SlideShare ist ein Scribd-Unternehmen logo
1 von 25
Downloaden Sie, um offline zu lesen
Peter Buschman
Operation Unthinkable
November 28th, 2019
Storage Forum Leipzig
Software Defined Storage
About
● Majority part of Booking Holdings (formerly Priceline Group)
● One of the largest e-Commerce websites in the world
● The largest online accommodation website in the world
● >1.5 Million properties in 220+ countries and territories
● 1.55 Million room nights booked every 24 hours
● >15,000 employees in 198 offices in 70 countries
● 1000s of LUNs, NFS shares, and S3 buckets
● Managed by a storage team of only 4 people
(as of October, 2019)
Defining the Unthinkable
"Developing and supporting our own
enterprise storage hardware and software."
Attributes of Enterprise Storage
●Gigantic Chassis
●Specialized Power
●Custom Cooling
●Difficult to Service
●Massively Expensive
"Habbakuks" "Kugelpanzers"
●Non-Standard Chassis
●Complicated Supply Chain
●Little Documentation
●Single Purpose
●Relatively Cheap
LIMITED RE-USABILITY
The complexity of storage...
The complexity of storage...
How to calculate complexity
complexity = (u -1) x c / n
u: number of rack-units
c: number of cables
n: number of racks
How complex is this?
Attributes of SDS Hardware
●Compact Chassis
●Standard Form Factor
●Standard Power
●Standard Cooling
●Easy to Service
"Shermans and T34s"
●Extensible Design
●Off the Shelf Ordering
●Optimized Supply Chain
●Multi Purpose
●Cost Effective
MAXIMUM RE-USABILITY
Re-thinking the problem...
??? 2U Empty Space
Empty 100Gb Switch Ports
Building a BOMB
● 2U max height
● 90cm max depth
● ~500W power
● 100Gb networking
● Non-volatile memory
● NVMe and high-capacity disk
● Maximize Terabytes/Watt
● Broadest software eco-system possible
● Simple enough for remote hands to install
???
Thinking inside the box....
(Bill of Materials)
The plan "crazy enough to work"
Image credit: Dutch National Archives Image credit: U.S. Public Domain
The Booking BOM Gen1 "SFF" Intel
● 2 x Intel Skylake 6146 CPUs
● 12 x 32GB DDR4 RAM (384GB total)
● 12 x 16GB NVDIMM-N (192GB total)
● 1 x 100Gb NIC
● 4 - 16 x 15.36TB (245.76TB total)
● 1 x HHHL FPGA storage accelerator (optional)
● Tool-less L-bracket rails
● Color-coded C13 to C14 power-cords
● 50cm and 1m network cables in the box
● Installs in < 5mins
The Booking BOM Gen2 "SFF" AMD
● 1 x AMD EPYC Rome 7402P 32-core CPU
● 12 x 128GB DDR4 RAM (1.5TB total)
● 4 x 32GB NVDIMM-N (128GB total)
● 1 x 100Gb NIC
● 4 - 24 x 15.36TB U.2 NVMe (368.64TB max)
● 1 x HHHL FPGA storage accelerator (optional)
● Tool-less L-bracket rails
● Color-coded C13 to C14 power-cords
● 50cm and 1m network cables in the box
● Installs in < 5mins
The Booking BOM Gen1 "LFF"
● 2 x Intel Skylake 6132 CPUs
● 12 x 32GB DDR4 RAM (384GB total)
● 12 x 16GB NVDIMM-N (192GB total)
● 1 x 100Gb NIC
● 12 x 14TB 7200rpm SATA disks (224TB total)
● 2 x 15.36TB HHHL NVMe (30.72 TB total)
● Tool-less L-bracket rails
● Color-coded C13 to C14 power-cords
● 50cm and 1m network cables in the box
● Installs in < 5mins
Tool-less drive caddies
2.5" Small Form Factor (SFF)
Tool-less drive caddies
3.5" Large Form Factor (LFF)
Re-thinking the solution...
12 + 4 = 16 x 14TB = 224TB Disk Capacity
2 x 15.36 = 30.72TB NVMe SSD Capacity
12 x 14TB Disks 4 x 14TB Disks
1 x 100Gb NIC
2 x 15.36TB NVMe
How complex is this?
1 2
3 4
4 cables x (2U - 1) / 1 rack = factor 4 complexity
Lots of failures...
● Ruined €100k of 11TB NVMe SSDs
● Early all-NVMe chassis were very buggy
● Painful lessons from early CPU choices
● Failed miserably with RDMA / RoCE
● Lots of NIC, BIOS, and SSD firmware issues
● Lots of SW + HW support matrix issues
● Struggles with vendor communications
● Struggles with vendor business cycles
● ...but we kept iterating!
Image credit: Imperial War Museum, UK © IWM (NA 13896)
What we have achieved so far...
● Deployed ~100 2U storage nodes
● Eliminated dedicated storage racks
● Cut power draw by over 50%
● Gone "all in" on software defined storage
● Eliminated storage hardware maintenance
● Switched entirely to software subscriptions
● Increased utilization while reducing costs
● Put total storage spending on a downward trend
● ...despite continued high data growth!
You need Allies!
What is your Unthinkable?
Image credit: Michael Coppins - Wikimedia Commons
● Plans you do not want to execute can yield great outcomes
● Define your own Unthinkable when it comes to SDS
● Build a BOM and don't compromise
● Recruit allies - you will need them!
● Expect failure and work through it
● Take your time
Closing.thoughts
Operation Unthinkable – Software Defined Storage @ Booking.com (Peter Buschman)

Weitere ähnliche Inhalte

Was ist angesagt?

Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...
Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...
Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...
DataStax
 

Was ist angesagt? (20)

Disaster Recovery with MySQL InnoDB ClusterSet - What is it and how do I use it?
Disaster Recovery with MySQL InnoDB ClusterSet - What is it and how do I use it?Disaster Recovery with MySQL InnoDB ClusterSet - What is it and how do I use it?
Disaster Recovery with MySQL InnoDB ClusterSet - What is it and how do I use it?
 
[2018] MySQL 이중화 진화기
[2018] MySQL 이중화 진화기[2018] MySQL 이중화 진화기
[2018] MySQL 이중화 진화기
 
Almost Perfect Service Discovery and Failover with ProxySQL and Orchestrator
Almost Perfect Service Discovery and Failover with ProxySQL and OrchestratorAlmost Perfect Service Discovery and Failover with ProxySQL and Orchestrator
Almost Perfect Service Discovery and Failover with ProxySQL and Orchestrator
 
MySQL Replication Performance Tuning for Fun and Profit!
MySQL Replication Performance Tuning for Fun and Profit!MySQL Replication Performance Tuning for Fun and Profit!
MySQL Replication Performance Tuning for Fun and Profit!
 
Oracle 12c and its pluggable databases
Oracle 12c and its pluggable databasesOracle 12c and its pluggable databases
Oracle 12c and its pluggable databases
 
PostgreSQL Security. How Do We Think?
PostgreSQL Security. How Do We Think?PostgreSQL Security. How Do We Think?
PostgreSQL Security. How Do We Think?
 
PMM database open source monitoring solution
PMM database open source monitoring solutionPMM database open source monitoring solution
PMM database open source monitoring solution
 
The Parquet Format and Performance Optimization Opportunities
The Parquet Format and Performance Optimization OpportunitiesThe Parquet Format and Performance Optimization Opportunities
The Parquet Format and Performance Optimization Opportunities
 
Oracle 12c Multitenant architecture
Oracle 12c Multitenant architectureOracle 12c Multitenant architecture
Oracle 12c Multitenant architecture
 
binary log と 2PC と Group Commit
binary log と 2PC と Group Commitbinary log と 2PC と Group Commit
binary log と 2PC と Group Commit
 
Deep Dive into GPU Support in Apache Spark 3.x
Deep Dive into GPU Support in Apache Spark 3.xDeep Dive into GPU Support in Apache Spark 3.x
Deep Dive into GPU Support in Apache Spark 3.x
 
Best Practices for the Most Impactful Oracle Database 18c and 19c Features
Best Practices for the Most Impactful Oracle Database 18c and 19c FeaturesBest Practices for the Most Impactful Oracle Database 18c and 19c Features
Best Practices for the Most Impactful Oracle Database 18c and 19c Features
 
BlueStore: a new, faster storage backend for Ceph
BlueStore: a new, faster storage backend for CephBlueStore: a new, faster storage backend for Ceph
BlueStore: a new, faster storage backend for Ceph
 
SSD Deployment Strategies for MySQL
SSD Deployment Strategies for MySQLSSD Deployment Strategies for MySQL
SSD Deployment Strategies for MySQL
 
Tuning Autovacuum in Postgresql
Tuning Autovacuum in PostgresqlTuning Autovacuum in Postgresql
Tuning Autovacuum in Postgresql
 
Deep Dive on Amazon Aurora PostgreSQL Performance Tuning (DAT428-R1) - AWS re...
Deep Dive on Amazon Aurora PostgreSQL Performance Tuning (DAT428-R1) - AWS re...Deep Dive on Amazon Aurora PostgreSQL Performance Tuning (DAT428-R1) - AWS re...
Deep Dive on Amazon Aurora PostgreSQL Performance Tuning (DAT428-R1) - AWS re...
 
The Art of Monitoring (2016).pdf
The Art of Monitoring (2016).pdfThe Art of Monitoring (2016).pdf
The Art of Monitoring (2016).pdf
 
MySQL Load Balancers - Maxscale, ProxySQL, HAProxy, MySQL Router & nginx - A ...
MySQL Load Balancers - Maxscale, ProxySQL, HAProxy, MySQL Router & nginx - A ...MySQL Load Balancers - Maxscale, ProxySQL, HAProxy, MySQL Router & nginx - A ...
MySQL Load Balancers - Maxscale, ProxySQL, HAProxy, MySQL Router & nginx - A ...
 
Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...
Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...
Apache Cassandra Multi-Datacenter Essentials (Julien Anguenot, iLand Internet...
 
MySQL Database Architectures - MySQL InnoDB ClusterSet 2021-11
MySQL Database Architectures - MySQL InnoDB ClusterSet 2021-11MySQL Database Architectures - MySQL InnoDB ClusterSet 2021-11
MySQL Database Architectures - MySQL InnoDB ClusterSet 2021-11
 

Ähnlich wie Operation Unthinkable – Software Defined Storage @ Booking.com (Peter Buschman)

Boosting I/O Performance with KVM io_uring
Boosting I/O Performance with KVM io_uringBoosting I/O Performance with KVM io_uring
Boosting I/O Performance with KVM io_uring
ShapeBlue
 

Ähnlich wie Operation Unthinkable – Software Defined Storage @ Booking.com (Peter Buschman) (20)

High-Density Top-Loading Storage for Cloud Scale Applications
High-Density Top-Loading Storage for Cloud Scale Applications High-Density Top-Loading Storage for Cloud Scale Applications
High-Density Top-Loading Storage for Cloud Scale Applications
 
Modern network servers
Modern network serversModern network servers
Modern network servers
 
Kernel Recipes 2016 - Speeding up development by setting up a kernel build farm
Kernel Recipes 2016 - Speeding up development by setting up a kernel build farmKernel Recipes 2016 - Speeding up development by setting up a kernel build farm
Kernel Recipes 2016 - Speeding up development by setting up a kernel build farm
 
Ceph Day Tokyo -- Ceph on All-Flash Storage
Ceph Day Tokyo -- Ceph on All-Flash StorageCeph Day Tokyo -- Ceph on All-Flash Storage
Ceph Day Tokyo -- Ceph on All-Flash Storage
 
Storage Spaces Direct - the new Microsoft SDS star - Carsten Rachfahl
Storage Spaces Direct - the new Microsoft SDS star - Carsten RachfahlStorage Spaces Direct - the new Microsoft SDS star - Carsten Rachfahl
Storage Spaces Direct - the new Microsoft SDS star - Carsten Rachfahl
 
MySQL Cluster (NDB) - Best Practices Percona Live 2017
MySQL Cluster (NDB) - Best Practices Percona Live 2017MySQL Cluster (NDB) - Best Practices Percona Live 2017
MySQL Cluster (NDB) - Best Practices Percona Live 2017
 
Cassandra Day Chicago 2015: DataStax Enterprise & Apache Cassandra Hardware B...
Cassandra Day Chicago 2015: DataStax Enterprise & Apache Cassandra Hardware B...Cassandra Day Chicago 2015: DataStax Enterprise & Apache Cassandra Hardware B...
Cassandra Day Chicago 2015: DataStax Enterprise & Apache Cassandra Hardware B...
 
Workshop actualización SVG CESGA 2012
Workshop actualización SVG CESGA 2012 Workshop actualización SVG CESGA 2012
Workshop actualización SVG CESGA 2012
 
Open Source Data Deduplication
Open Source Data DeduplicationOpen Source Data Deduplication
Open Source Data Deduplication
 
Cy7 introduction
Cy7 introductionCy7 introduction
Cy7 introduction
 
Linux Stammtisch Munich: Ceph - Overview, Experiences and Outlook
Linux Stammtisch Munich: Ceph - Overview, Experiences and OutlookLinux Stammtisch Munich: Ceph - Overview, Experiences and Outlook
Linux Stammtisch Munich: Ceph - Overview, Experiences and Outlook
 
cachegrand: A Take on High Performance Caching
cachegrand: A Take on High Performance Cachingcachegrand: A Take on High Performance Caching
cachegrand: A Take on High Performance Caching
 
UNIT 2 P1
UNIT 2 P1UNIT 2 P1
UNIT 2 P1
 
Achieving the ultimate performance with KVM
Achieving the ultimate performance with KVMAchieving the ultimate performance with KVM
Achieving the ultimate performance with KVM
 
Boosting I/O Performance with KVM io_uring
Boosting I/O Performance with KVM io_uringBoosting I/O Performance with KVM io_uring
Boosting I/O Performance with KVM io_uring
 
Backup management with Ceph Storage - Camilo Echevarne, Félix Barbeira
Backup management with Ceph Storage - Camilo Echevarne, Félix BarbeiraBackup management with Ceph Storage - Camilo Echevarne, Félix Barbeira
Backup management with Ceph Storage - Camilo Echevarne, Félix Barbeira
 
Shootout at the PAAS Corral
Shootout at the PAAS CorralShootout at the PAAS Corral
Shootout at the PAAS Corral
 
Presentation for IGDCloud meetup: The clouds arena AWS ver. others
Presentation for IGDCloud meetup: The clouds arena AWS ver. othersPresentation for IGDCloud meetup: The clouds arena AWS ver. others
Presentation for IGDCloud meetup: The clouds arena AWS ver. others
 
Erasure codes and storage tiers on gluster
Erasure codes and storage tiers on glusterErasure codes and storage tiers on gluster
Erasure codes and storage tiers on gluster
 
The Power of HPC with Next Generation Supermicro Systems
The Power of HPC with Next Generation Supermicro Systems The Power of HPC with Next Generation Supermicro Systems
The Power of HPC with Next Generation Supermicro Systems
 

Mehr von data://disrupted®

Rook: Storage for Containers in Containers – data://disrupted® 2020
Rook: Storage for Containers in Containers  – data://disrupted® 2020Rook: Storage for Containers in Containers  – data://disrupted® 2020
Rook: Storage for Containers in Containers – data://disrupted® 2020
data://disrupted®
 
Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020
Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020
Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020
data://disrupted®
 
HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...
HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...
HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...
data://disrupted®
 

Mehr von data://disrupted® (18)

Benchmarking your cloud performance with top 4 global public clouds
Benchmarking your cloud performance with top 4 global public cloudsBenchmarking your cloud performance with top 4 global public clouds
Benchmarking your cloud performance with top 4 global public clouds
 
Achieving the Ultimate Performance with KVM
Achieving the Ultimate Performance with KVMAchieving the Ultimate Performance with KVM
Achieving the Ultimate Performance with KVM
 
​Muss es wirklich wieder Tape sein?
​Muss es wirklich wieder Tape sein? ​Muss es wirklich wieder Tape sein?
​Muss es wirklich wieder Tape sein?
 
​Tape-basierter Object-Storage als S3 Speicherklasse und Cloud-Absicherung
​Tape-basierter Object-Storage als S3 Speicherklasse und Cloud-Absicherung​Tape-basierter Object-Storage als S3 Speicherklasse und Cloud-Absicherung
​Tape-basierter Object-Storage als S3 Speicherklasse und Cloud-Absicherung
 
Rook: Storage for Containers in Containers – data://disrupted® 2020
Rook: Storage for Containers in Containers  – data://disrupted® 2020Rook: Storage for Containers in Containers  – data://disrupted® 2020
Rook: Storage for Containers in Containers – data://disrupted® 2020
 
Storage Benchmarks - Voodoo oder Wissenschaft? – data://disrupted® 2020
Storage Benchmarks - Voodoo oder Wissenschaft? – data://disrupted® 2020Storage Benchmarks - Voodoo oder Wissenschaft? – data://disrupted® 2020
Storage Benchmarks - Voodoo oder Wissenschaft? – data://disrupted® 2020
 
Datenspeicherung 2020 bis 2030 – immer noch auf Festplatten? – data://disrupt...
Datenspeicherung 2020 bis 2030 – immer noch auf Festplatten? – data://disrupt...Datenspeicherung 2020 bis 2030 – immer noch auf Festplatten? – data://disrupt...
Datenspeicherung 2020 bis 2030 – immer noch auf Festplatten? – data://disrupt...
 
Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020
Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020
Speichermedium Tape – Warum es keine Alternative gibt – data://disrupted® 2020
 
Ransomware: Ohne Air Gap & Tape sind Sie verloren! – data://disrupted® 2020
Ransomware: Ohne Air Gap & Tape sind Sie verloren! – data://disrupted® 2020Ransomware: Ohne Air Gap & Tape sind Sie verloren! – data://disrupted® 2020
Ransomware: Ohne Air Gap & Tape sind Sie verloren! – data://disrupted® 2020
 
HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...
HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...
HCI einfach einfach! IT-Infrastruktur wie ein Smartphone! – data://disrupted®...
 
Erasure coding stief.tech 2020-03
Erasure coding stief.tech 2020-03Erasure coding stief.tech 2020-03
Erasure coding stief.tech 2020-03
 
Nextcloud als On-Premises Lösung für hochsicheren Datenaustausch (Frank Karli...
Nextcloud als On-Premises Lösung für hochsicheren Datenaustausch (Frank Karli...Nextcloud als On-Premises Lösung für hochsicheren Datenaustausch (Frank Karli...
Nextcloud als On-Premises Lösung für hochsicheren Datenaustausch (Frank Karli...
 
Die IBM 3592 Speicherlösung: Ein Vorgeschmack auf die Zukunft (Anne Ingenhaag)
Die IBM 3592 Speicherlösung: Ein Vorgeschmack auf die Zukunft (Anne Ingenhaag)Die IBM 3592 Speicherlösung: Ein Vorgeschmack auf die Zukunft (Anne Ingenhaag)
Die IBM 3592 Speicherlösung: Ein Vorgeschmack auf die Zukunft (Anne Ingenhaag)
 
CANDIDATE EXPERIENCE – Was Bewerber tatsächlich erwarten.
CANDIDATE EXPERIENCE – Was Bewerber tatsächlich erwarten.CANDIDATE EXPERIENCE – Was Bewerber tatsächlich erwarten.
CANDIDATE EXPERIENCE – Was Bewerber tatsächlich erwarten.
 
Cloud/Object-basierte Datenspeicherung mit HSM/ILM in S3 Speicherklassen (Tho...
Cloud/Object-basierte Datenspeicherung mit HSM/ILM in S3 Speicherklassen (Tho...Cloud/Object-basierte Datenspeicherung mit HSM/ILM in S3 Speicherklassen (Tho...
Cloud/Object-basierte Datenspeicherung mit HSM/ILM in S3 Speicherklassen (Tho...
 
Buzzword Bingo Storage Edition 2019 (Wolfgang Stief)
Buzzword Bingo Storage Edition 2019 (Wolfgang Stief)Buzzword Bingo Storage Edition 2019 (Wolfgang Stief)
Buzzword Bingo Storage Edition 2019 (Wolfgang Stief)
 
Hochleistungsspeichersysteme für Datenanalyse an der TU Dresden (Michael Kluge)
Hochleistungsspeichersysteme für Datenanalyse an der TU Dresden (Michael Kluge)Hochleistungsspeichersysteme für Datenanalyse an der TU Dresden (Michael Kluge)
Hochleistungsspeichersysteme für Datenanalyse an der TU Dresden (Michael Kluge)
 
Intelligent Edge - breaking the storage hype (Michael Beeck, mibeeck GmbH)
Intelligent Edge - breaking the storage hype (Michael Beeck, mibeeck GmbH)Intelligent Edge - breaking the storage hype (Michael Beeck, mibeeck GmbH)
Intelligent Edge - breaking the storage hype (Michael Beeck, mibeeck GmbH)
 

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Top 10 Most Downloaded Games on Play Store in 2024
Top 10 Most Downloaded Games on Play Store in 2024Top 10 Most Downloaded Games on Play Store in 2024
Top 10 Most Downloaded Games on Play Store in 2024
 
GenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdfGenAI Risks & Security Meetup 01052024.pdf
GenAI Risks & Security Meetup 01052024.pdf
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
presentation ICT roal in 21st century education
presentation ICT roal in 21st century educationpresentation ICT roal in 21st century education
presentation ICT roal in 21st century education
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
 
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsTop 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
 
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
 

Operation Unthinkable – Software Defined Storage @ Booking.com (Peter Buschman)

  • 1. Peter Buschman Operation Unthinkable November 28th, 2019 Storage Forum Leipzig Software Defined Storage
  • 2. About ● Majority part of Booking Holdings (formerly Priceline Group) ● One of the largest e-Commerce websites in the world ● The largest online accommodation website in the world ● >1.5 Million properties in 220+ countries and territories ● 1.55 Million room nights booked every 24 hours ● >15,000 employees in 198 offices in 70 countries ● 1000s of LUNs, NFS shares, and S3 buckets ● Managed by a storage team of only 4 people (as of October, 2019)
  • 3. Defining the Unthinkable "Developing and supporting our own enterprise storage hardware and software."
  • 4. Attributes of Enterprise Storage ●Gigantic Chassis ●Specialized Power ●Custom Cooling ●Difficult to Service ●Massively Expensive "Habbakuks" "Kugelpanzers" ●Non-Standard Chassis ●Complicated Supply Chain ●Little Documentation ●Single Purpose ●Relatively Cheap LIMITED RE-USABILITY
  • 5. The complexity of storage...
  • 6. The complexity of storage...
  • 7. How to calculate complexity complexity = (u -1) x c / n u: number of rack-units c: number of cables n: number of racks
  • 9. Attributes of SDS Hardware ●Compact Chassis ●Standard Form Factor ●Standard Power ●Standard Cooling ●Easy to Service "Shermans and T34s" ●Extensible Design ●Off the Shelf Ordering ●Optimized Supply Chain ●Multi Purpose ●Cost Effective MAXIMUM RE-USABILITY
  • 10. Re-thinking the problem... ??? 2U Empty Space Empty 100Gb Switch Ports
  • 11. Building a BOMB ● 2U max height ● 90cm max depth ● ~500W power ● 100Gb networking ● Non-volatile memory ● NVMe and high-capacity disk ● Maximize Terabytes/Watt ● Broadest software eco-system possible ● Simple enough for remote hands to install ??? Thinking inside the box.... (Bill of Materials)
  • 12. The plan "crazy enough to work" Image credit: Dutch National Archives Image credit: U.S. Public Domain
  • 13. The Booking BOM Gen1 "SFF" Intel ● 2 x Intel Skylake 6146 CPUs ● 12 x 32GB DDR4 RAM (384GB total) ● 12 x 16GB NVDIMM-N (192GB total) ● 1 x 100Gb NIC ● 4 - 16 x 15.36TB (245.76TB total) ● 1 x HHHL FPGA storage accelerator (optional) ● Tool-less L-bracket rails ● Color-coded C13 to C14 power-cords ● 50cm and 1m network cables in the box ● Installs in < 5mins
  • 14. The Booking BOM Gen2 "SFF" AMD ● 1 x AMD EPYC Rome 7402P 32-core CPU ● 12 x 128GB DDR4 RAM (1.5TB total) ● 4 x 32GB NVDIMM-N (128GB total) ● 1 x 100Gb NIC ● 4 - 24 x 15.36TB U.2 NVMe (368.64TB max) ● 1 x HHHL FPGA storage accelerator (optional) ● Tool-less L-bracket rails ● Color-coded C13 to C14 power-cords ● 50cm and 1m network cables in the box ● Installs in < 5mins
  • 15. The Booking BOM Gen1 "LFF" ● 2 x Intel Skylake 6132 CPUs ● 12 x 32GB DDR4 RAM (384GB total) ● 12 x 16GB NVDIMM-N (192GB total) ● 1 x 100Gb NIC ● 12 x 14TB 7200rpm SATA disks (224TB total) ● 2 x 15.36TB HHHL NVMe (30.72 TB total) ● Tool-less L-bracket rails ● Color-coded C13 to C14 power-cords ● 50cm and 1m network cables in the box ● Installs in < 5mins
  • 16. Tool-less drive caddies 2.5" Small Form Factor (SFF)
  • 17. Tool-less drive caddies 3.5" Large Form Factor (LFF)
  • 18. Re-thinking the solution... 12 + 4 = 16 x 14TB = 224TB Disk Capacity 2 x 15.36 = 30.72TB NVMe SSD Capacity 12 x 14TB Disks 4 x 14TB Disks 1 x 100Gb NIC 2 x 15.36TB NVMe
  • 19. How complex is this? 1 2 3 4 4 cables x (2U - 1) / 1 rack = factor 4 complexity
  • 20. Lots of failures... ● Ruined €100k of 11TB NVMe SSDs ● Early all-NVMe chassis were very buggy ● Painful lessons from early CPU choices ● Failed miserably with RDMA / RoCE ● Lots of NIC, BIOS, and SSD firmware issues ● Lots of SW + HW support matrix issues ● Struggles with vendor communications ● Struggles with vendor business cycles ● ...but we kept iterating! Image credit: Imperial War Museum, UK © IWM (NA 13896)
  • 21. What we have achieved so far... ● Deployed ~100 2U storage nodes ● Eliminated dedicated storage racks ● Cut power draw by over 50% ● Gone "all in" on software defined storage ● Eliminated storage hardware maintenance ● Switched entirely to software subscriptions ● Increased utilization while reducing costs ● Put total storage spending on a downward trend ● ...despite continued high data growth!
  • 23. What is your Unthinkable? Image credit: Michael Coppins - Wikimedia Commons
  • 24. ● Plans you do not want to execute can yield great outcomes ● Define your own Unthinkable when it comes to SDS ● Build a BOM and don't compromise ● Recruit allies - you will need them! ● Expect failure and work through it ● Take your time Closing.thoughts