SlideShare a Scribd company logo
Keeping your latency SLA
No matter what!
Glauber Costa
VP Field Engineering ScyllaDB
Presenter bio
Glauber Costa is VP of Field Engineering at ScyllaDB. He shares
his time between the engineering department working on
upcoming Scylla features and helping customers succeed.
Before ScyllaDB, Glauber worked with Virtualization in the Linux
Kernel for 10 years, with contributions into the Xen and KVM
Hypervisors and all sorts of guest functionality and containers.
Where do latencies come from?
Where do latencies come from?
Storage I/O
▪ 4kB reads
▪ 150+ concurrent
requests
▪ 300𝜇s avg latency
Storage I/O
▪ 128kB reads
▪ ~30 concurrent
requests
▪ 2ms avg latency
Storage I/O
▪ AWS i3.metal write bandwidth: 6GB/s
Storage I/O
▪ AWS i3.metal write bandwidth: 6GB/s
▪ AWS i3.metal read bandwidth: 15GB/s
Storage I/O
128kB write128kB write128kB write
12ms worth of latency
Storage
Storage I/O
512-byte
read
128kB write128kB write128kB write
12ms worth of latency
Storage
at least 12ms!
Scylla 2.2 vs Scylla 2.3
Scylla 2.2
--max-io-requests=150
Scylla 2.3
read_iops: 9001
read_bandwidth: 160MB
write_iops: 9723
write_bandwidth: 162MB
Scylla 2.2 vs Scylla 2.3 (AWS EBS, reads vs compaction)
Scylla 2.2 Scylla 2.3
average 8.4ms 6.0ms 40%
p50 5.6ms 3.1ms 80%
p95 15ms 8ms 85%
p99 67ms 63ms 6.3%
CPU Scheduling
read write read Compaction
CPU
SSTable write
CPU Scheduling
read write read Compaction
CPU
CPU
Compaction
SSTable write
SSTable write
read write readread write read
CPU Scheduling
▪ Preemption every x 𝜇s - the task quota
▪ Expected latency increase when n more classes run:
• n * task_quota
• Scylla 2.3 task quota: 500 𝜇s
Which tasks to run?
100 shares
100 shares
Which tasks to run?
100 shares
50 shares
Which tasks to run?
Workload changes, Scylla adapts
Scylla 2.1, no controller
Scylla 2.2, compaction controller
Comparative results
Scylla 2.1 Scylla 2.2
Writes/s 354K 418K +18%
CPU Scheduling
▪ Preemption every x 𝜇s - the task quota
▪ Expected latency increase when n more classes run:
• n * task_quota
• Scylla 2.3 task quota: 500 𝜇s
CPU Scheduling
▪ Preemption every x 𝜇s - the task quota
▪ Expected latency increase when n more classes run:
• n * task_quota
• Scylla 2.3 task quota: 500 𝜇s
▪ Failure to preempt: task quota violation
Task quota violations
▪ Large memory allocations
▪ Memory allocator itself
▪ The Cache
▪ I/O Subsystem
▪ SSTable writing
▪ Bloom Filter generation
▪ The Linux Kernel
Scylla 2.2 vs Scylla 2.3 (99.9th percentile write latency)
Scylla 2.2 - 4ms to 6ms Scylla 2.3 - < 4ms
It doesn’t end there
We need to preempt more
▪ Preemption uses a thread that sets a flag
• Source of context switches
▪ Every preemption does I/O
• It is a good thing, but I/O goes to the kernel
▪ New polling API merged to the Linux kernel
• No context switches for I/O! Can poll a lot more often
We need to preempt more - Linux 4.19
Completion queue shared
between kernel and user
Disk Write Completion
Disk Read Completion
Socket is readable
Disk Read Completion
Socket is readable
Database dequeues event notifications
from shared ring without a system call
Kernel appends event notifications
from interrupts
SCYLLA
KERNEL
Shard-aware driver - Scylla 2.3
There is no network hop, both
CPUs are in the same system.
But worst case, it adds a task
quota to request processing.
CPU that owns
connection
CPU that owns
the token
Token
Shard-aware driver - Scylla 2.3
Driver opens connections to all
CPUs, and learns which tokens
belong to each CPU.
Request is sent directly to
owner.
CPU that owns
the token
Token
Summary
▪ Keeping latencies bounded and predictable is important
▪ Latency-inducing events really come from
everywhere - from the Linux kernel itself to
the most surprising places in Scylla
▪ We keep investing time year after year to make
sure our latencies are even more predictable
Thank You
Any Questions ?
Please stay in touch
glauber@scylladb.com
@glcst

More Related Content

What's hot

Scylla Summit 2016: Compose on Containing the Database
Scylla Summit 2016: Compose on Containing the DatabaseScylla Summit 2016: Compose on Containing the Database
Scylla Summit 2016: Compose on Containing the Database
ScyllaDB
 
Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...
Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...
Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...
ScyllaDB
 
How We Made Scylla Maintenance Easier, Safer and Faster
How We Made Scylla Maintenance Easier, Safer and FasterHow We Made Scylla Maintenance Easier, Safer and Faster
How We Made Scylla Maintenance Easier, Safer and Faster
ScyllaDB
 
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
ScyllaDB
 
Back to the future with C++ and Seastar
Back to the future with C++ and SeastarBack to the future with C++ and Seastar
Back to the future with C++ and Seastar
Tzach Livyatan
 
Seastar Summit 2019 Keynote
Seastar Summit 2019 KeynoteSeastar Summit 2019 Keynote
Seastar Summit 2019 Keynote
ScyllaDB
 
How to Monitor and Size Workloads on AWS i3 instances
How to Monitor and Size Workloads on AWS i3 instancesHow to Monitor and Size Workloads on AWS i3 instances
How to Monitor and Size Workloads on AWS i3 instances
ScyllaDB
 
Scylla Summit 2016: Scylla at Samsung SDS
Scylla Summit 2016: Scylla at Samsung SDSScylla Summit 2016: Scylla at Samsung SDS
Scylla Summit 2016: Scylla at Samsung SDS
ScyllaDB
 
Scylla Summit 2018: How We Made Large Partition Scans Over Two Times Faster
Scylla Summit 2018: How We Made Large Partition Scans Over Two Times FasterScylla Summit 2018: How We Made Large Partition Scans Over Two Times Faster
Scylla Summit 2018: How We Made Large Partition Scans Over Two Times Faster
ScyllaDB
 
Scylla’s Journey Towards Being an Elastic Cloud Native Database
Scylla’s Journey Towards Being an Elastic Cloud Native DatabaseScylla’s Journey Towards Being an Elastic Cloud Native Database
Scylla’s Journey Towards Being an Elastic Cloud Native Database
ScyllaDB
 
How we got to 1 millisecond latency in 99% under repair, compaction, and flus...
How we got to 1 millisecond latency in 99% under repair, compaction, and flus...How we got to 1 millisecond latency in 99% under repair, compaction, and flus...
How we got to 1 millisecond latency in 99% under repair, compaction, and flus...
ScyllaDB
 
How Scylla Manager Handles Backups
How Scylla Manager Handles BackupsHow Scylla Manager Handles Backups
How Scylla Manager Handles Backups
ScyllaDB
 
Lightweight Transactions at Lightning Speed
Lightweight Transactions at Lightning SpeedLightweight Transactions at Lightning Speed
Lightweight Transactions at Lightning Speed
ScyllaDB
 
mParticle's Journey to Scylla from Cassandra
mParticle's Journey to Scylla from CassandramParticle's Journey to Scylla from Cassandra
mParticle's Journey to Scylla from Cassandra
ScyllaDB
 
Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...
Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...
Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...
ScyllaDB
 
Comparing Apache Cassandra 4.0, 3.0, and ScyllaDB
Comparing Apache Cassandra 4.0, 3.0, and ScyllaDBComparing Apache Cassandra 4.0, 3.0, and ScyllaDB
Comparing Apache Cassandra 4.0, 3.0, and ScyllaDB
ScyllaDB
 
Scylla: 1 Million CQL operations per second per server
Scylla: 1 Million CQL operations per second per serverScylla: 1 Million CQL operations per second per server
Scylla: 1 Million CQL operations per second per server
Avi Kivity
 
Scylla Summit 2018: What's New in Scylla Manager?
Scylla Summit 2018: What's New in Scylla Manager?Scylla Summit 2018: What's New in Scylla Manager?
Scylla Summit 2018: What's New in Scylla Manager?
ScyllaDB
 
10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...
10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...
10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...
DevOpsDays Tel Aviv
 
Scylla Summit 2018: OLAP or OLTP? Why Not Both?
Scylla Summit 2018: OLAP or OLTP? Why Not Both?Scylla Summit 2018: OLAP or OLTP? Why Not Both?
Scylla Summit 2018: OLAP or OLTP? Why Not Both?
ScyllaDB
 

What's hot (20)

Scylla Summit 2016: Compose on Containing the Database
Scylla Summit 2016: Compose on Containing the DatabaseScylla Summit 2016: Compose on Containing the Database
Scylla Summit 2016: Compose on Containing the Database
 
Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...
Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...
Scylla Summit 2018: Make Scylla Fast Again! Find out how using Tools, Talent,...
 
How We Made Scylla Maintenance Easier, Safer and Faster
How We Made Scylla Maintenance Easier, Safer and FasterHow We Made Scylla Maintenance Easier, Safer and Faster
How We Made Scylla Maintenance Easier, Safer and Faster
 
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
Scylla Summit 2016: Outbrain Case Study - Lowering Latency While Doing 20X IO...
 
Back to the future with C++ and Seastar
Back to the future with C++ and SeastarBack to the future with C++ and Seastar
Back to the future with C++ and Seastar
 
Seastar Summit 2019 Keynote
Seastar Summit 2019 KeynoteSeastar Summit 2019 Keynote
Seastar Summit 2019 Keynote
 
How to Monitor and Size Workloads on AWS i3 instances
How to Monitor and Size Workloads on AWS i3 instancesHow to Monitor and Size Workloads on AWS i3 instances
How to Monitor and Size Workloads on AWS i3 instances
 
Scylla Summit 2016: Scylla at Samsung SDS
Scylla Summit 2016: Scylla at Samsung SDSScylla Summit 2016: Scylla at Samsung SDS
Scylla Summit 2016: Scylla at Samsung SDS
 
Scylla Summit 2018: How We Made Large Partition Scans Over Two Times Faster
Scylla Summit 2018: How We Made Large Partition Scans Over Two Times FasterScylla Summit 2018: How We Made Large Partition Scans Over Two Times Faster
Scylla Summit 2018: How We Made Large Partition Scans Over Two Times Faster
 
Scylla’s Journey Towards Being an Elastic Cloud Native Database
Scylla’s Journey Towards Being an Elastic Cloud Native DatabaseScylla’s Journey Towards Being an Elastic Cloud Native Database
Scylla’s Journey Towards Being an Elastic Cloud Native Database
 
How we got to 1 millisecond latency in 99% under repair, compaction, and flus...
How we got to 1 millisecond latency in 99% under repair, compaction, and flus...How we got to 1 millisecond latency in 99% under repair, compaction, and flus...
How we got to 1 millisecond latency in 99% under repair, compaction, and flus...
 
How Scylla Manager Handles Backups
How Scylla Manager Handles BackupsHow Scylla Manager Handles Backups
How Scylla Manager Handles Backups
 
Lightweight Transactions at Lightning Speed
Lightweight Transactions at Lightning SpeedLightweight Transactions at Lightning Speed
Lightweight Transactions at Lightning Speed
 
mParticle's Journey to Scylla from Cassandra
mParticle's Journey to Scylla from CassandramParticle's Journey to Scylla from Cassandra
mParticle's Journey to Scylla from Cassandra
 
Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...
Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...
Scylla Summit 2022: Operating at Monstrous Scales: Benchmarking Petabyte Work...
 
Comparing Apache Cassandra 4.0, 3.0, and ScyllaDB
Comparing Apache Cassandra 4.0, 3.0, and ScyllaDBComparing Apache Cassandra 4.0, 3.0, and ScyllaDB
Comparing Apache Cassandra 4.0, 3.0, and ScyllaDB
 
Scylla: 1 Million CQL operations per second per server
Scylla: 1 Million CQL operations per second per serverScylla: 1 Million CQL operations per second per server
Scylla: 1 Million CQL operations per second per server
 
Scylla Summit 2018: What's New in Scylla Manager?
Scylla Summit 2018: What's New in Scylla Manager?Scylla Summit 2018: What's New in Scylla Manager?
Scylla Summit 2018: What's New in Scylla Manager?
 
10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...
10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...
10 Devops-Friendly Database Must-Haves - Dor Laor, ScyllaDB - DevOpsDays Tel ...
 
Scylla Summit 2018: OLAP or OLTP? Why Not Both?
Scylla Summit 2018: OLAP or OLTP? Why Not Both?Scylla Summit 2018: OLAP or OLTP? Why Not Both?
Scylla Summit 2018: OLAP or OLTP? Why Not Both?
 

Similar to Scylla Summit 2018: Keeping Your Latency SLAs No Matter What!

How is Kafka so Fast?
How is Kafka so Fast?How is Kafka so Fast?
How is Kafka so Fast?
Ricardo Paiva
 
Scality S3 Server: Node js Meetup Presentation
Scality S3 Server: Node js Meetup PresentationScality S3 Server: Node js Meetup Presentation
Scality S3 Server: Node js Meetup Presentation
Scality
 
AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...
AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...
AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...
ScyllaDB
 
Under The Hood Of A Shard-Per-Core Database Architecture
Under The Hood Of A Shard-Per-Core Database ArchitectureUnder The Hood Of A Shard-Per-Core Database Architecture
Under The Hood Of A Shard-Per-Core Database Architecture
ScyllaDB
 
Storage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, WhiptailStorage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, Whiptail
Internet World
 
Memory, Big Data, NoSQL and Virtualization
Memory, Big Data, NoSQL and VirtualizationMemory, Big Data, NoSQL and Virtualization
Memory, Big Data, NoSQL and Virtualization
Bigstep
 
CLFS 2010
CLFS 2010CLFS 2010
CLFS 2010
bergwolf
 
The Smug Mug Tale
The Smug Mug TaleThe Smug Mug Tale
The Smug Mug Tale
MySQLConference
 
Measuring Database Performance on Bare Metal AWS Instances
Measuring Database Performance on Bare Metal AWS InstancesMeasuring Database Performance on Bare Metal AWS Instances
Measuring Database Performance on Bare Metal AWS Instances
ScyllaDB
 
Ceph on All Flash Storage -- Breaking Performance Barriers
Ceph on All Flash Storage -- Breaking Performance BarriersCeph on All Flash Storage -- Breaking Performance Barriers
Ceph on All Flash Storage -- Breaking Performance Barriers
Ceph Community
 
Storage and performance, Whiptail
Storage and performance, Whiptail Storage and performance, Whiptail
Storage and performance, Whiptail
Internet World
 
P99 Pursuit: 8 Years of Battling P99 Latency
P99 Pursuit: 8 Years of Battling P99 LatencyP99 Pursuit: 8 Years of Battling P99 Latency
P99 Pursuit: 8 Years of Battling P99 Latency
ScyllaDB
 
Introduction to Apache Kafka- Part 1
Introduction to Apache Kafka- Part 1Introduction to Apache Kafka- Part 1
Introduction to Apache Kafka- Part 1
Knoldus Inc.
 
Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...
Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...
Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...
Amazon Web Services
 
Aurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 Gaming
Aurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 GamingAurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 Gaming
Aurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 Gaming
Amazon Web Services Korea
 
How to achieve no compromise performance and availability
How to achieve no compromise performance and availabilityHow to achieve no compromise performance and availability
How to achieve no compromise performance and availability
ScyllaDB
 
AF Ceph: Ceph Performance Analysis and Improvement on Flash
AF Ceph: Ceph Performance Analysis and Improvement on FlashAF Ceph: Ceph Performance Analysis and Improvement on Flash
AF Ceph: Ceph Performance Analysis and Improvement on Flash
Ceph Community
 
4 use cases for C* to Scylla
4 use cases for C*  to Scylla4 use cases for C*  to Scylla
4 use cases for C* to Scylla
◄ ★ Jack Pavlov ★ ►
 
Scylla db deck, july 2017
Scylla db deck, july 2017Scylla db deck, july 2017
Scylla db deck, july 2017
Dor Laor
 
Real time data pipline with kafka streams
Real time data pipline with kafka streamsReal time data pipline with kafka streams
Real time data pipline with kafka streams
Yoni Farin
 

Similar to Scylla Summit 2018: Keeping Your Latency SLAs No Matter What! (20)

How is Kafka so Fast?
How is Kafka so Fast?How is Kafka so Fast?
How is Kafka so Fast?
 
Scality S3 Server: Node js Meetup Presentation
Scality S3 Server: Node js Meetup PresentationScality S3 Server: Node js Meetup Presentation
Scality S3 Server: Node js Meetup Presentation
 
AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...
AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...
AdGear Use Case with Scylla - 1M Queries Per Second with Single-Digit Millise...
 
Under The Hood Of A Shard-Per-Core Database Architecture
Under The Hood Of A Shard-Per-Core Database ArchitectureUnder The Hood Of A Shard-Per-Core Database Architecture
Under The Hood Of A Shard-Per-Core Database Architecture
 
Storage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, WhiptailStorage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, Whiptail
 
Memory, Big Data, NoSQL and Virtualization
Memory, Big Data, NoSQL and VirtualizationMemory, Big Data, NoSQL and Virtualization
Memory, Big Data, NoSQL and Virtualization
 
CLFS 2010
CLFS 2010CLFS 2010
CLFS 2010
 
The Smug Mug Tale
The Smug Mug TaleThe Smug Mug Tale
The Smug Mug Tale
 
Measuring Database Performance on Bare Metal AWS Instances
Measuring Database Performance on Bare Metal AWS InstancesMeasuring Database Performance on Bare Metal AWS Instances
Measuring Database Performance on Bare Metal AWS Instances
 
Ceph on All Flash Storage -- Breaking Performance Barriers
Ceph on All Flash Storage -- Breaking Performance BarriersCeph on All Flash Storage -- Breaking Performance Barriers
Ceph on All Flash Storage -- Breaking Performance Barriers
 
Storage and performance, Whiptail
Storage and performance, Whiptail Storage and performance, Whiptail
Storage and performance, Whiptail
 
P99 Pursuit: 8 Years of Battling P99 Latency
P99 Pursuit: 8 Years of Battling P99 LatencyP99 Pursuit: 8 Years of Battling P99 Latency
P99 Pursuit: 8 Years of Battling P99 Latency
 
Introduction to Apache Kafka- Part 1
Introduction to Apache Kafka- Part 1Introduction to Apache Kafka- Part 1
Introduction to Apache Kafka- Part 1
 
Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...
Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...
Deep Dive on the Amazon Aurora MySQL-compatible Edition - DAT301 - re:Invent ...
 
Aurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 Gaming
Aurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 GamingAurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 Gaming
Aurora는 어떻게 다른가 - 김일호 솔루션즈 아키텍트:: AWS Cloud Track 3 Gaming
 
How to achieve no compromise performance and availability
How to achieve no compromise performance and availabilityHow to achieve no compromise performance and availability
How to achieve no compromise performance and availability
 
AF Ceph: Ceph Performance Analysis and Improvement on Flash
AF Ceph: Ceph Performance Analysis and Improvement on FlashAF Ceph: Ceph Performance Analysis and Improvement on Flash
AF Ceph: Ceph Performance Analysis and Improvement on Flash
 
4 use cases for C* to Scylla
4 use cases for C*  to Scylla4 use cases for C*  to Scylla
4 use cases for C* to Scylla
 
Scylla db deck, july 2017
Scylla db deck, july 2017Scylla db deck, july 2017
Scylla db deck, july 2017
 
Real time data pipline with kafka streams
Real time data pipline with kafka streamsReal time data pipline with kafka streams
Real time data pipline with kafka streams
 

More from ScyllaDB

Optimizing NoSQL Performance Through Observability
Optimizing NoSQL Performance Through ObservabilityOptimizing NoSQL Performance Through Observability
Optimizing NoSQL Performance Through Observability
ScyllaDB
 
Event-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream ProcessingEvent-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream Processing
ScyllaDB
 
Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...
Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...
Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...
ScyllaDB
 
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
ScyllaDB
 
Developer Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLDeveloper Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQL
ScyllaDB
 
What Developers Need to Unlearn for High Performance NoSQL
What Developers Need to Unlearn for High Performance NoSQLWhat Developers Need to Unlearn for High Performance NoSQL
What Developers Need to Unlearn for High Performance NoSQL
ScyllaDB
 
Low Latency at Extreme Scale: Proven Practices & Pitfalls
Low Latency at Extreme Scale: Proven Practices & PitfallsLow Latency at Extreme Scale: Proven Practices & Pitfalls
Low Latency at Extreme Scale: Proven Practices & Pitfalls
ScyllaDB
 
Dissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance DilemmasDissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance Dilemmas
ScyllaDB
 
Beyond Linear Scaling: A New Path for Performance with ScyllaDB
Beyond Linear Scaling: A New Path for Performance with ScyllaDBBeyond Linear Scaling: A New Path for Performance with ScyllaDB
Beyond Linear Scaling: A New Path for Performance with ScyllaDB
ScyllaDB
 
Dissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance DilemmasDissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance Dilemmas
ScyllaDB
 
Database Performance at Scale Masterclass: Workload Characteristics by Felipe...
Database Performance at Scale Masterclass: Workload Characteristics by Felipe...Database Performance at Scale Masterclass: Workload Characteristics by Felipe...
Database Performance at Scale Masterclass: Workload Characteristics by Felipe...
ScyllaDB
 
Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...
Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...
Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...
ScyllaDB
 
Database Performance at Scale Masterclass: Driver Strategies by Piotr Sarna
Database Performance at Scale Masterclass: Driver Strategies by Piotr SarnaDatabase Performance at Scale Masterclass: Driver Strategies by Piotr Sarna
Database Performance at Scale Masterclass: Driver Strategies by Piotr Sarna
ScyllaDB
 
Replacing Your Cache with ScyllaDB
Replacing Your Cache with ScyllaDBReplacing Your Cache with ScyllaDB
Replacing Your Cache with ScyllaDB
ScyllaDB
 
Powering Real-Time Apps with ScyllaDB_ Low Latency & Linear Scalability
Powering Real-Time Apps with ScyllaDB_ Low Latency & Linear ScalabilityPowering Real-Time Apps with ScyllaDB_ Low Latency & Linear Scalability
Powering Real-Time Apps with ScyllaDB_ Low Latency & Linear Scalability
ScyllaDB
 
7 Reasons Not to Put an External Cache in Front of Your Database.pptx
7 Reasons Not to Put an External Cache in Front of Your Database.pptx7 Reasons Not to Put an External Cache in Front of Your Database.pptx
7 Reasons Not to Put an External Cache in Front of Your Database.pptx
ScyllaDB
 
Getting the most out of ScyllaDB
Getting the most out of ScyllaDBGetting the most out of ScyllaDB
Getting the most out of ScyllaDB
ScyllaDB
 
NoSQL Database Migration Masterclass - Session 2: The Anatomy of a Migration
NoSQL Database Migration Masterclass - Session 2: The Anatomy of a MigrationNoSQL Database Migration Masterclass - Session 2: The Anatomy of a Migration
NoSQL Database Migration Masterclass - Session 2: The Anatomy of a Migration
ScyllaDB
 
NoSQL Database Migration Masterclass - Session 3: Migration Logistics
NoSQL Database Migration Masterclass - Session 3: Migration LogisticsNoSQL Database Migration Masterclass - Session 3: Migration Logistics
NoSQL Database Migration Masterclass - Session 3: Migration Logistics
ScyllaDB
 
NoSQL Data Migration Masterclass - Session 1 Migration Strategies and Challenges
NoSQL Data Migration Masterclass - Session 1 Migration Strategies and ChallengesNoSQL Data Migration Masterclass - Session 1 Migration Strategies and Challenges
NoSQL Data Migration Masterclass - Session 1 Migration Strategies and Challenges
ScyllaDB
 

More from ScyllaDB (20)

Optimizing NoSQL Performance Through Observability
Optimizing NoSQL Performance Through ObservabilityOptimizing NoSQL Performance Through Observability
Optimizing NoSQL Performance Through Observability
 
Event-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream ProcessingEvent-Driven Architecture Masterclass: Challenges in Stream Processing
Event-Driven Architecture Masterclass: Challenges in Stream Processing
 
Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...
Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...
Event-Driven Architecture Masterclass: Integrating Distributed Data Stores Ac...
 
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
Event-Driven Architecture Masterclass: Engineering a Robust, High-performance...
 
Developer Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLDeveloper Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQL
 
What Developers Need to Unlearn for High Performance NoSQL
What Developers Need to Unlearn for High Performance NoSQLWhat Developers Need to Unlearn for High Performance NoSQL
What Developers Need to Unlearn for High Performance NoSQL
 
Low Latency at Extreme Scale: Proven Practices & Pitfalls
Low Latency at Extreme Scale: Proven Practices & PitfallsLow Latency at Extreme Scale: Proven Practices & Pitfalls
Low Latency at Extreme Scale: Proven Practices & Pitfalls
 
Dissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance DilemmasDissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance Dilemmas
 
Beyond Linear Scaling: A New Path for Performance with ScyllaDB
Beyond Linear Scaling: A New Path for Performance with ScyllaDBBeyond Linear Scaling: A New Path for Performance with ScyllaDB
Beyond Linear Scaling: A New Path for Performance with ScyllaDB
 
Dissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance DilemmasDissecting Real-World Database Performance Dilemmas
Dissecting Real-World Database Performance Dilemmas
 
Database Performance at Scale Masterclass: Workload Characteristics by Felipe...
Database Performance at Scale Masterclass: Workload Characteristics by Felipe...Database Performance at Scale Masterclass: Workload Characteristics by Felipe...
Database Performance at Scale Masterclass: Workload Characteristics by Felipe...
 
Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...
Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...
Database Performance at Scale Masterclass: Database Internals by Pavel Emelya...
 
Database Performance at Scale Masterclass: Driver Strategies by Piotr Sarna
Database Performance at Scale Masterclass: Driver Strategies by Piotr SarnaDatabase Performance at Scale Masterclass: Driver Strategies by Piotr Sarna
Database Performance at Scale Masterclass: Driver Strategies by Piotr Sarna
 
Replacing Your Cache with ScyllaDB
Replacing Your Cache with ScyllaDBReplacing Your Cache with ScyllaDB
Replacing Your Cache with ScyllaDB
 
Powering Real-Time Apps with ScyllaDB_ Low Latency & Linear Scalability
Powering Real-Time Apps with ScyllaDB_ Low Latency & Linear ScalabilityPowering Real-Time Apps with ScyllaDB_ Low Latency & Linear Scalability
Powering Real-Time Apps with ScyllaDB_ Low Latency & Linear Scalability
 
7 Reasons Not to Put an External Cache in Front of Your Database.pptx
7 Reasons Not to Put an External Cache in Front of Your Database.pptx7 Reasons Not to Put an External Cache in Front of Your Database.pptx
7 Reasons Not to Put an External Cache in Front of Your Database.pptx
 
Getting the most out of ScyllaDB
Getting the most out of ScyllaDBGetting the most out of ScyllaDB
Getting the most out of ScyllaDB
 
NoSQL Database Migration Masterclass - Session 2: The Anatomy of a Migration
NoSQL Database Migration Masterclass - Session 2: The Anatomy of a MigrationNoSQL Database Migration Masterclass - Session 2: The Anatomy of a Migration
NoSQL Database Migration Masterclass - Session 2: The Anatomy of a Migration
 
NoSQL Database Migration Masterclass - Session 3: Migration Logistics
NoSQL Database Migration Masterclass - Session 3: Migration LogisticsNoSQL Database Migration Masterclass - Session 3: Migration Logistics
NoSQL Database Migration Masterclass - Session 3: Migration Logistics
 
NoSQL Data Migration Masterclass - Session 1 Migration Strategies and Challenges
NoSQL Data Migration Masterclass - Session 1 Migration Strategies and ChallengesNoSQL Data Migration Masterclass - Session 1 Migration Strategies and Challenges
NoSQL Data Migration Masterclass - Session 1 Migration Strategies and Challenges
 

Recently uploaded

How to write a program in any programming language
How to write a program in any programming languageHow to write a program in any programming language
How to write a program in any programming language
Rakesh Kumar R
 
Oracle Database 19c New Features for DBAs and Developers.pptx
Oracle Database 19c New Features for DBAs and Developers.pptxOracle Database 19c New Features for DBAs and Developers.pptx
Oracle Database 19c New Features for DBAs and Developers.pptx
Remote DBA Services
 
Using Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional SafetyUsing Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional Safety
Ayan Halder
 
socradar-q1-2024-aviation-industry-report.pdf
socradar-q1-2024-aviation-industry-report.pdfsocradar-q1-2024-aviation-industry-report.pdf
socradar-q1-2024-aviation-industry-report.pdf
SOCRadar
 
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
kalichargn70th171
 
Unveiling the Advantages of Agile Software Development.pdf
Unveiling the Advantages of Agile Software Development.pdfUnveiling the Advantages of Agile Software Development.pdf
Unveiling the Advantages of Agile Software Development.pdf
brainerhub1
 
2024 eCommerceDays Toulouse - Sylius 2.0.pdf
2024 eCommerceDays Toulouse - Sylius 2.0.pdf2024 eCommerceDays Toulouse - Sylius 2.0.pdf
2024 eCommerceDays Toulouse - Sylius 2.0.pdf
Łukasz Chruściel
 
UI5con 2024 - Keynote: Latest News about UI5 and it’s Ecosystem
UI5con 2024 - Keynote: Latest News about UI5 and it’s EcosystemUI5con 2024 - Keynote: Latest News about UI5 and it’s Ecosystem
UI5con 2024 - Keynote: Latest News about UI5 and it’s Ecosystem
Peter Muessig
 
Microservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we workMicroservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we work
Sven Peters
 
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOMLORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
lorraineandreiamcidl
 
Why Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise Edition
Why Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise EditionWhy Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise Edition
Why Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise Edition
Envertis Software Solutions
 
Webinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for EmbeddedWebinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for Embedded
ICS
 
Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)
Julian Hyde
 
Using Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query PerformanceUsing Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query Performance
Grant Fritchey
 
E-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian Companies
E-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian CompaniesE-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian Companies
E-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian Companies
Quickdice ERP
 
原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样
原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样
原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样
mz5nrf0n
 
LORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptx
LORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptxLORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptx
LORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptx
lorraineandreiamcidl
 
Hand Rolled Applicative User Validation Code Kata
Hand Rolled Applicative User ValidationCode KataHand Rolled Applicative User ValidationCode Kata
Hand Rolled Applicative User Validation Code Kata
Philip Schwarz
 
Revolutionizing Visual Effects Mastering AI Face Swaps.pdf
Revolutionizing Visual Effects Mastering AI Face Swaps.pdfRevolutionizing Visual Effects Mastering AI Face Swaps.pdf
Revolutionizing Visual Effects Mastering AI Face Swaps.pdf
Undress Baby
 
KuberTENes Birthday Bash Guadalajara - Introducción a Argo CD
KuberTENes Birthday Bash Guadalajara - Introducción a Argo CDKuberTENes Birthday Bash Guadalajara - Introducción a Argo CD
KuberTENes Birthday Bash Guadalajara - Introducción a Argo CD
rodomar2
 

Recently uploaded (20)

How to write a program in any programming language
How to write a program in any programming languageHow to write a program in any programming language
How to write a program in any programming language
 
Oracle Database 19c New Features for DBAs and Developers.pptx
Oracle Database 19c New Features for DBAs and Developers.pptxOracle Database 19c New Features for DBAs and Developers.pptx
Oracle Database 19c New Features for DBAs and Developers.pptx
 
Using Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional SafetyUsing Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional Safety
 
socradar-q1-2024-aviation-industry-report.pdf
socradar-q1-2024-aviation-industry-report.pdfsocradar-q1-2024-aviation-industry-report.pdf
socradar-q1-2024-aviation-industry-report.pdf
 
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
 
Unveiling the Advantages of Agile Software Development.pdf
Unveiling the Advantages of Agile Software Development.pdfUnveiling the Advantages of Agile Software Development.pdf
Unveiling the Advantages of Agile Software Development.pdf
 
2024 eCommerceDays Toulouse - Sylius 2.0.pdf
2024 eCommerceDays Toulouse - Sylius 2.0.pdf2024 eCommerceDays Toulouse - Sylius 2.0.pdf
2024 eCommerceDays Toulouse - Sylius 2.0.pdf
 
UI5con 2024 - Keynote: Latest News about UI5 and it’s Ecosystem
UI5con 2024 - Keynote: Latest News about UI5 and it’s EcosystemUI5con 2024 - Keynote: Latest News about UI5 and it’s Ecosystem
UI5con 2024 - Keynote: Latest News about UI5 and it’s Ecosystem
 
Microservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we workMicroservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we work
 
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOMLORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
 
Why Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise Edition
Why Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise EditionWhy Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise Edition
Why Choose Odoo 17 Community & How it differs from Odoo 17 Enterprise Edition
 
Webinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for EmbeddedWebinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for Embedded
 
Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)
 
Using Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query PerformanceUsing Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query Performance
 
E-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian Companies
E-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian CompaniesE-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian Companies
E-Invoicing Implementation: A Step-by-Step Guide for Saudi Arabian Companies
 
原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样
原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样
原版定制美国纽约州立大学奥尔巴尼分校毕业证学位证书原版一模一样
 
LORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptx
LORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptxLORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptx
LORRAINE ANDREI_LEQUIGAN_HOW TO USE WHATSAPP.pptx
 
Hand Rolled Applicative User Validation Code Kata
Hand Rolled Applicative User ValidationCode KataHand Rolled Applicative User ValidationCode Kata
Hand Rolled Applicative User Validation Code Kata
 
Revolutionizing Visual Effects Mastering AI Face Swaps.pdf
Revolutionizing Visual Effects Mastering AI Face Swaps.pdfRevolutionizing Visual Effects Mastering AI Face Swaps.pdf
Revolutionizing Visual Effects Mastering AI Face Swaps.pdf
 
KuberTENes Birthday Bash Guadalajara - Introducción a Argo CD
KuberTENes Birthday Bash Guadalajara - Introducción a Argo CDKuberTENes Birthday Bash Guadalajara - Introducción a Argo CD
KuberTENes Birthday Bash Guadalajara - Introducción a Argo CD
 

Scylla Summit 2018: Keeping Your Latency SLAs No Matter What!

  • 1. Keeping your latency SLA No matter what! Glauber Costa VP Field Engineering ScyllaDB
  • 2. Presenter bio Glauber Costa is VP of Field Engineering at ScyllaDB. He shares his time between the engineering department working on upcoming Scylla features and helping customers succeed. Before ScyllaDB, Glauber worked with Virtualization in the Linux Kernel for 10 years, with contributions into the Xen and KVM Hypervisors and all sorts of guest functionality and containers.
  • 3. Where do latencies come from?
  • 4. Where do latencies come from?
  • 5. Storage I/O ▪ 4kB reads ▪ 150+ concurrent requests ▪ 300𝜇s avg latency
  • 6. Storage I/O ▪ 128kB reads ▪ ~30 concurrent requests ▪ 2ms avg latency
  • 7. Storage I/O ▪ AWS i3.metal write bandwidth: 6GB/s
  • 8. Storage I/O ▪ AWS i3.metal write bandwidth: 6GB/s ▪ AWS i3.metal read bandwidth: 15GB/s
  • 9. Storage I/O 128kB write128kB write128kB write 12ms worth of latency Storage
  • 10. Storage I/O 512-byte read 128kB write128kB write128kB write 12ms worth of latency Storage at least 12ms!
  • 11. Scylla 2.2 vs Scylla 2.3 Scylla 2.2 --max-io-requests=150 Scylla 2.3 read_iops: 9001 read_bandwidth: 160MB write_iops: 9723 write_bandwidth: 162MB
  • 12. Scylla 2.2 vs Scylla 2.3 (AWS EBS, reads vs compaction) Scylla 2.2 Scylla 2.3 average 8.4ms 6.0ms 40% p50 5.6ms 3.1ms 80% p95 15ms 8ms 85% p99 67ms 63ms 6.3%
  • 13. CPU Scheduling read write read Compaction CPU SSTable write
  • 14. CPU Scheduling read write read Compaction CPU CPU Compaction SSTable write SSTable write read write readread write read
  • 15. CPU Scheduling ▪ Preemption every x 𝜇s - the task quota ▪ Expected latency increase when n more classes run: • n * task_quota • Scylla 2.3 task quota: 500 𝜇s
  • 16. Which tasks to run? 100 shares 100 shares
  • 17. Which tasks to run? 100 shares 50 shares
  • 20. Scylla 2.1, no controller
  • 22. Comparative results Scylla 2.1 Scylla 2.2 Writes/s 354K 418K +18%
  • 23. CPU Scheduling ▪ Preemption every x 𝜇s - the task quota ▪ Expected latency increase when n more classes run: • n * task_quota • Scylla 2.3 task quota: 500 𝜇s
  • 24. CPU Scheduling ▪ Preemption every x 𝜇s - the task quota ▪ Expected latency increase when n more classes run: • n * task_quota • Scylla 2.3 task quota: 500 𝜇s ▪ Failure to preempt: task quota violation
  • 25. Task quota violations ▪ Large memory allocations ▪ Memory allocator itself ▪ The Cache ▪ I/O Subsystem ▪ SSTable writing ▪ Bloom Filter generation ▪ The Linux Kernel
  • 26. Scylla 2.2 vs Scylla 2.3 (99.9th percentile write latency) Scylla 2.2 - 4ms to 6ms Scylla 2.3 - < 4ms
  • 28. We need to preempt more ▪ Preemption uses a thread that sets a flag • Source of context switches ▪ Every preemption does I/O • It is a good thing, but I/O goes to the kernel ▪ New polling API merged to the Linux kernel • No context switches for I/O! Can poll a lot more often
  • 29. We need to preempt more - Linux 4.19 Completion queue shared between kernel and user Disk Write Completion Disk Read Completion Socket is readable Disk Read Completion Socket is readable Database dequeues event notifications from shared ring without a system call Kernel appends event notifications from interrupts SCYLLA KERNEL
  • 30. Shard-aware driver - Scylla 2.3 There is no network hop, both CPUs are in the same system. But worst case, it adds a task quota to request processing. CPU that owns connection CPU that owns the token Token
  • 31. Shard-aware driver - Scylla 2.3 Driver opens connections to all CPUs, and learns which tokens belong to each CPU. Request is sent directly to owner. CPU that owns the token Token
  • 32. Summary ▪ Keeping latencies bounded and predictable is important ▪ Latency-inducing events really come from everywhere - from the Linux kernel itself to the most surprising places in Scylla ▪ We keep investing time year after year to make sure our latencies are even more predictable
  • 33. Thank You Any Questions ? Please stay in touch glauber@scylladb.com @glcst