SlideShare a Scribd company logo
How NVMe over TCP runs
PostgreSQL in Quicksilver
mode!
Sagy Volkov
Distinguished Performance Architect
Lightbits Labs
Lightbits Labs Proprietary and Confidential | 2
● Intro to NVMe/TCP (or NVMe Over TCP)
● Why NVMe/TCP
● The PostgreSQL (art of) Partitioning
● Some Performance Comparison (and how I ran them)
● Q&A
Agenda
Lightbits Labs Proprietary and Confidential | 3
● NVM Express™ (NVMe) standard released in March 2011
○ Architecture, command set, and queueing interface for PCIe SSDs
■ Optimized for direct attached NVM PCIe® SSDs
■ The goal was a single interface that is scalable from client to enterprise
● NVMe™ over Fabrics (NVMe-oF™) standard released in June 2016
○ Extends the architecture, command set, and queueing interface of NVMe to
fabric attached storage
○ Two initial transport binding specifications
■ RDMA (Infiniband, RoCE, iWARP)
■ Fibre Channel
● TCP transport binding specification (NVMe™/TCP) specified for NVMe-oF in
November of 2018
NVMe/TCP - a history lesson…
Lightbits Labs Proprietary and Confidential | 4
● Also in 2018, Lightbits Labs introduced the first storage solution that was
designed and built for NVMe/TCP
● Added to the Linux kernel in 4.10
● Part of all major Linux enterprise distros.
● Today: Lightbits, Ceph, Dell/EMC, NetApp, Pure Storage and more.
NVMe/TCP - more lesson…
Lightbits Labs Proprietary and Confidential | 5
NVMe-oF & NVMe/TCP - Protocol Layers
Lightbits Labs Proprietary and Confidential | 6
NVMe Transport Taxonomy
Lightbits Labs Proprietary and Confidential | 7
NVMe/TCP Queue Mapping
Lightbits Labs Proprietary and Confidential | 8
Why NVMe/TCP
(or why do we need another NVMe-oF transport?)
▪ PCIe®
̶ Great for direct attached NVMe SSDs
̶ Does not scale well to large topologies
▪ RDMA (Infiniband, RoCE, iWARP) and Fibre Channel
̶ Provides a high degree of scalability, but requires special networks and hardware
▪ TCP
̶ Ubiquitous (does not require special networks or hardware)
̶ Scalable allowing large scale deployments and operation over long distances
̶ Can provide performance (throughput and latency) that is (almost) comparable to
direct attached NVMe SSDs
̶ Provides a reliable error-free in-order bidirectional byte stream service between
a sender and receiver
Lightbits Labs Proprietary and Confidential | 9
Things to consider:
● Remember that you can use NVMe/TCP, wherever you have TCP available. Simple
as that.
● Not all NVMe/TCP storage systems are the same
● Cloud: Instances usually have higher bandwidth for network than for native
storage from the cloud provider.
● Cloud: Network resources are billed differently depending on your cloud provider
(ingress/egress)
The PostgreSQL (art of) Partitioning
Lightbits Labs Proprietary and Confidential | 11
The Art of Partitioning
● Not really an art, just an overlooked feature :)
● Version 10 (actually 9.6)
● Why use it?
○ Improved query performance*
○ Easy to manage data
○ Improve data load and index creation
● Types:
○ Range (like dates)
○ List (like specific category)
○ Hash (evenly distributed)
Performance
Lightbits Labs Proprietary and Confidential | 13
How I measure performance
● Why PostgreSQL?
● Why Kubernetes?
● Why CNPG?
● Why Sherlock? and sysbench?
Lightbits Labs Proprietary and Confidential | 14
Some basic performance
Random I/O, replication factor 2
Workload Block Size IOPs BW
Read Only
4KB 14M 53GB/s
128KB 535K 65GB/s
70% Read / 30% Write
4KB 5.3M 20GB/s
128KB 325K 40GB/s
50% Read / 50% Write
4KB 3.8M 14.6GB/s
128KB 205K 25GB/s
Write Only
4KB 2.3M 8.9GB/s
128KB 107K 13GB/s
4KB Random I/O, replication factor 2
Workload IOPS
R/W
latency
Avg
(usec)
99th
(usec)
100% Read 3.8 Million read 158.76 292.86
70% Read
2.4 Million
read 225.96 497.66
30% Write write 338.41 651.26
50% Read
2 Million
read 254.40 569.34
50% Write write 398.10 765.95
100% Write 1.3 Million write 493.30 946.18
Lightbits Labs Proprietary and Confidential | 15
Availability Zone
VPC
AWS Cloud
I4i.8xlarge
PGbench Performance testing setup in AWS
18.75
Gbps
18.75
Gbps
18.75
Gbps
I4i.8xlarge
I4i.8xlarge
PVCs (Lightbits volumes) >
Openshift: 4.13
5 x Worker nodes: r5n.8x (32 cores,
256GB mem)
CNPG (5 clusters):
● CPU: 18
● Mem: 10Gi
● 1 x 100GB PVC PGDATA
● 1 x 100GB PVC WAL
● 8 x 100GB PVCs = 8 TBs = 8
Partitions for accounts and
history tables
Lightbits: version 3.7.1
● 3 x I4i.8xlarge
● Only 2 NVMe devices per
instance
● Instance B/W limited to
18.75Gbps
3 x Replication at the
Storage level using
Lightbits (SC)
Lightbits Labs Proprietary and Confidential | 16
Performance tests using pgbench
These numbers are
using the smallest
Lightbits cluster that is
available in the cloud
with the least number
of NVMe devices and
limited network
bandwidth.
Imagine how many
CNPG clusters we can
run using i4i.metal (for
example) or on-prem
servers!!!
Lightbits Labs Proprietary and Confidential | 17
Performance tests using pgbench - on-prem
● 12 partitions
● Same number of
cores for the
workers, older
CPU generation.
● More SSD Drives
Lightbits Labs Proprietary and Confidential | 18
Lightbits performance differences compared to CEPH RBD
storage
Lightbits Delivers More Density - More Performance - Lower Latencies - Less Hardware
Q&A and Thanks for
having me!
Yael
Ad

More Related Content

Recently uploaded (20)

Setup & Implementation of OutSystems Cloud Connector ODC
Setup & Implementation of OutSystems Cloud Connector ODCSetup & Implementation of OutSystems Cloud Connector ODC
Setup & Implementation of OutSystems Cloud Connector ODC
outsystemspuneusergr
 
Sermon_How To Change Your Life For Better_2.0.pptx
Sermon_How To Change Your Life For Better_2.0.pptxSermon_How To Change Your Life For Better_2.0.pptx
Sermon_How To Change Your Life For Better_2.0.pptx
oxofoegbu
 
Lec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptx
Lec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptxLec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptx
Lec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptx
TayyabaSiddiqui12
 
A One Day Workshop on English Language Learning
A One Day Workshop on English Language LearningA One Day Workshop on English Language Learning
A One Day Workshop on English Language Learning
Disha Kariya
 
Microsoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdf
Microsoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdfMicrosoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdf
Microsoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdf
MinniePfeiffer
 
APEC - POWER POINT.pptxtrabajo de investig
APEC - POWER POINT.pptxtrabajo de investigAPEC - POWER POINT.pptxtrabajo de investig
APEC - POWER POINT.pptxtrabajo de investig
MyriamRaquelCoronado
 
THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...
THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...
THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...
ASHISHKUMAR504404
 
Yellow and Black Modern Minimalist Workshop Presentation.pptx
Yellow and Black Modern Minimalist Workshop Presentation.pptxYellow and Black Modern Minimalist Workshop Presentation.pptx
Yellow and Black Modern Minimalist Workshop Presentation.pptx
alibabattying
 
Bidding World Conference 2027 - Ghana.pptx
Bidding World Conference 2027 - Ghana.pptxBidding World Conference 2027 - Ghana.pptx
Bidding World Conference 2027 - Ghana.pptx
ISGF - International Scout and Guide Fellowship
 
Speech 3-A Vision for Tomorrow for GE2025
Speech 3-A Vision for Tomorrow for GE2025Speech 3-A Vision for Tomorrow for GE2025
Speech 3-A Vision for Tomorrow for GE2025
Noraini Yunus
 
Key Elements of a Procurement Plan.docx.
Key Elements of a Procurement Plan.docx.Key Elements of a Procurement Plan.docx.
Key Elements of a Procurement Plan.docx.
NeoRakodu
 
Operation market garden presentation.pptx
Operation market garden presentation.pptxOperation market garden presentation.pptx
Operation market garden presentation.pptx
antoniopoljica
 
NASIG ISSN 2025 updated for the_4-30meeting.pptx
NASIG ISSN 2025 updated for the_4-30meeting.pptxNASIG ISSN 2025 updated for the_4-30meeting.pptx
NASIG ISSN 2025 updated for the_4-30meeting.pptx
reine1
 
fundamentals of communicationclass notes.pptx
fundamentals of communicationclass notes.pptxfundamentals of communicationclass notes.pptx
fundamentals of communicationclass notes.pptx
Sunkod
 
CAE. final pptx.pptx h
CAE. final pptx.pptx                        hCAE. final pptx.pptx                        h
CAE. final pptx.pptx h
Test761
 
The Business Dynamics of Quick Commerce.pdf
The Business Dynamics of Quick Commerce.pdfThe Business Dynamics of Quick Commerce.pdf
The Business Dynamics of Quick Commerce.pdf
RDinuRao
 
Besu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptx
Besu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptxBesu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptx
Besu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptx
Rajdeep Chakraborty
 
2025-04-27 A New Day Dawns 02 (shared slides).pptx
2025-04-27 A New Day Dawns 02 (shared slides).pptx2025-04-27 A New Day Dawns 02 (shared slides).pptx
2025-04-27 A New Day Dawns 02 (shared slides).pptx
Dale Wells
 
ICONX - Presentation - PDF - english - international
ICONX - Presentation - PDF - english - internationalICONX - Presentation - PDF - english - international
ICONX - Presentation - PDF - english - international
Bitcoin Mining RACE
 
Hartcliffe Betrayed Bristol Radical History Festival April 2025.pptx
Hartcliffe Betrayed Bristol Radical History Festival April 2025.pptxHartcliffe Betrayed Bristol Radical History Festival April 2025.pptx
Hartcliffe Betrayed Bristol Radical History Festival April 2025.pptx
Paul Smith
 
Setup & Implementation of OutSystems Cloud Connector ODC
Setup & Implementation of OutSystems Cloud Connector ODCSetup & Implementation of OutSystems Cloud Connector ODC
Setup & Implementation of OutSystems Cloud Connector ODC
outsystemspuneusergr
 
Sermon_How To Change Your Life For Better_2.0.pptx
Sermon_How To Change Your Life For Better_2.0.pptxSermon_How To Change Your Life For Better_2.0.pptx
Sermon_How To Change Your Life For Better_2.0.pptx
oxofoegbu
 
Lec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptx
Lec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptxLec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptx
Lec 3 - Chapter 2 Carl Jung’s Theory of Personality.pptx
TayyabaSiddiqui12
 
A One Day Workshop on English Language Learning
A One Day Workshop on English Language LearningA One Day Workshop on English Language Learning
A One Day Workshop on English Language Learning
Disha Kariya
 
Microsoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdf
Microsoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdfMicrosoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdf
Microsoft Azure Data Fundamentals (DP-900) Exam Dumps & Questions 2025.pdf
MinniePfeiffer
 
APEC - POWER POINT.pptxtrabajo de investig
APEC - POWER POINT.pptxtrabajo de investigAPEC - POWER POINT.pptxtrabajo de investig
APEC - POWER POINT.pptxtrabajo de investig
MyriamRaquelCoronado
 
THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...
THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...
THE SEXUAL HARASSMENT OF WOMAN AT WORKPLACE (PREVENTION, PROHIBITION & REDRES...
ASHISHKUMAR504404
 
Yellow and Black Modern Minimalist Workshop Presentation.pptx
Yellow and Black Modern Minimalist Workshop Presentation.pptxYellow and Black Modern Minimalist Workshop Presentation.pptx
Yellow and Black Modern Minimalist Workshop Presentation.pptx
alibabattying
 
Speech 3-A Vision for Tomorrow for GE2025
Speech 3-A Vision for Tomorrow for GE2025Speech 3-A Vision for Tomorrow for GE2025
Speech 3-A Vision for Tomorrow for GE2025
Noraini Yunus
 
Key Elements of a Procurement Plan.docx.
Key Elements of a Procurement Plan.docx.Key Elements of a Procurement Plan.docx.
Key Elements of a Procurement Plan.docx.
NeoRakodu
 
Operation market garden presentation.pptx
Operation market garden presentation.pptxOperation market garden presentation.pptx
Operation market garden presentation.pptx
antoniopoljica
 
NASIG ISSN 2025 updated for the_4-30meeting.pptx
NASIG ISSN 2025 updated for the_4-30meeting.pptxNASIG ISSN 2025 updated for the_4-30meeting.pptx
NASIG ISSN 2025 updated for the_4-30meeting.pptx
reine1
 
fundamentals of communicationclass notes.pptx
fundamentals of communicationclass notes.pptxfundamentals of communicationclass notes.pptx
fundamentals of communicationclass notes.pptx
Sunkod
 
CAE. final pptx.pptx h
CAE. final pptx.pptx                        hCAE. final pptx.pptx                        h
CAE. final pptx.pptx h
Test761
 
The Business Dynamics of Quick Commerce.pdf
The Business Dynamics of Quick Commerce.pdfThe Business Dynamics of Quick Commerce.pdf
The Business Dynamics of Quick Commerce.pdf
RDinuRao
 
Besu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptx
Besu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptxBesu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptx
Besu Shibpur Enquesta 2012 Intra College General Quiz Prelims.pptx
Rajdeep Chakraborty
 
2025-04-27 A New Day Dawns 02 (shared slides).pptx
2025-04-27 A New Day Dawns 02 (shared slides).pptx2025-04-27 A New Day Dawns 02 (shared slides).pptx
2025-04-27 A New Day Dawns 02 (shared slides).pptx
Dale Wells
 
ICONX - Presentation - PDF - english - international
ICONX - Presentation - PDF - english - internationalICONX - Presentation - PDF - english - international
ICONX - Presentation - PDF - english - international
Bitcoin Mining RACE
 
Hartcliffe Betrayed Bristol Radical History Festival April 2025.pptx
Hartcliffe Betrayed Bristol Radical History Festival April 2025.pptxHartcliffe Betrayed Bristol Radical History Festival April 2025.pptx
Hartcliffe Betrayed Bristol Radical History Festival April 2025.pptx
Paul Smith
 

Featured (20)

AI Trends in Creative Operations 2024 by Artwork Flow.pdf
AI Trends in Creative Operations 2024 by Artwork Flow.pdfAI Trends in Creative Operations 2024 by Artwork Flow.pdf
AI Trends in Creative Operations 2024 by Artwork Flow.pdf
marketingartwork
 
Skeleton Culture Code
Skeleton Culture CodeSkeleton Culture Code
Skeleton Culture Code
Skeleton Technologies
 
PEPSICO Presentation to CAGNY Conference Feb 2024
PEPSICO Presentation to CAGNY Conference Feb 2024PEPSICO Presentation to CAGNY Conference Feb 2024
PEPSICO Presentation to CAGNY Conference Feb 2024
Neil Kimberley
 
Content Methodology: A Best Practices Report (Webinar)
Content Methodology: A Best Practices Report (Webinar)Content Methodology: A Best Practices Report (Webinar)
Content Methodology: A Best Practices Report (Webinar)
contently
 
How to Prepare For a Successful Job Search for 2024
How to Prepare For a Successful Job Search for 2024How to Prepare For a Successful Job Search for 2024
How to Prepare For a Successful Job Search for 2024
Albert Qian
 
Social Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie InsightsSocial Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie Insights
Kurio // The Social Media Age(ncy)
 
Trends In Paid Search: Navigating The Digital Landscape In 2024
Trends In Paid Search: Navigating The Digital Landscape In 2024Trends In Paid Search: Navigating The Digital Landscape In 2024
Trends In Paid Search: Navigating The Digital Landscape In 2024
Search Engine Journal
 
5 Public speaking tips from TED - Visualized summary
5 Public speaking tips from TED - Visualized summary5 Public speaking tips from TED - Visualized summary
5 Public speaking tips from TED - Visualized summary
SpeakerHub
 
ChatGPT and the Future of Work - Clark Boyd
ChatGPT and the Future of Work - Clark Boyd ChatGPT and the Future of Work - Clark Boyd
ChatGPT and the Future of Work - Clark Boyd
Clark Boyd
 
Getting into the tech field. what next
Getting into the tech field. what next Getting into the tech field. what next
Getting into the tech field. what next
Tessa Mero
 
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Google's Just Not That Into You: Understanding Core Updates & Search IntentGoogle's Just Not That Into You: Understanding Core Updates & Search Intent
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Lily Ray
 
How to have difficult conversations
How to have difficult conversations How to have difficult conversations
How to have difficult conversations
Rajiv Jayarajah, MAppComm, ACC
 
Introduction to Data Science
Introduction to Data ScienceIntroduction to Data Science
Introduction to Data Science
Christy Abraham Joy
 
Time Management & Productivity - Best Practices
Time Management & Productivity -  Best PracticesTime Management & Productivity -  Best Practices
Time Management & Productivity - Best Practices
Vit Horky
 
The six step guide to practical project management
The six step guide to practical project managementThe six step guide to practical project management
The six step guide to practical project management
MindGenius
 
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
RachelPearson36
 
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Applitools
 
12 Ways to Increase Your Influence at Work
12 Ways to Increase Your Influence at Work12 Ways to Increase Your Influence at Work
12 Ways to Increase Your Influence at Work
GetSmarter
 
ChatGPT webinar slides
ChatGPT webinar slidesChatGPT webinar slides
ChatGPT webinar slides
Alireza Esmikhani
 
More than Just Lines on a Map: Best Practices for U.S Bike Routes
More than Just Lines on a Map: Best Practices for U.S Bike RoutesMore than Just Lines on a Map: Best Practices for U.S Bike Routes
More than Just Lines on a Map: Best Practices for U.S Bike Routes
Project for Public Spaces & National Center for Biking and Walking
 
AI Trends in Creative Operations 2024 by Artwork Flow.pdf
AI Trends in Creative Operations 2024 by Artwork Flow.pdfAI Trends in Creative Operations 2024 by Artwork Flow.pdf
AI Trends in Creative Operations 2024 by Artwork Flow.pdf
marketingartwork
 
PEPSICO Presentation to CAGNY Conference Feb 2024
PEPSICO Presentation to CAGNY Conference Feb 2024PEPSICO Presentation to CAGNY Conference Feb 2024
PEPSICO Presentation to CAGNY Conference Feb 2024
Neil Kimberley
 
Content Methodology: A Best Practices Report (Webinar)
Content Methodology: A Best Practices Report (Webinar)Content Methodology: A Best Practices Report (Webinar)
Content Methodology: A Best Practices Report (Webinar)
contently
 
How to Prepare For a Successful Job Search for 2024
How to Prepare For a Successful Job Search for 2024How to Prepare For a Successful Job Search for 2024
How to Prepare For a Successful Job Search for 2024
Albert Qian
 
Social Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie InsightsSocial Media Marketing Trends 2024 // The Global Indie Insights
Social Media Marketing Trends 2024 // The Global Indie Insights
Kurio // The Social Media Age(ncy)
 
Trends In Paid Search: Navigating The Digital Landscape In 2024
Trends In Paid Search: Navigating The Digital Landscape In 2024Trends In Paid Search: Navigating The Digital Landscape In 2024
Trends In Paid Search: Navigating The Digital Landscape In 2024
Search Engine Journal
 
5 Public speaking tips from TED - Visualized summary
5 Public speaking tips from TED - Visualized summary5 Public speaking tips from TED - Visualized summary
5 Public speaking tips from TED - Visualized summary
SpeakerHub
 
ChatGPT and the Future of Work - Clark Boyd
ChatGPT and the Future of Work - Clark Boyd ChatGPT and the Future of Work - Clark Boyd
ChatGPT and the Future of Work - Clark Boyd
Clark Boyd
 
Getting into the tech field. what next
Getting into the tech field. what next Getting into the tech field. what next
Getting into the tech field. what next
Tessa Mero
 
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Google's Just Not That Into You: Understanding Core Updates & Search IntentGoogle's Just Not That Into You: Understanding Core Updates & Search Intent
Google's Just Not That Into You: Understanding Core Updates & Search Intent
Lily Ray
 
Time Management & Productivity - Best Practices
Time Management & Productivity -  Best PracticesTime Management & Productivity -  Best Practices
Time Management & Productivity - Best Practices
Vit Horky
 
The six step guide to practical project management
The six step guide to practical project managementThe six step guide to practical project management
The six step guide to practical project management
MindGenius
 
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
RachelPearson36
 
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Applitools
 
12 Ways to Increase Your Influence at Work
12 Ways to Increase Your Influence at Work12 Ways to Increase Your Influence at Work
12 Ways to Increase Your Influence at Work
GetSmarter
 
Ad

stackconf 2025 | How NVMe over TCP runs PostgreSQL in Quicksilver mode! by Sagy Volkov.pdf

  • 1. How NVMe over TCP runs PostgreSQL in Quicksilver mode! Sagy Volkov Distinguished Performance Architect Lightbits Labs
  • 2. Lightbits Labs Proprietary and Confidential | 2 ● Intro to NVMe/TCP (or NVMe Over TCP) ● Why NVMe/TCP ● The PostgreSQL (art of) Partitioning ● Some Performance Comparison (and how I ran them) ● Q&A Agenda
  • 3. Lightbits Labs Proprietary and Confidential | 3 ● NVM Express™ (NVMe) standard released in March 2011 ○ Architecture, command set, and queueing interface for PCIe SSDs ■ Optimized for direct attached NVM PCIe® SSDs ■ The goal was a single interface that is scalable from client to enterprise ● NVMe™ over Fabrics (NVMe-oF™) standard released in June 2016 ○ Extends the architecture, command set, and queueing interface of NVMe to fabric attached storage ○ Two initial transport binding specifications ■ RDMA (Infiniband, RoCE, iWARP) ■ Fibre Channel ● TCP transport binding specification (NVMe™/TCP) specified for NVMe-oF in November of 2018 NVMe/TCP - a history lesson…
  • 4. Lightbits Labs Proprietary and Confidential | 4 ● Also in 2018, Lightbits Labs introduced the first storage solution that was designed and built for NVMe/TCP ● Added to the Linux kernel in 4.10 ● Part of all major Linux enterprise distros. ● Today: Lightbits, Ceph, Dell/EMC, NetApp, Pure Storage and more. NVMe/TCP - more lesson…
  • 5. Lightbits Labs Proprietary and Confidential | 5 NVMe-oF & NVMe/TCP - Protocol Layers
  • 6. Lightbits Labs Proprietary and Confidential | 6 NVMe Transport Taxonomy
  • 7. Lightbits Labs Proprietary and Confidential | 7 NVMe/TCP Queue Mapping
  • 8. Lightbits Labs Proprietary and Confidential | 8 Why NVMe/TCP (or why do we need another NVMe-oF transport?) ▪ PCIe® ̶ Great for direct attached NVMe SSDs ̶ Does not scale well to large topologies ▪ RDMA (Infiniband, RoCE, iWARP) and Fibre Channel ̶ Provides a high degree of scalability, but requires special networks and hardware ▪ TCP ̶ Ubiquitous (does not require special networks or hardware) ̶ Scalable allowing large scale deployments and operation over long distances ̶ Can provide performance (throughput and latency) that is (almost) comparable to direct attached NVMe SSDs ̶ Provides a reliable error-free in-order bidirectional byte stream service between a sender and receiver
  • 9. Lightbits Labs Proprietary and Confidential | 9 Things to consider: ● Remember that you can use NVMe/TCP, wherever you have TCP available. Simple as that. ● Not all NVMe/TCP storage systems are the same ● Cloud: Instances usually have higher bandwidth for network than for native storage from the cloud provider. ● Cloud: Network resources are billed differently depending on your cloud provider (ingress/egress)
  • 10. The PostgreSQL (art of) Partitioning
  • 11. Lightbits Labs Proprietary and Confidential | 11 The Art of Partitioning ● Not really an art, just an overlooked feature :) ● Version 10 (actually 9.6) ● Why use it? ○ Improved query performance* ○ Easy to manage data ○ Improve data load and index creation ● Types: ○ Range (like dates) ○ List (like specific category) ○ Hash (evenly distributed)
  • 13. Lightbits Labs Proprietary and Confidential | 13 How I measure performance ● Why PostgreSQL? ● Why Kubernetes? ● Why CNPG? ● Why Sherlock? and sysbench?
  • 14. Lightbits Labs Proprietary and Confidential | 14 Some basic performance Random I/O, replication factor 2 Workload Block Size IOPs BW Read Only 4KB 14M 53GB/s 128KB 535K 65GB/s 70% Read / 30% Write 4KB 5.3M 20GB/s 128KB 325K 40GB/s 50% Read / 50% Write 4KB 3.8M 14.6GB/s 128KB 205K 25GB/s Write Only 4KB 2.3M 8.9GB/s 128KB 107K 13GB/s 4KB Random I/O, replication factor 2 Workload IOPS R/W latency Avg (usec) 99th (usec) 100% Read 3.8 Million read 158.76 292.86 70% Read 2.4 Million read 225.96 497.66 30% Write write 338.41 651.26 50% Read 2 Million read 254.40 569.34 50% Write write 398.10 765.95 100% Write 1.3 Million write 493.30 946.18
  • 15. Lightbits Labs Proprietary and Confidential | 15 Availability Zone VPC AWS Cloud I4i.8xlarge PGbench Performance testing setup in AWS 18.75 Gbps 18.75 Gbps 18.75 Gbps I4i.8xlarge I4i.8xlarge PVCs (Lightbits volumes) > Openshift: 4.13 5 x Worker nodes: r5n.8x (32 cores, 256GB mem) CNPG (5 clusters): ● CPU: 18 ● Mem: 10Gi ● 1 x 100GB PVC PGDATA ● 1 x 100GB PVC WAL ● 8 x 100GB PVCs = 8 TBs = 8 Partitions for accounts and history tables Lightbits: version 3.7.1 ● 3 x I4i.8xlarge ● Only 2 NVMe devices per instance ● Instance B/W limited to 18.75Gbps 3 x Replication at the Storage level using Lightbits (SC)
  • 16. Lightbits Labs Proprietary and Confidential | 16 Performance tests using pgbench These numbers are using the smallest Lightbits cluster that is available in the cloud with the least number of NVMe devices and limited network bandwidth. Imagine how many CNPG clusters we can run using i4i.metal (for example) or on-prem servers!!!
  • 17. Lightbits Labs Proprietary and Confidential | 17 Performance tests using pgbench - on-prem ● 12 partitions ● Same number of cores for the workers, older CPU generation. ● More SSD Drives
  • 18. Lightbits Labs Proprietary and Confidential | 18 Lightbits performance differences compared to CEPH RBD storage Lightbits Delivers More Density - More Performance - Lower Latencies - Less Hardware
  • 19. Q&A and Thanks for having me! Yael