Cost optimization - Don't overspend on AWS

Editor's Notes

• #3: AWS serves hundreds of thousands of customers in more than 190 countries. Amazon CloudFront and Amazon Route 53 services are offered at AWS Edge Locations
• #4: AWS has developed the broadest collection of services (more than 90) available from any cloud provider. Foundation services across compute, storage, security, and networking offer customers flexibility in their architecture. We have a full spectrum of options to meet most price-to-performance scenarios. We offer the capability for both managed and unmanaged database options. The offerings for Analytics and Application Services enable advanced data processing and workloads. Our management tools offer a lot of insight and flexibility to let you manage your AWS resources through either our tools or the management tools you’re already familiar with. Recent expansion into enterprise applications has been entirely driven by customer feedback on where they’d like us to deliver value.
• #5: TALKING POINTS Customers have selected AWS for years because we have proven ourselves committed to customer success. We believe we stand apart in the market because of seven factors: Security, Service Breadth and Depth, Experience, Global Footprint, Artificial Intelligence , Partner Ecosystem and Enterprise Leadership recognized by Gartner
• #7: TCO Business value Cost optimisation
• #13: We’re operating in a new IT consumption model, where IT can be consumed on demand. With this utility pricing model, everyone pays for what is used. It’s like your water bill. If your faucet is running… you’re paying your water bill at the end of the month based on what you used. But what happens if you leave that faucet running? Or have a teenager that likes long showers? Or have a leaky pipe? Cost optimization is about focusing on what is needed and identifying ways that we can turn the corner between “usage” and “need” to ensure we’re as lean and efficient as possible
• #14: Now Lets have a look at where you can optimise from a technical hands-on perspective
• #15: There are many ways in which you can save on AWS. Firstly there are some easier options to save such as: [Click] Consolidated billing – which is linking your accounts to a single payer account to help trigger volume discounts and maximize Reserved instances benefits Rightsizing before migration or stating your instances – it’s easy to Deleting unused resources – the equivalent of finding a $50 dollar bill on the ground and picking it up (tagging helps significantly with this) And Matching usage to storage class – S3 now has S3 analytics which identifies by bucket and tag, what is not used frequently and policies can be applied to automatically move files to infrequent access, which costs less. Slightly more complex ways include: [Click] Planning/ Architecting for cost before migration – here mature customers will design using Serverless services first, then consider containerization if that’s not possible, and then finally EC2 as a last resourt Turning instances off outside of working hours – typically 70% time in a week is non-working time. Many dev, test, staging applications can be turned off. Tagging enables this which we will cover later today. And using reserved instances – reserved instances provide discount for commitment, and are commonly used for EC2. However they also exist for RDS, Redshift, Elasticache and DynamoDB which many customers run 24-7 always on. And finally some more harder ways to save [Click] Spot instances – I’ve seen companies operate 70% of their production workload in Spot with architecture that performs as well as on-demand Scaling instances to meet external demand – websites are a perfect example of this where in busy seasons you want to ramp up and ramp down and not pay for those resources in quiet periods Rightsizing after migration saves significant cost and is supported by tooling mentioned later in this talk, Architecting for cost after migration, or breaking apart monolith applications into micro-services or taking a SaaS approach to applications where you only pay for what you use. Using containerization to maximize utilization (or squeezing more into less) And the immensely powerful CloudFormation allowing hardware to be automated, resulting in immense cost savings from a staff-time perspective. For example, Bristol Myers Squibb developed Automation using CloudFormation scripts to spin up entire secure medical analysis environments, saving 98% computation time relative to on-premise, saving analyst staff time, at a low cost, in a repeatable fashion With all of this, what is considered difficult melts away with the appropriate training, knowledge, and tools.
• #16: So how much do these activities save? [Click] Short answer? a lot! For example, architecting for cost and using pay for use services like Lambda or DynamoDB can easily save over 80% of your application cost. Deleting unused resources such as idle EC2, instantly saves 100% of the resource cost I’ve seen customers with thousands of dollars of incomplete S3 transfers sitting there, unused ELBs, and years worth of backup volumes that are not needed.
• #19: We see infra for Dev and Test team single timezone Really easy turn off when bed Easier still if non prod separate account In a moment we’ll look at tools large customers use
• #20: Amazon RDS for MySQL, MariaDB, PostgreSQL, Oracle and SQL Server now allows you to easily stop and start your database instances. This makes it easy and affordable to use databases for development and test purposes, where the database is not required to be running all of the time.  Stopping and starting a database instance requires just a few clicks in the AWS Management Console, or a single call using the AWS API or AWS Command Line Interface, and takes just a few minutes. While your database instance is stopped, you are charged for provisioned storage, manual snapshots and automated backup storage within your specified retention window, but not for database instance hours. While a database instance is stopped, Amazon RDS does not delete any of your automatic backups or transaction logs. This means you can do a point-in-time restore to any point within your specified automated backup retention window, even after an instance is started. Starting an instance restores it to the same configuration as it had when stopped, including its endpoint, DB parameter group, security group, and option group membership.  You can stop an instance for up to 7 days at a time. After 7 days, it will be automatically started. For more details on stopping and starting a database instance, please refer to Stopping and Starting a DB Instance in the Amazon RDS User Guide.  The stop/start feature is available for database instances running in a Single-AZ deployment which are not part of a Read Replica (both source and replica) configuration. Please see Amazon RDS Pricing for regional availability. 
• #24: Customers can primarily choose from 3 different purchasing options for EC2 – On Demand, RIs and Spot. On Demand is a pay as you go model with no commitments. You can increase or decrease your compute capacity depending on the demands of your application and only pay the specified hourly rate for the instances you use. This is suitable for customers that have unpredictable workloads or for applications that are being developed & tested on EC2 for the first time. RIs – This is where we’ll spend most of our time today. RIs offer a significant discount (upto 75%) over On Demand in exchange for a 1 year or 3 year commitment, and also provide an optional capacity reservation. RIs are best suited for customers that have steady state or predictable usage. Spot – AWS offers you the ability to bid on spare EC2 capacity for up to 90% discount off On Demand prices. Customers who have workloads that have flexible start and end times or who have urgent computing needs for large amounts of capacity can benefit from this model.
• #27: There are three payment options for Reserved Instances, each with a different upfront amount – No Upfront—You are billed a discounted hourly rate for every hour within the term, regardless of usage, and no upfront payment is required. 2. Partial Upfront—A portion of the cost must be paid upfront and the remaining hours in the term are billed at a discounted hourly rate, regardless of usage. 3. All Upfront—Full payment is made at the start of the term, with no other costs incurred for the remainder of the term, regardless of hours used. The All Upfront payment option provides the highest discount.
• #29: There are two types of RIs – Standard RIs and Convertible RIs. A Standard RI offers you the most amount of discount, when you commit to a 1 year or 3 year term. 3 year Standard RIs offer a higher discount than their 1 year counterparts. You can purchase a Standard RI using the AWS CLI, APIs or the AWS management console. At the time of purchase, you need to specify the instance type, OS, Tenancy and AZ.
• #30: Convertible RIs offer you the flexibility to change the attributes of your RI configuration, while still receiving a significant discount (usually around 45% off On Demand prices). The Convertible RI is useful for customers who can commit for a three-year term in exchange for a significant discount on their overall EC2 usage, but are uncertain about their instance needs in the future. When you purchase a Convertible RI, you can modify the RI across instance families, sizes, OS types or tenancy.
• #35: Spot Instances provide the ability for customers to purchase compute capacity with no upfront commitment and at hourly rates usually lower than the On-Demand rate, often as much as 90% cheaper - for those wondering what is a 90% discount? It is about 1c per core hour. Ask yourself what could your best people do, or how well could your application perform with a 10,000 core data center that costs just $100 per hour.. What applications should be considered for spot? Applications that have flexible start and end times Experiments that can only be conducted at very low compute prices (Brookhaven and Fermi – analyzing the origins of our universe). Or business that need extremely low infrastructure costs to achieve profitability such as Adtech. Users with urgent computing needs or large amounts of additional capacity
• #40: If you think about the typical lifecycle of data, newly created active data is access very frequently. In our example take a new video clip you share with your friends and family. People will be consuming this new data actively, this new video will be played back frequently, shared and commented on very frequently. As this video becomes older, a smaller number of people will engage, it will be LESS FREQUENTLY accessed. S3 Standard-IA is a new LOW COST storage class on Amazon S3 we launched just a few weeks ago. Designed for data that is access infrequently called “Standard-Infrequent-Access”. This new storage class offers the same great durability and Performance of Amazon S3 Standard Storage class with a slightly lower availability… ideal for workloads that are “COLDER” and “less frequently accessed”. If you don’t want to think about your data access patterns but just want to high durability, availability and performance for Amazon S3 you can simply select S3 Standard. For data that is less-frequently accessed, you can leverage Amazon S3 Standard-IA to save on cost while still benefiting from the great durability and performance as S3 Standard. At some point in time your data will be ready to be archived because no one if actively interacting with your data and you need to archive that away for record keeping etc. In addition to transitioning your data to S-IA as its characteristics change, you can also leverage Amazon S3 Standard-IA for new data that fits the bill for Infrequently accessed data. For example you can leverage the S-IA storage class to stored detailed applications logs that you analyst in-frequently and save on storage cost.
• #42: If you’re not responsible for the hands-on actioning of cost savings, there are mechanisms that you can set up so that your organisation doesn’t need to go through an effort intensive cost optimisation blitz every 6 months.
• #43: First of all, what do we mean by mechanisms? Jeff once said, “Good intentions never work, you need good mechanisms to make anything happen” As much as teams want to save on AWS, you need mechanisms to drive action
• #44: Once we have gathered the party, we can start setting up the remaining mechanisms These include: Tagging & Enforcement of Tagging Measures and Targets of Success Reporting & Tooling Governance and Controls And Ongoing Education
• #45: For resource provisioning, there are several options. One is using AWS service catalogue with IAM roles and permissions requiring approval to access more expensive resources for example.
• #46: AWS Organizations can help you define policies by Organizational Unit. For example, allowing the Finance team to only access S3 buckets for billing files. The link at the bottom provides further detail on AWS organisations
• #47: So what should we measure? When starting out, go for measures that won’t give anyone a heart attack (e.g. non-financials) Financial measures such as current and potential savings by team invites contention and possibly defensive behavior. Try to use financial measures when you know teams are ready and have bought-in to the calculation method. Without measures teams I’ve seen teams claim to be optimized via anecdotal cases, but you have no real view of overall optimisation.
• #48: Out of the box, from a reporting standpoint, AWS provides Cost Explorer which is an excellent starting point. It allows you to answer questions such as where is my data transfer spend coming from? How much did I spend on backup storage last month? And You can also save views for later by clicking save report.
• #51: AWS Trusted Advisor can point you towards inefficiencies. Note that some of the savings will not be realizable, but you can treat it as a compass that points to potential savings.
• #52: In-house developed reporting can be as simple as feeding the Cost and Usage Report into Redshift or Athena and visualizing in a reporting tool such as Quicksight. The link at the bottom provides a guide on how to set this up. Or simply google, cost and usage report redshift and quicksight
• #53: When we talk about governance and controls, we ask questions such as: What’s your cost review process? How do you control visibility based on organizational unit and level? What happens when costs increase unexpectedly? Who gets to switch on resources? What happens to resources that are not tagged? What happens when a business unit nears or exceeds budget? The video link provided gives you a view of what advanced governance and cost management looks like at Netflix I encourage you to have a look. Because these questions are highly individual to each business, and it’s nearly impossible to plan in advance what to do. Therefore an agile and clearly communicated approach to governance that involves relevant stakeholders will set you up for successs.
• #54: 3rd party options also provide a lot of depth and functionality in cost optimization and cost management reporting and tooling. They allow you to have instant access to thousands of hours of development time otherwise spent in building your own reporting and tooling, and makes a lot of financial sense for some customers to use. Some of these provide advanced function like reserved instance cost and benefits allocation by account and tag. And extend past cost control into security and automation.
• #55: Here are some High quality talks that will help guide best practice and inspire your teams to take action. Each of these are less than an hour long and watching these will put you well ahead of the curve with regards to Cost Optimisation knowledge. If you’re willing to invest a little more time, there is also a course on A Cloud Guru on cost management.