instance in another account. This is not always feasible, however, because individual vendors and business needs drive instance selection.
RIs are the best pricing breaks you can get in EC2. By creating a centralized RI governance model, you can ensure that you capture the maximum benefit from an RI investment.
Cloud app planning is a much broader discipline than just instance selection; it requires taking a look at the “whole application”—that is, the application itself and the components that make up the deployment (e.g., load balancers, proxy servers, cache, database servers, log management, performance management, and so forth).
Cloud app planning means mapping the data transfer between these and choosing an optimal resource placement that takes into account such factors as network and data transfer costs and data durability.
Lack of a holistic view of costs before deploying the “whole application” can often lead to pricing surprises. For example, before you deploy an application in an HA configuration, you should be able to quantify the network characteristics of the application before the deployment and understand their effect on pricing. An application that generates a lot of traffic could generate more in cross-AZ and cross-region transfer costs than the business benefit of High Availability.
Applications need to be modeled in their entirety, not just the individual instances. Their interactions and network traffic between components need to be captured and analyzed before making recommendations on placement.
When planning your application, make sure that you treat your application as a collection of services that communicate with each other, and not just as a collection of individual instances. For example, before deploying an HA architecture across regions, factor in the cost of having your application communicate across regions. Similarly, before deciding on S3 as a storage mechanism, factor in not just your storage costs, but also your retrieval costs.
Our solution, HyperCloud, has a Cloud App Planning service that allows you to model an entire application, with all its dependencies. The service will recommend a placement for the application (along with costs) that factors in such details as data and network transfer costs and storage tiers. It also allows for easy what-if scenarios that help with optimal placement of test, dev, production and DR instances of the application.
By modeling applications in their entirety and factoring the bigger picture into not just instance choice but also placement, you can ensure that you have a durable deployment architecture without pricing surprises.
CLICK HERE to see this recorded webinar.
Greybeard Consulting's President, Chris Gerhardt is featured. Chris talks about how tools like HyperGrid accelerate cloud adoption and streamline IT, while cutting costs and mitigating risk.
"Top 10 Strategies to Manage Cost and Continuously Optimize AWS"
Read all ten tips now.