Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Upcoming SlideShare
What to Upload to SlideShare
What to Upload to SlideShare
Loading in …3
1 of 50

Amazon EC2 Instances, Featuring Performance Optimisation Best Practices



by Enda Kenny, AWS Solutions Architect

Related Books

Free with a 30 day trial from Scribd

See all

Related Audiobooks

Free with a 30 day trial from Scribd

See all

Amazon EC2 Instances, Featuring Performance Optimisation Best Practices

  1. 1. © 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Enda Kenny, Solutions Architect 09th April 2018 Deep Dive on EC2 AWS Pop-up Loft Dublin
  2. 2. Amazon EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  3. 3. Amazon EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  4. 4. Amazon Elastic Compute Cloud (EC2): Virtual servers in the cloud Physical Servers in AWS Global Regions Host server Hypervisor Guest 1 Guest 2 Guest n
  5. 5. Amazon EC2 12 years ago… Scale up or down quickly, as needed Pay for what you use M1 “One size fits all”
  6. 6. EC2 instance growth since then 3 5 7 11 12 13 23 42 52 60 70 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
  7. 7. EC2 instance characteristics i3.xlarge Instancefamily Instancegeneration Instancesize Instancetype CPU Memory Storage Network Perf
  8. 8. Amazon Machine Images (AMIs) Amazon maintained Set of Linux and Windows images Kept up-to-date by Amazon in each region Community maintained Images published by other AWS users Managed and maintained by Marketplace partners Your machine images AMIs you have created from EC2 instances Can be kept private or shared with other accounts
  9. 9. EC2 instances General purpose Dense storage Compute optimised FPGAsGPU Compute Storage optimised Graphics intensive Memory optimised High I/O P2M4 D2 X1 G2T2 R4I3C5 F1M5 P3H1 EC2 Bare MetalG3T2Unlimited X1eI2C4 High I/O General purpose burstable Direct access to physical server resources
  10. 10. General Purpose instance workloads Web/app servers Enterprise apps Gaming servers Caching fleets Analytics applications Dev/test environments
  11. 11. M5: Next-Generation General Purpose instance • Powered by 2.5 GHz Intel Xeon Scalable Processors (Skylake) • New larger instance size—m5.24xlarge with 96 vCPUs and 384 GiB of memory (4:1 Memory:vCPU ratio) • Improved network and EBS performance on smaller instance sizes • Support for Intel AVX-512 offering up to twice the performance for vector and floating point workloads 14% price/performance improvement With M5 M4 M5
  12. 12. Opportunity: Most instances aren’t very busy
  13. 13. T2: General Purpose Burstable instances T2 Burstable Performance instances provide a generous baseline level of CPU performance with the ability to burst above the baseline Lowest cost EC2 instance at $0.004 per hour, and available on AWS Free Tier WithT2 Unlimited, burst whenever you want for as long as you want Just $0.05 per vCPU-hour over baseline, averaged over 24 hours T2.nano 0.5GiB 1 vCPU Base perf 5% …7 sizes T2.2xlarge 32 GiB 8 vCPU Base perf 135%
  14. 14. R4: Memory Optimised instances In-memory databases In-memory analyticsIn-memory caches • 8:1 GiB to vCPU ratio • Memory-optimised instances with Intel Xeon (Broadwell) processors • Up to 25 Gbps NW bandwidth r4.large 15.3 GiB 2 vCPU 488 GiB 64 vCPU r4.16xlarge …6 sizes
  15. 15. X1 and X1e—Large-Scale Memory-Optimised For memory-intensive workloads and very large in-memory workloads 32:1 GiB to vCPU ratio High-performance databases, Large in- memory databases (e.g. SAP HANA), and DB workloads with vCPU based licensing (Oracle, SAP) For large in-memory workloads 16:1 GiB to vCPU ratio In-memory databases (e.g., SAP HANA), big data processing engines (Apache Spark, Presto), in-memory analytics X1eX1 1 TB 64 vCPU x1.16xlarge 2 TB 128 vCPU x1.32xlarge 2 TB 64 vCPU X1e.16xlarge 4 TB 128 vCPU x1e.32xlarge …6 sizes Roadmap through 2018: up to 16TB memory instances! …2 sizes
  16. 16. I3: I/O optimised instances 9X as many IOPS as I2 I2 I3 • Intel Xeon E5 v4 (Broadwell) processors, with up to 15.2TB of locally attached NVMe SSD storage, 64 vCPUs, and 488 GiB memory • Lowest cost per IOPS ($/IOPS) • Offers very high Random I/O (up to 3.3 million IOPS) and disk throughput (up to 16 GB/s) • Up to 25 Gbps NW bandwidth High-perf databases Real-time analytics No SQL databasesTransactional workloads SQL
  17. 17. EC2 Bare Metal EC2 Bare Metal Run bare metal workloads on EC2 with all the elasticity, security, scale, and services of AWS i3.metal 36 hyperthreaded cores 15.2 TB SSD-based NVMe storage 512 GiB RAM Designed for workloads that are not virtualised, require specific types of hypervisors, or have licensing models that restrict virtualization Powers the VMware Cloud on AWS
  18. 18. Dense Storage workloads—D2 and H1 Data warehousing Log processing HDFS d2.8xlarge 244 GiB 36 vCPU 48 TB HDD • Lowest cost per storage ($/GB) • Supports high sequential disk throughput • More vCPUs and memory per terabyte of disk • Lower costs for big data uses cases 256 GiB 64 vCPU 16 TB HDD h1.16xlarge Big data Kafka MapReduce
  19. 19. Compute-intensive workloads Batch processing Distributed analytics High-perf computing (HPC) Ad serving Multiplayer gaming Video encoding
  20. 20. C5: Compute-optimised instances based on Intel Skylake • Based on 3.0 GHz Intel Xeon Scalable Processors (Skylake) • Up to 72 vCPUs and 144 GiB of memory (2:1 Memory:vCPU ratio) • 25 Gbps NW bandwidth • Support for Intel AVX-512 25% price/performance improvement over C4 C4 C5 “We saw significant performance improvement on Amazon EC2 C5, with up to a 140% performance improvement in industry standardCPU benchmarks over C4.”
  21. 21. Accelerated computing on AWS Parallelism increases throughout CPU: High speed, highly flexible GPUs and FPGAs can provide massive parallelism and higher efficiency than CPUs for many categories of applications
  22. 22. High-performance graphics with G3 Seismic exploration and analytics for oil and gas “The exploration and production models are increasingly complex with very large datasets, 3D and dynamic algorithms, security, and global reach... . Amazon EC2 G3 instances enable Landmark to deliver value to our clients in ways that were not possible before.” - Chandra Yeleshwarapu, Global Head of Services and Cloud Landmark, Halliburton Ideal for workloads needing massive parallel processing power Visualizations Cloud workstation 3D rendering Video encoding Virtual reality 4 GPUs, 64 vCPUs, 488 GiB of host memory, and 20 Gbps of network bandwidth Tesla M60 GPU offers 8 GB of GPU memory, 2048 parallel processing cores and a hardware encoder 10 H.265 (HEVC) 1080p30 streams 18 H.264 1080p30 streams
  23. 23. Graphics Acceleration: Elastic GPUs Allows customers to add low-cost graphics acceleration to Amazon EC2 instances over the network Come in a wide range of sizes; you can attach GPUs to a wide range of EC2 instances to achieve optimal performance OpenGL compliant, giving you the confidence to run any graphics-intensive application 1GiB GPU Memory 2 GiB 4 GiB 8 GiB C u r r e n t G e n e r a t i o n E C 2 I n s t a n c e
  24. 24. Use Cases for GPU Compute Machine learning/AI High-performance computing Natural language processing Image and video recognition Autonomous vehicle systems Recommendation systems Computational fluid dynamics Financial and data analytics Weather simulation Computational chemistry
  25. 25. Next Generation of GPU Compute Instances—P3 Instances • Industry’s most powerful GPU-based platform • Based on NVIDIA’s latest GPUTeslaV100 • 1 PetaFLOP of computational performance in a single instance • Provides up to 14X performance improvement over P2 for machine learning use cases • Up to 2.6X performance improvement over P2 for HPC use cases P3
  26. 26. Amazon EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  27. 27. Amazon EC2 instance store • Local to instance • Non-persistent data store • Data not replicated (by default) • No snapshot support • SSD or HDD EC2 instances Physical Host Instance Store or
  28. 28. Amazon Elastic Block Store (EBS) EC2 instance EBS volume • Block storage as a service • Create, attach volumes through an API • Service accessed over the network • Select storage and compute based on your workload • Volumes persist independent of EC2 • Detach and attach between instances • Choice of magnetic and SSD-based volume types • Supports Snapshots: Point-in-time backup of modified volume blockssc1st1 io1gp2 EBS SSD-backed volumes EBS HDD-backed volumes ElasticVolumes let you increase volume size or change volume type
  29. 29. Amazon EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  30. 30. Amazon Virtual Private Cloud (VPC) Flow LogsNATGatewayVirtual Private Cloud Provision a logically isolated cloud where you can launch AWS resources into a virtual network SecurityGroups & ACLs
  31. 31. EC2 Resources Recap AMI Virtual Machine Configuration Instance Running or StoppedVM VPC Amazon S3 EBS EBS EBS VPC EBS EBS EBS EBS Snapshots
  32. 32. Amazon EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  33. 33. AWS global infrastructure
  34. 34. Placement Groups CLUSTER SPREAD Placement Groups enable you to influence our selection of capacity for member instances, optimizing the experience for a workload EC2 places instances closely together in order to optimise the performance of inter-instance network traffic EC2 places instances on distinct hardware in order to help reduce correlated failures
  35. 35. Spread Placement Groups Database Cluster When deploying a No SQL database cluster in EC2, Spread Placement will ensure the instances in your cluster are on distinct hardware, helping to insulate a single hardware failure to a single node
  36. 36. Elastic Load Balancing Load balancer used to route incoming requests to multiple EC2 instances, Containers, or IP addresses in yourVPC ElasticLoad Balancing provides high-availability byutilizing multiple AvailabilityZones ELB EC2 Instance EC2 Instance EC2 Instance
  37. 37. Auto Scaling Auto Scaling groupAuto Scaling group Dynamic scaling ELB EC2 instances ELB CPU Utilization EC2 instances Replace unhealthy instances Fleet management
  38. 38. EC2 Foundations Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  39. 39. Launching Instances with Launch Templates Tags Launch Parameters User Data Network Interface Placement AMI ID EBSVolume InstanceType Console CLI API Instances Launch Block Device Mapping
  40. 40. Launch Templates CONSISTENT EXPERIENCE Templatise launch requests in order to streamline and simplify future launches in Auto Scaling, Spot Fleet, and On-Demand Instances SIMPLE PERMISSIONS GOVERNANCE & BEST PRACTICES INCREASE PRODUCTIVITY
  41. 41. Amazon CloudWatch • Monitoring service for AWS cloud resources and the applications you run on AWS • You can use Amazon CloudWatch to collect and track metrics, collect and monitor log files, set alarms, and automatically react to changes in your AWS resources
  42. 42. AWS Systems Manager • Securely manage Windows and Linux instances, EC2, or on-premises • Stay compliant with patching, config drift management, and software inventory • Automate daily tasks with delegated administration and approval • Centrally manage secrets and config items AWS cloud On-premise data center IT Admin, DevOps Engineer Granular Role-based Access Control with Audit Trail
  43. 43. Amazon EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  44. 44. EC2 Purchasing Options On-Demand Pay for compute capacity by the second with no long-term commitments Spiky workloads, to define needs Reserved Make a 1- or 3-year commitment and receive a significant discount off of On-Demand prices Committed, steady-state usage Spot Spare EC2 capacity at a savings of up to 90% off of On-Demand prices Fault-tolerant, dev/test, time-flexible, stateless workloads Per Second Billing for EC2 Linux instances & EBS volumes
  45. 45. EC2 Reserved Pricing Discount up to 75% off of the On- Demand price Steady state and committed usage 1- and 3-year terms Payment flexibility with 3 upfront payment options (all, partial, none) ReserveCapacity or opt for flexibility across AZs and instance sizes Convertible RIs Change instance family, OS, tenancy, and payment 1-Year Convertible RIs
  46. 46. Reserved Instance Recommendations
  47. 47. EC2 Spot Pricing Turbo Boost your results with Spot Fleet Spare EC2Capacity that AWS can reclaim with 2-minutes notice Savings up to 90% off of theOn- Demand price Eliminate the bid! No need to learn new APIs Pause and resume with Stop/Start and Hibernate
  48. 48. To optimise EC2, combine all 3 options 1. Use Reserved Instances for known/steady-state workloads 2. Scale using Spot, On-Demand or both 3. AWS services make this easy and efficient (e.g., Auto Scaling, Spot fleet, ECS, EMR, Thinkbox Deadline, AWS Batch) 0 2 4 6 8 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 /Spot On-Demand Spot Reserved
  49. 49. EC2 Resources Instances Storage Networking Availability Regions and AZs Placement Groups Load Balancing Auto Scaling Management Deployment Monitoring Administration Purchase Options On Demand Reserved Spot
  50. 50. Thank you Enda Kenny, Solutions Architect

Editor's Notes

  • Good morning everybody, Welcome to the inaugural AWS Pop-up Loft in Dublin. Thanks for coming to the first session of the week.
    This session is a Deep Dive on EC2
    My name is , I’m a solution architect here at Amazon Web Services
    If you have any questions come and find me after, ill be manning the Ask and Architect booths today and tomorrow and many of my colleagues are here also to help
    3 questions to audience – 1) How many are new to the Cloud?
    2) Who has spun up an EC2 server?
    3) How many are using EC2 for production workloads and are here to learn more about whats new on EC2?
  • This graphic shows the topics and the order of what we will cover today
    Over the next ¾ of an hour we will discuss EC2 in these terms
    Resources: the EC2 instances themselves, the storage they use and networking concepts in effect on AWS – will spend much of the time chatting about this subject as it is the meat of the EC2 story
    Availability: We will discuss how you can design your applications on EC2 for HA
    Management : As your fleet grows you need tools and services to manage it so we spend some time on that
    Purchase Options: How would you like to pay for your instances?
    Segway – we will hone in on resources to begin …. Click …We will come back to this image as we progress to show how the different components relate to each other

  • Specifically we will talk about EC2 instances
    First lets focus on EC2 resources….. Click……. Starting with EC2 instances

  • Message: This is where EC2 sits relative to physical servers and hypervisors
    You can think as the hypervisor as mechanism to carve up the physical server into virtual machines.

    [Transition] VMs on AWS are Amazon EC2 instances
    We will take care of security of the Cloud up to the hypervisor level while we provide you with the tools to look after security in the Cloud

  • In 2006 we had a goal of ‘the on-demand delivery of IT resources via the Internet with pay-as-you-go pricing’
    We wanted to create compute as a service –scale up an down, pay for what you use and that one size would fit all use cases
    The first two still hold true today, we still want elasticity, the ability to scale up and down as needed, you still pay only for what you use but customers told us one size does not fit all.
    [Segway]…. Click Now have over 100 instances types
  • Now have over 100 instance types.
    15 instance families, across 6 instance categories.

  • Explain the term I mentioned previously ..
    I3.xlarge is the API name
    3 is generation number
    Xlarge is the tshirt size – all the sizes have the same ratio but they get bigger

    [Segway] – what runs on these instances .. Click…
  • Defines the software and the configuration of these instances.
    Marketplace have over 4000 products from over 1200 ISV.
  • Will talk about some of these in more detail in a second but this represents the spectrum of the EC2 instance families
    Everything with a ‘New’ over it has been released in the last 18 month

    Segway - Lets start with workloads, because everything starts with workloads - click….. and work back to the instances that meet these needs
  • Good balance of CPU to memory
    Segway – we just introduced the latest M5 ….click
    8 mins
  • M1 only had 1 vCPU and 2 GiB – we've come a long way.
    Also have smaller t-shirt sizes in same ratio
    We have 25 Gbs of networking
    AVX512 vector extensions which lets you double your performance on certain floating point workloads

    Segway – when new looked at the CPU utilisation for our general purpose instances we notices that they are not at all busy …. Click

    9 mins
  • Left hand size represents 0% utilisation , right hand side 100%
    Most usage is in the region of 20 – 30 % CPU usage
    Typical pattern is not busy, then get web request and spike and then go back to not being busy.

    Segway - We wanted to come up with a way to make use of the unused CPU capacity and then to pass that saving on to our customer .. Click
    9 mins
  • We came up with a way to share the unused CPU across customers, by overselling instances and passing on the savings to our customers in the form of burst credits
    Customers loved these instances so we kept adding instance types to the family
    But if they run over the baseline consistently we throttle them down, to address this we Introduced T2 unlimited
    Burst whenever you want for as long as you want, you will be charged a small sum if you your average usage is above averaged CPU usage.

    Segway – we talked about general purpose instances having 4:1 Memory(GB): vCPU ratio … click.. R4 offer 8:1 …

    Mention workload type
  • Ideal for these workloads
    Segway - Feedback we got was that customers have larger datasets, SAP Hana, in memory caches, apache spark .. Click
  • For this we released the X1 instance family, this is great for these workloads

    Customers shared with us that they wanted us to grow with their needs, so we released the X1e family - 32:1 GiB to vCPU ratio, up to 4 TB

    [Message: is that we are growing ahead of the customer requirement – that’s why we show the second last and the last instance type in each family]

    We are releasing bigger instance sizes as per the roadmap to 16TB of memory
  • 2 types of storage:
    High performance high IO SSD
    High throughput , magnetic storage
    Lets start with I/O optimised,
    3.3 million IOPs, 9 times the IOPs offered in the previous generation I2

    Segway – first instance where we have a baremetal option .. Click
  • Have been working on offloading from the hypervisor to hardware offloads, that sit inside the host but don’t use up all the processer and memory
    Offloading software define networking, packet processing, EBS management, encryption
    Running on hardware offloads means you can have all the resources that you were previously used for these operations
    Now the hypervisor is optional, you can bring your own hypervisor, or bring OSs that don’t run on vms because of licensing

    Segway – some workloads are constrained by the disk throughput .. Click .. For these workloads we have D2

  • D2 - Workloads don’t need high IOs per second but need high sequential disk throughput
    D2 – Lowest cost per storage

    H1 Segway - Since we introduced this new workloads have come to the cloud - Big data, Kafka, MapReduce – that don’t need that ratio, need more vCPU for the disk, and memory for the disk they have
    Shipped the H1 - for Big data and Kafka they get lower costs it has same memory but twice as many vCPU - more CPUs and memory per TB of disk

    Segway – on to the workhorses of EC2 .. Click
  • The workloads are ..

    Segway - For these workloads we introduced the C5
  • Like M5 they use Skylake, but we worked with them to create a Skylake chip that has extremely high performance for these workloads, 3 GHz frequency
    25% price/ performance improvement over previous generation.
    AVX-512 – makes inference and vector based processing much faster
    Leverage AVX-512 vector extensions Netflix saw 140% performance improvement

    Segway - Compute is not just about CPUs anymore…
  • Trend in the industry with ML and HPC to get more done faster
    Metaphor, think of business jet, similar to a CPU, get 15-30 folks somewhere very quickly
    But a bullet train is not quite as fast but can carry thousands of passengers – bullet train is your man. like GPU/FPGA
    Parallelisation is key

    CPUs are constrained by the number of cores you can put on a die, and the number of ALUs (Arithmitic Logic Units) on a core
    GPUs can have 5000 cores, 8 on single machine
    FPGA can have millions of programmable logic cells
    Can do things in parallel that would take an age in a CPU

    Segway - we are leading the industry in adaption and acceleration of GPUs
  • Haliburton use G3s to process large datasets in the Oil and gas exploration and production industry.
    Segway - If you need just a little GPU
    Landmark’s DecisionSpace 365, the industry’s first end-to-end Exploration and Production SaaS solution, is empowering oil and gas companies to make the shift to cloud where they’re experiencing breakthroughs in efficiency across the E&P lifecycle
  • If you need a little GPU, Elastic GPUs lets you attach a GPU over the network
    We manage remoting the GPU calls to the Elastic GPU, use standard graphics drivers – open GL compliant
    Take you choice of instance, add your GPU according to you size, pay only for what you need.

    Segway – GPUs are not just about graphics workloads…. Click .. They are also used for compute – more general purpose use cases
  • They are also used for natural language processing, ML

    Segway – for these workloads we have the GPU powerhouse … click .. The P3
  • Provides 14x performance for ML
    2.6x for HPC

    1 Petaflop of computational performance
  • We discussed the various instance types but what about the storage these instances run on
    Segway – when we first launched EC2 in 2006 we only offered instance store ephemeral storage …… click
  • Instance store is still available today
    The management of backups and persistent is managed elsewhere as is the case for desktop, laptops and on premise servers

    Segway – but most of our customers use EBS
  • Volumes are logical, under the covers its more than one disk
    Can choose how much disk you need.
    Variety of volumes SSD, or provisioned IOPs – gp2 – General Purpose SSD; io1 – provisioned IOPs, st1 - throughput optimised HDD
    Can be snapshotted, 1st snapshot, copy every block - 2nd and subsequent on snapshots changed blocks
    Elastic volumes, change the volumes without service interruption
    Segway – so that’s the storage the instances use, so what are the networking considerations
  • so that’s the storage the instances use, so what are the networking considerations
  • Mention the Ugur session on VPC that is coming up at 11:30 – VPC – Networking Fundamentals and Connectivity.. .
    Logically isolated part of the AWS Cloud
    NAT GW, map a public ip to a private one to allow private instances to access resources on the internet
    Flow logs - look at traffic flow on your network adaptors and see what traffic was accepted and declined
    All accessible from on premise resources using DX or VPN
    Segway – lets recap on the EC2 resources we have covered thus far … click
  • So to bring together what we have discussed so far

    Segway -
  • So we are back to out guiding graphic, heres where we are, we have talked about AWS resources, now lets look at the what underpins the high availability that can be achieved using EC2
  • I wanted to share with you this image which show the Amazon Global network that interconnects our regions.
    Traverses Atlantic, Pacific, and Indian Oceans as well as the Mediterranean, Red and South China seas
    Its redundant 100 GbE cables, operate without impact through a link cut
    It not passed from one provider to another provider to another interconnection site to another interconnection but rather network is operated by one company, so we will give better quality service.

    Segway – Placement groups
  • Historically used to place instances close together, we have supported this for some time.
    Last year we introduced Spread
    Segway – so what's a typically use case for Spread . .click
  • Just a parameter in the run instances called strategy = Cluster, or Strategy = Spread
    Segway - ELB
  • Intsances across Azs, active/active config, not disruption.

    Segway -
  • Transitions – watch out!
    Segway -
  • Immediate Segway - As your fleets get bigger, you no longer just have 1 or a handful of servers to manage but you have hundreds, thousands or tens of thousands of servers, you need to get clever about how to manage your fleet.
  • In the beginning you need to specify many parameters to get you instances launched
    We launched launch templates, encapsulate all these parameters into an EC2 resources, which parameters are optional, which are editable etc.
    You can say that resources can only be created with certain launch templates

    Segway -
  • Segway -
  • Pro tip use Cloudwatch, identify issues and mitigate them before they are noticed by your customers
  • Many customers have capabilities on premises for patch, config drift etc
    Can now do in cloud native way to accomplish they management tasks that they have done on premise with a variety of tools
    There is no additional cost for using the AWS Systems Manager, you pay only for the underlying resources managed and created by AWS Systems Manager

    Segway – so that covers a sample of our management tools, click , now let me tell you about the purchase options you have available to you for EC2
  • Segway -
  • Will cover the highlights on how you should think about purchasing EC2 capacity
    Ondemand – getting started, you dont know what you need
    Reserved – when you know your workload
    Spot – unused capacity

    Segway -
  • Feedback we got , historically RI1 gives you 1) Capacity reservation and 2) discount
    Cust were only interested in discount, we said if they waive your capacity reservation we will let your RIs float across the Azs and instance size
    Segway -
  • How do I know how much RIs to buy
    RI recommendations in Cost Explorer
    Segway – we will talk about Spot
  • Let you take advantage of excess capacity by purchasing at a discount
    Customers typically use sot for workloads that they wouldn’t otherwise pay for, things like the 30th or 40th simulation, how much is that worth to you.
    Catch is 2 min warning when we need the capacity when we need it during peak times
    New - Eliminate the bid: Predictive pricing - 70 - 90 discount on price, no longer need to bid, you will get the instance, and you will know its at a good price.
    New - Spot acquisition experience has been integrated into the run instances API , pass a market parameter to use spot instances and if they are available you will get them
    New - Hibernate Spot Instance, Can now hibernate the Spot instance when you get this 2 minute warning. We will intercept warning, freeze the running state of the instance to disk like you would if you shut laptop lid, when the capacity is available again we will hydrate that into your spot instance. This makes spot much mores accessible to a broader set of use case

    Segway -
  • Which to use? All of the above.

    Segway -
  • So we have completed the loop, we talked about EC2 in terms of it Resources, how we implement high availability, Management and Purchasing option
    Segway - I hope you have learned something and most important of all that you enjoyed todays session

  • Ill be available in the ask the architect booths today and tomorrow, come and introduce yourself and bring any questions you have on this or anything else AWS
  • ×