SlideShare a Scribd company logo
Running BSD
on AWS
Julien Simon, Principal Technical Evangelist
@julsimon
Nicolas David, EMEA Technical Trainer
@nuage_ninja
Who we are
Nicolas
https://nuage.ninja
@nuage_ninja
Julien
https://medium.com/@julsimon
@julsimon
Agenda
• AWS Infrastructure
• Instances, VMs, OSes
• Building BSD AMIs
• Benchmarking ‘buildworld’
• Q&A
AWS Infrastructure
AWS Global Infrastructure
16 Regions
44 Availability Zones
91 Edge Locations
https://aws.amazon.com/about-aws/global-infrastructure/
Example AWS Region
AZ
AZ
AZ AZ AZ
Transit
Transit
AWS re:Invent 2016 | Tuesday Night Live with James Hamilton
https://www.youtube.com/watch?v=AyOAjFNPAbA
Example AWS Availability Zone
AZ
AZ
AZ AZ AZ
Transit
Transit
Example AWS Data Center
Instances, Virtual Machines
& Operating Systems
Amazon EC2
• Infrastructure as a Service, launched in 2006
• Virtual machines (“EC2 instances”) and images (“Amazon Machine Image”, “AMI”)
• Amazon AMIs, vendor AMIs (“EC2 Marketplace”), community AMIs, or your own
• All-inclusive: networking (Virtual Private Cloud), storage (Elastic Block Storage),
firewalling (Security Group), load balancing (Elastic Load Balancing), high availability
(Availability Zones), automatic scaling (Auto Scaling groups), monitoring
(Cloudwatch)
• Until now, pay on an hourly basis. Reserved Instances and Spot for large savings
• Starting October 2nd, pay on a per second basis, minimum 1 minute (boot time)
https://aws.amazon.com/ec2/
http://aws.amazon.com/free/
https://aws.amazon.com/ec2/pricing/reserved-instances/
https://aws.amazon.com/ec2/spot/
« I can get less expensive VMs at X, Y or Z »
• Comparing apples and oranges?
• Take a long hard look at:
– Geographical coverage
– Width and depth of technical services
– High availability: not all “regions” are born equal
– Scalability
– Security
– Compliance
05/01/16 https://aws.amazon.com/blogs/aws/happy-new-year-ec2-price-reduction-c4-m4-and-r3-instances/
11/08/16 https://aws.amazon.com/blogs/aws/amazon-elastic-block-store-ebs-update-snapshot-price-reduction-more-piopsgib/
14/11/16 https://aws.amazon.com/blogs/aws/ec2-price-reduction-c4-m4-and-t2-instances/
03/05/17 https://aws.amazon.com/blogs/aws/ec2-price-reductions-reserved-instances-m4-instances/
Instances Types
<Family><Generation>.<Size>, e.g. m4.xlarge
General purpose: t2 (burstable), m4
Compute-optimized: c4
Storage-optimized: i3 (I/O), d2 (Density)
Memory-optimized: r4, x1
GPU: g3, p2
t2.nano: 1 vCPU, 512MB RAM, EBS storage
x1e.32xlarge: 128 vCPU, 4TB RAM, 2x1920 GB SSD, 25Gb network
https://aws.amazon.com/ec2/instance-types/
AWS EC2 Instances with Intel Xeon
AWS
Instance
Type
High
Memory
X1
Compute-
Optimized
C4
Storage-
Optimized
D2
General
Purpose
M4
Memory-
Optimized
R4
IO-
Optimized
I3
Graphics-
Optimized
G3
Burstable
Performance
T2
Intel Xeon
Processor
E7-8880 v3 E5-2666 v3 E5-2676 v3
E5-2686 v4
E5-2676 v3
E5-2686 v4 E5-2686 v4 E5-2686 v4
Intel Xeon
Family
Intel AVX AVX 2.0 AVX 2.0 AVX 2.0 AVX 2.0 Yes Yes Yes Yes
Intel AES-NI Yes Yes Yes Yes Yes Yes Yes No
Intel Turbo
Boost
Yes Yes Yes Yes Yes Yes Yes Yes
Intel TSX Yes No No No No No No No
Per core P-
and C-state
control
No
Yes
(8xlarge
only)
No No No No No No
SSD
Storage
EBS
Optimized by
default
EBS
Optimized by
default
No
EBS:
Optimized by
default
Yes Yes
EBS:
Optimized by
default
EBS only
X1e Instance - Tons of Memory
• Features up to 4TB of memory and 128 vCPU.
• Uses Intel E7 v3 Haswell processors.
• Features the new generation 25Gbps ENI.
• Is designed for demanding enterprise workloads,
including production installations of SAP HANA,
Microsoft SQL Server, Apache Spark, and Presto.
C4 Instance - Tons of Compute
• Uses Intel Xeon E5-2666 v3 Haswell processors
@ 2.9Ghz
• Features up to 36 vCPU and 60GB of RAM
• Designed to deliver maximum single core
performance (C-state, P-state, TurboBoost up to
3.5GHz) to compute-intensive applications
I3 Instance - Tons of I/O
• Uses Intel Xeon E5-2686 v4 Broadwell processors @ 2.3Ghz
• Features up to 488GB of memory, 64 vCPU.
• Up to 15.2TB of SSD Storage: 8 NVMe SSDs, 3.3M IOPS
• Features the new generation 25Gbps ENI.
• Is designed for high throughput and low latency including
relational databases, NoSQL databases, search engines, data
warehouses, real-time analytics, and disk-based caches.
AWS EC2 Storage Options – EBS Volumes
Solid-State Drives (SSD) Hard Disk Drives (HDD)
General Purpose SSD Provisioned IOPS SSD
Throughput-Optimized
HDD Cold HDD
Max volume size 16 TiB 16 TiB 16 TiB 16 TiB
Max IOPS/volume 10,000 20,000 500 250
Max
throughput/volume 160 MiB/s 320 MiB/s 500 MiB/s 250 MiB/s
Use cases
• Recommended for most
workloads
• System boot volumes
• Virtual desktops
• Low-latency interactive
apps
• Development and test
environments
• I/O-intensive workloads
• Relational DBs
• NoSQL DBs
• Streaming workloads
requiring consistent, fast
throughput at a low price
• Big data
• Data warehouses
• Log processing
• Cannot be a boot volume
• Throughput-oriented
storage for large
volumes of data that
is infrequently
accessed
• Scenarios where the
lowest storage cost is
important
• Cannot be a boot
volume
AWS EC2 Storage Options – Instance Storage
• Is local, complimentary direct attached block storage.
• Includes availability, number of disks, and size based
on EC2 instance type.
• Is optimized for up to 3.3M IOPS.
• Is SSD or magnetic.
• Has no persistence.
• Automatically deletes data when an EC2 instance
stops, fails or is terminated.
Benchmarking ‘buildworld’
C4 X1 I3
RAM 60GB 2TB 488GB
Compute 36 vCPU
Intel Haswell
128 vCPU
Intel Haswell
64 vCPU
Intel Broadwell
Storage EBS
10k PIOPS SSD
Instance store
2 x 1.92TB SSD
Instance store
8 x 1.92TB NVMe
Filesystem UFS UFS ZFS
(2 4-disk pools)
SSD time
Ramdisk time
Price
Benchmarking ‘buildworld’
• x1.32xlarge
/usr/src on first local SSD, /usr/obj on second local SSD (ufs)
make –j128
• c4.8xlarge
/usr/src and /usr/obj on same EBS volume (SSD, 10k IOPS, ufs)
make –j36
• i3.16xlarge
/usr/src and /usr/obj on 2 ZFS pools (4 local NVMe SSD each)
make –j64
Building BSD AMIs
130+ UNIX AMIs on the AWS Marketplace
https://aws.amazon.com/marketplace/
Baking your own AMI
• AWS CLI/AWS Shell CLI: aws ec2 create-image.
• Aminator: Netflix tool, EC2 only for Red Hat and
CentOS.
• Packer: Hashicorp tool, more features.
http://docs.aws.amazon.com/cli/latest/reference/ec2/create-image.html
https://github.com/Netflix/aminator
https://www.packer.io
BSD AMI factory – 1/4
https://aws.amazon.com/getting-started/projects/set-up-ci-cd-pipeline/
BSD AMI factory – 2/4
• OpenBSD host w/ 12gb available (ami+4gb tmp
files)
• create-ami.sh (Thanks ajacoutot@ !)
– doas, curl, ec2-api-tools, awscli and vmdktool packages installed.
– AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY set
– MIRROR set to the closest AWS Region
BSD AMI factory – 3/4
BSD AMI factory – 4/4
DevOps for AMIs !
1. Commit to git
2. Bake the AMI
3. Notify Teams & Code Pipeline
4. Deploy Infrastructure for UAT Environment + use new AMI
5. Test !
1. Security/Compliance ? AWS Inspector
2. Load ? Bees with Machine Guns
3. Other stuff ? Features, Load+Security, etc.
6. Move on to Production once UAT results are satisfactory
http://davini.co/ebc-openbsd-2017
https://github.com/newsapps/beeswithmachineguns
Takeways
• DevOps is for AMIs, also for Containers
• No servers, just services
• Security is for everyone, and everywhere
• Pay by the usage
How you can help
1. Test FreeBSD on AWS and report issues
2. Write « Getting started with FreeBSD on AWS »
3. Work on « instant server » metaports
– Instant webserver, instant Wordpress, etc.
Get in touch with Colin Percival
<cperciva@freebsd.org>
Benchmarking results
C4 X1 I3
RAM 60GB 2TB 488GB
Compute 36 vCPU
Intel Haswell
128 vCPU
Intel Haswell
64 vCPU
Intel Broadwell
Storage EBS
10k PIOPS SSD
Instance store
2 x 1.92TB SSD
Instance store
8 x 1.92TB NVMe
Filesystem UFS UFS ZFS
(2 4-disk pools)
SSD time 11mn 39s 11mn 40s 10mn 58s
Ramdisk time
Price
C4 X1 I3
RAM 60GB 2TB 488GB
Compute 36 vCPU
Intel Haswell
128 vCPU
Intel Haswell
64 vCPU
Intel Broadwell
Storage EBS
10k PIOPS SSD
Instance store
2 x 1.92TB SSD
Instance store
8 x 1.92TB NVMe
Filesystem UFS UFS ZFS
(2 4-disk pools)
SSD time 11mn 39s 11mn 40s 10mn 58s
Ramdisk time 11mn 10s 11mn 26 11mn 07s
Price
C4 X1 I3
RAM 60GB 2TB 488GB
Compute 36 vCPU
Intel Haswell
128 vCPU
Intel Haswell
64 vCPU
Intel Broadwell
Storage EBS
10k PIOPS SSD
Instance store
2 x 1.92TB SSD
Instance store
8 x 1.92TB NVMe
Filesystem UFS UFS ZFS
(2 4-disk pools)
SSD time 11mn 39s 11mn 40s 10mn 58s
Ramdisk time 11mn 10s 11mn 26 11mn 07s
Price $1.591 $13.338 $4.992
Conclusion
Running BSD on AWS
AWS is a rich and lively environment for BSD and Open Source platforms
Your choice: DIY, Marketplace, Partners, AWS Managed Services
The tools & projects you love, without the infrastructure drama
Built-in high availability, scalability, security & compliance
Focus on creativity and productivity, not on plumbing
AWS User Groups
Lille
Paris
Rennes
Nantes
Bordeaux
Lyon
Montpellier
Toulouse
Côte d’Azur
Grand Est
@aws_actus
Thank you!
Julien Simon, Principal Technical Evangelist
@julsimon
Nicolas David, EMEA Technical Trainer
@nuage_ninja

More Related Content

PPTX
Advanced Scheduling with Amazon ECS (September 2017)
PPTX
Picking the right AWS backend for your application (September 2017)
PDF
Deep Learning with AWS (November 2016)
PPTX
FPGAs in the cloud? (October 2017)
PDF
Docker clusters on AWS with Amazon ECS and Kubernetes
PDF
Deep Dive on Amazon EC2 Instances (March 2017)
PDF
Picking the right AWS backend for your Java application (May 2017)
PDF
Picking the right AWS backend for your Java application (May 2017)
Advanced Scheduling with Amazon ECS (September 2017)
Picking the right AWS backend for your application (September 2017)
Deep Learning with AWS (November 2016)
FPGAs in the cloud? (October 2017)
Docker clusters on AWS with Amazon ECS and Kubernetes
Deep Dive on Amazon EC2 Instances (March 2017)
Picking the right AWS backend for your Java application (May 2017)
Picking the right AWS backend for your Java application (May 2017)

What's hot (7)

PDF
AWS EC2
PDF
Advanced Task Scheduling with Amazon ECS (June 2017)
PDF
AWS meetup「Apache Spark on EMR」
PPTX
Amazon ec2
PPTX
AWS re:Invent 2016 : announcement, technical demos and feedbacks
PPTX
Build, train, and deploy Machine Learning models at scale (May 2018)
PDF
Amazon Ec2 Application Design
AWS EC2
Advanced Task Scheduling with Amazon ECS (June 2017)
AWS meetup「Apache Spark on EMR」
Amazon ec2
AWS re:Invent 2016 : announcement, technical demos and feedbacks
Build, train, and deploy Machine Learning models at scale (May 2018)
Amazon Ec2 Application Design
Ad

Similar to Running BSD on AWS (14)

PPTX
Migrating enterprise workloads to AWS
PDF
AWS Developer Fundamentals
PPTX
Re invent announcements_2016_hcls_use_cases_mchampion
PDF
Amazon Web Services (AWS) : Fundamentals
PPTX
CloudOverviewAWS.pptx
PPTX
Basic ppt on cloud computing on amazon web
PPTX
Best Practices for running the Oracle Database on EC2 webinar
PPTX
Amazon EC2 Instance Types.pptx
PDF
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
PDF
The iot academy_awstraining_part1_aws_introduction
PPTX
Debian Cloud - building the Debian AMIs
PDF
Oracle Cloud Infrastructure – Compute
PPT
Centricity EMRCPS_Platform_Architecture_Performance
PDF
Kubernetes for HCL Connections Component Pack - Build or Buy?
Migrating enterprise workloads to AWS
AWS Developer Fundamentals
Re invent announcements_2016_hcls_use_cases_mchampion
Amazon Web Services (AWS) : Fundamentals
CloudOverviewAWS.pptx
Basic ppt on cloud computing on amazon web
Best Practices for running the Oracle Database on EC2 webinar
Amazon EC2 Instance Types.pptx
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
The iot academy_awstraining_part1_aws_introduction
Debian Cloud - building the Debian AMIs
Oracle Cloud Infrastructure – Compute
Centricity EMRCPS_Platform_Architecture_Performance
Kubernetes for HCL Connections Component Pack - Build or Buy?
Ad

More from Julien SIMON (20)

PDF
Implementing high-quality and cost-effiient AI applications with small langua...
PDF
Optimiser vos workloads AI/ML sur Amazon EC2 et AWS Graviton
PDF
Trying to figure out MCP by actually building an app from scratch with open s...
PDF
Arcee AI - building and working with small language models (06/25)
PDF
deep_dive_multihead_latent_attention.pdf
PDF
Deep Dive: Model Distillation with DistillKit
PDF
Deep Dive: Parameter-Efficient Model Adaptation with LoRA and Spectrum
PDF
Building High-Quality Domain-Specific Models with Mergekit
PDF
Tailoring Small Language Models for Enterprise Use Cases
PDF
Tailoring Small Language Models for Enterprise Use Cases
PDF
Julien Simon - Deep Dive: Compiling Deep Learning Models
PDF
Tailoring Small Language Models for Enterprise Use Cases
PDF
Julien Simon - Deep Dive - Optimizing LLM Inference
PDF
Julien Simon - Deep Dive - Accelerating Models with Better Attention Layers
PDF
Julien Simon - Deep Dive - Quantizing LLMs
PDF
Julien Simon - Deep Dive - Model Merging
PDF
An introduction to computer vision with Hugging Face
PDF
Reinventing Deep Learning
 with Hugging Face Transformers
PDF
Building NLP applications with Transformers
PPTX
Building Machine Learning Models Automatically (June 2020)
Implementing high-quality and cost-effiient AI applications with small langua...
Optimiser vos workloads AI/ML sur Amazon EC2 et AWS Graviton
Trying to figure out MCP by actually building an app from scratch with open s...
Arcee AI - building and working with small language models (06/25)
deep_dive_multihead_latent_attention.pdf
Deep Dive: Model Distillation with DistillKit
Deep Dive: Parameter-Efficient Model Adaptation with LoRA and Spectrum
Building High-Quality Domain-Specific Models with Mergekit
Tailoring Small Language Models for Enterprise Use Cases
Tailoring Small Language Models for Enterprise Use Cases
Julien Simon - Deep Dive: Compiling Deep Learning Models
Tailoring Small Language Models for Enterprise Use Cases
Julien Simon - Deep Dive - Optimizing LLM Inference
Julien Simon - Deep Dive - Accelerating Models with Better Attention Layers
Julien Simon - Deep Dive - Quantizing LLMs
Julien Simon - Deep Dive - Model Merging
An introduction to computer vision with Hugging Face
Reinventing Deep Learning
 with Hugging Face Transformers
Building NLP applications with Transformers
Building Machine Learning Models Automatically (June 2020)

Recently uploaded (20)

PDF
The Rise and Fall of 3GPP – Time for a Sabbatical?
PPTX
MYSQL Presentation for SQL database connectivity
PPTX
20250228 LYD VKU AI Blended-Learning.pptx
PPTX
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
PDF
Building Integrated photovoltaic BIPV_UPV.pdf
PDF
Review of recent advances in non-invasive hemoglobin estimation
PDF
Dropbox Q2 2025 Financial Results & Investor Presentation
PDF
CIFDAQ's Market Insight: SEC Turns Pro Crypto
PDF
NewMind AI Monthly Chronicles - July 2025
PDF
Machine learning based COVID-19 study performance prediction
PDF
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
PDF
Reach Out and Touch Someone: Haptics and Empathic Computing
PDF
Diabetes mellitus diagnosis method based random forest with bat algorithm
PDF
Per capita expenditure prediction using model stacking based on satellite ima...
PPTX
Cloud computing and distributed systems.
PPT
Teaching material agriculture food technology
PDF
Approach and Philosophy of On baking technology
PPT
“AI and Expert System Decision Support & Business Intelligence Systems”
PDF
Bridging biosciences and deep learning for revolutionary discoveries: a compr...
PDF
Peak of Data & AI Encore- AI for Metadata and Smarter Workflows
The Rise and Fall of 3GPP – Time for a Sabbatical?
MYSQL Presentation for SQL database connectivity
20250228 LYD VKU AI Blended-Learning.pptx
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
Building Integrated photovoltaic BIPV_UPV.pdf
Review of recent advances in non-invasive hemoglobin estimation
Dropbox Q2 2025 Financial Results & Investor Presentation
CIFDAQ's Market Insight: SEC Turns Pro Crypto
NewMind AI Monthly Chronicles - July 2025
Machine learning based COVID-19 study performance prediction
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
Reach Out and Touch Someone: Haptics and Empathic Computing
Diabetes mellitus diagnosis method based random forest with bat algorithm
Per capita expenditure prediction using model stacking based on satellite ima...
Cloud computing and distributed systems.
Teaching material agriculture food technology
Approach and Philosophy of On baking technology
“AI and Expert System Decision Support & Business Intelligence Systems”
Bridging biosciences and deep learning for revolutionary discoveries: a compr...
Peak of Data & AI Encore- AI for Metadata and Smarter Workflows

Running BSD on AWS

  • 1. Running BSD on AWS Julien Simon, Principal Technical Evangelist @julsimon Nicolas David, EMEA Technical Trainer @nuage_ninja
  • 3. Agenda • AWS Infrastructure • Instances, VMs, OSes • Building BSD AMIs • Benchmarking ‘buildworld’ • Q&A
  • 5. AWS Global Infrastructure 16 Regions 44 Availability Zones 91 Edge Locations https://aws.amazon.com/about-aws/global-infrastructure/
  • 6. Example AWS Region AZ AZ AZ AZ AZ Transit Transit AWS re:Invent 2016 | Tuesday Night Live with James Hamilton https://www.youtube.com/watch?v=AyOAjFNPAbA
  • 7. Example AWS Availability Zone AZ AZ AZ AZ AZ Transit Transit
  • 9. Instances, Virtual Machines & Operating Systems
  • 10. Amazon EC2 • Infrastructure as a Service, launched in 2006 • Virtual machines (“EC2 instances”) and images (“Amazon Machine Image”, “AMI”) • Amazon AMIs, vendor AMIs (“EC2 Marketplace”), community AMIs, or your own • All-inclusive: networking (Virtual Private Cloud), storage (Elastic Block Storage), firewalling (Security Group), load balancing (Elastic Load Balancing), high availability (Availability Zones), automatic scaling (Auto Scaling groups), monitoring (Cloudwatch) • Until now, pay on an hourly basis. Reserved Instances and Spot for large savings • Starting October 2nd, pay on a per second basis, minimum 1 minute (boot time) https://aws.amazon.com/ec2/ http://aws.amazon.com/free/ https://aws.amazon.com/ec2/pricing/reserved-instances/ https://aws.amazon.com/ec2/spot/
  • 11. « I can get less expensive VMs at X, Y or Z » • Comparing apples and oranges? • Take a long hard look at: – Geographical coverage – Width and depth of technical services – High availability: not all “regions” are born equal – Scalability – Security – Compliance 05/01/16 https://aws.amazon.com/blogs/aws/happy-new-year-ec2-price-reduction-c4-m4-and-r3-instances/ 11/08/16 https://aws.amazon.com/blogs/aws/amazon-elastic-block-store-ebs-update-snapshot-price-reduction-more-piopsgib/ 14/11/16 https://aws.amazon.com/blogs/aws/ec2-price-reduction-c4-m4-and-t2-instances/ 03/05/17 https://aws.amazon.com/blogs/aws/ec2-price-reductions-reserved-instances-m4-instances/
  • 12. Instances Types <Family><Generation>.<Size>, e.g. m4.xlarge General purpose: t2 (burstable), m4 Compute-optimized: c4 Storage-optimized: i3 (I/O), d2 (Density) Memory-optimized: r4, x1 GPU: g3, p2 t2.nano: 1 vCPU, 512MB RAM, EBS storage x1e.32xlarge: 128 vCPU, 4TB RAM, 2x1920 GB SSD, 25Gb network https://aws.amazon.com/ec2/instance-types/
  • 13. AWS EC2 Instances with Intel Xeon AWS Instance Type High Memory X1 Compute- Optimized C4 Storage- Optimized D2 General Purpose M4 Memory- Optimized R4 IO- Optimized I3 Graphics- Optimized G3 Burstable Performance T2 Intel Xeon Processor E7-8880 v3 E5-2666 v3 E5-2676 v3 E5-2686 v4 E5-2676 v3 E5-2686 v4 E5-2686 v4 E5-2686 v4 Intel Xeon Family Intel AVX AVX 2.0 AVX 2.0 AVX 2.0 AVX 2.0 Yes Yes Yes Yes Intel AES-NI Yes Yes Yes Yes Yes Yes Yes No Intel Turbo Boost Yes Yes Yes Yes Yes Yes Yes Yes Intel TSX Yes No No No No No No No Per core P- and C-state control No Yes (8xlarge only) No No No No No No SSD Storage EBS Optimized by default EBS Optimized by default No EBS: Optimized by default Yes Yes EBS: Optimized by default EBS only
  • 14. X1e Instance - Tons of Memory • Features up to 4TB of memory and 128 vCPU. • Uses Intel E7 v3 Haswell processors. • Features the new generation 25Gbps ENI. • Is designed for demanding enterprise workloads, including production installations of SAP HANA, Microsoft SQL Server, Apache Spark, and Presto.
  • 15. C4 Instance - Tons of Compute • Uses Intel Xeon E5-2666 v3 Haswell processors @ 2.9Ghz • Features up to 36 vCPU and 60GB of RAM • Designed to deliver maximum single core performance (C-state, P-state, TurboBoost up to 3.5GHz) to compute-intensive applications
  • 16. I3 Instance - Tons of I/O • Uses Intel Xeon E5-2686 v4 Broadwell processors @ 2.3Ghz • Features up to 488GB of memory, 64 vCPU. • Up to 15.2TB of SSD Storage: 8 NVMe SSDs, 3.3M IOPS • Features the new generation 25Gbps ENI. • Is designed for high throughput and low latency including relational databases, NoSQL databases, search engines, data warehouses, real-time analytics, and disk-based caches.
  • 17. AWS EC2 Storage Options – EBS Volumes Solid-State Drives (SSD) Hard Disk Drives (HDD) General Purpose SSD Provisioned IOPS SSD Throughput-Optimized HDD Cold HDD Max volume size 16 TiB 16 TiB 16 TiB 16 TiB Max IOPS/volume 10,000 20,000 500 250 Max throughput/volume 160 MiB/s 320 MiB/s 500 MiB/s 250 MiB/s Use cases • Recommended for most workloads • System boot volumes • Virtual desktops • Low-latency interactive apps • Development and test environments • I/O-intensive workloads • Relational DBs • NoSQL DBs • Streaming workloads requiring consistent, fast throughput at a low price • Big data • Data warehouses • Log processing • Cannot be a boot volume • Throughput-oriented storage for large volumes of data that is infrequently accessed • Scenarios where the lowest storage cost is important • Cannot be a boot volume
  • 18. AWS EC2 Storage Options – Instance Storage • Is local, complimentary direct attached block storage. • Includes availability, number of disks, and size based on EC2 instance type. • Is optimized for up to 3.3M IOPS. • Is SSD or magnetic. • Has no persistence. • Automatically deletes data when an EC2 instance stops, fails or is terminated.
  • 19. Benchmarking ‘buildworld’ C4 X1 I3 RAM 60GB 2TB 488GB Compute 36 vCPU Intel Haswell 128 vCPU Intel Haswell 64 vCPU Intel Broadwell Storage EBS 10k PIOPS SSD Instance store 2 x 1.92TB SSD Instance store 8 x 1.92TB NVMe Filesystem UFS UFS ZFS (2 4-disk pools) SSD time Ramdisk time Price
  • 20. Benchmarking ‘buildworld’ • x1.32xlarge /usr/src on first local SSD, /usr/obj on second local SSD (ufs) make –j128 • c4.8xlarge /usr/src and /usr/obj on same EBS volume (SSD, 10k IOPS, ufs) make –j36 • i3.16xlarge /usr/src and /usr/obj on 2 ZFS pools (4 local NVMe SSD each) make –j64
  • 22. 130+ UNIX AMIs on the AWS Marketplace https://aws.amazon.com/marketplace/
  • 23. Baking your own AMI • AWS CLI/AWS Shell CLI: aws ec2 create-image. • Aminator: Netflix tool, EC2 only for Red Hat and CentOS. • Packer: Hashicorp tool, more features. http://docs.aws.amazon.com/cli/latest/reference/ec2/create-image.html https://github.com/Netflix/aminator https://www.packer.io
  • 24. BSD AMI factory – 1/4 https://aws.amazon.com/getting-started/projects/set-up-ci-cd-pipeline/
  • 25. BSD AMI factory – 2/4 • OpenBSD host w/ 12gb available (ami+4gb tmp files) • create-ami.sh (Thanks ajacoutot@ !) – doas, curl, ec2-api-tools, awscli and vmdktool packages installed. – AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY set – MIRROR set to the closest AWS Region
  • 26. BSD AMI factory – 3/4
  • 27. BSD AMI factory – 4/4 DevOps for AMIs ! 1. Commit to git 2. Bake the AMI 3. Notify Teams & Code Pipeline 4. Deploy Infrastructure for UAT Environment + use new AMI 5. Test ! 1. Security/Compliance ? AWS Inspector 2. Load ? Bees with Machine Guns 3. Other stuff ? Features, Load+Security, etc. 6. Move on to Production once UAT results are satisfactory http://davini.co/ebc-openbsd-2017 https://github.com/newsapps/beeswithmachineguns
  • 28. Takeways • DevOps is for AMIs, also for Containers • No servers, just services • Security is for everyone, and everywhere • Pay by the usage
  • 29. How you can help 1. Test FreeBSD on AWS and report issues 2. Write « Getting started with FreeBSD on AWS » 3. Work on « instant server » metaports – Instant webserver, instant Wordpress, etc. Get in touch with Colin Percival <cperciva@freebsd.org>
  • 31. C4 X1 I3 RAM 60GB 2TB 488GB Compute 36 vCPU Intel Haswell 128 vCPU Intel Haswell 64 vCPU Intel Broadwell Storage EBS 10k PIOPS SSD Instance store 2 x 1.92TB SSD Instance store 8 x 1.92TB NVMe Filesystem UFS UFS ZFS (2 4-disk pools) SSD time 11mn 39s 11mn 40s 10mn 58s Ramdisk time Price
  • 32. C4 X1 I3 RAM 60GB 2TB 488GB Compute 36 vCPU Intel Haswell 128 vCPU Intel Haswell 64 vCPU Intel Broadwell Storage EBS 10k PIOPS SSD Instance store 2 x 1.92TB SSD Instance store 8 x 1.92TB NVMe Filesystem UFS UFS ZFS (2 4-disk pools) SSD time 11mn 39s 11mn 40s 10mn 58s Ramdisk time 11mn 10s 11mn 26 11mn 07s Price
  • 33. C4 X1 I3 RAM 60GB 2TB 488GB Compute 36 vCPU Intel Haswell 128 vCPU Intel Haswell 64 vCPU Intel Broadwell Storage EBS 10k PIOPS SSD Instance store 2 x 1.92TB SSD Instance store 8 x 1.92TB NVMe Filesystem UFS UFS ZFS (2 4-disk pools) SSD time 11mn 39s 11mn 40s 10mn 58s Ramdisk time 11mn 10s 11mn 26 11mn 07s Price $1.591 $13.338 $4.992
  • 36. AWS is a rich and lively environment for BSD and Open Source platforms Your choice: DIY, Marketplace, Partners, AWS Managed Services The tools & projects you love, without the infrastructure drama Built-in high availability, scalability, security & compliance Focus on creativity and productivity, not on plumbing
  • 38. Thank you! Julien Simon, Principal Technical Evangelist @julsimon Nicolas David, EMEA Technical Trainer @nuage_ninja