SlideShare a Scribd company logo
Aerospike aer . o . spike [air-oh- spahyk]
noun, 1. tip of a rocket that enhances speed and stability
Aerospike:
Maximizing
Performance
Improving Performance
© 2012 Aerospike. All rights reserved. Confidential Pg. 2
Network
Aerospike Nodes
Clients
Clients:
• CPU
• Network
• Storage
Aerospike Nodes:
• CPU
• Network
• Storage
Network:
• Network throughput
Improving Performance: Client
➤ CPU
 Loads placed on the CPU vary by language and
application. Easily seen through the top command.
For example, if your application creates random
data, this can take a lot of CPU resources to create
the data.
© 2013 Aerospike. All rights reserved. Confidential Pg. 3
Improving Performance: Client
➤ Network
 Number of connections. Some languages do not
relinquish connections easily. Use “netstat” to find
how many connections are open.
 Bandwidth. Use the sar command to find the
bandwidth usage (e.g. “sar –n DEV 10 10”).
 Transactions per second. This is more subtle, but
most servers have a problem exceeding 90,000
transactions per second due to context switching
limitations. This can most easily be seen by using
“top” and hitting the “1” key. If the percentage of
software interrupts is greater than around 30%, you
are probably hitting this limit.
© 2013 Aerospike. All rights reserved. Confidential Pg. 4
Improving Performance: Client
➤ Storage
 Depends on how data is generated, but disk is a
common bottleneck.
 If reading from file, check disk performance
 If reading from database, check on how fast data can be pulled
from the database.
© 2013 Aerospike. All rights reserved. Confidential Pg. 5
Improving Performance: Network
➤ Long delays due to multiple network hops
between client and server.
➤ Saturated network device (switch, router)
These generally require conversations with the
network admin, but are rarely the issue.
© 2013 Aerospike. All rights reserved. Confidential Pg. 6
Improving Performance: Aerospike Node
➤ CPU
 Easily seen through the top command.
Very rarely the issue
© 2013 Aerospike. All rights reserved. Confidential Pg. 7
➤ Network
 Number of connections
 Bandwidth
 Transactions per second
These are as with the client
© 2013 Aerospike. All rights reserved. Confidential Pg. 8
Improving Performance: Aerospike Node
Calculating Bandwidth Requirements for
Replicated Writes
One consideration in estimating network
bandwidth requirements is that writes
are replicated. So the bandwidth
required for writes must be multiplied
by the replication factor (usually 2).
1. Write sent to row master
2. Latch against simultaneous writes
3. Apply write to master memory
4. Apply write synchronously to
replica(s) memory
5. Queue operations to disk
6. Signal completed transaction
7. Master applies conflict resolution
policy – rollback / rollforward
© 2012 Aerospike. All rights reserved. Pg. 9
master replica
➤ Storage
 Check on disk utilization with iostat.
 Check to see what the performance on the SSD is
like with the ACT.
 Sometimes RAID controllers will be saturated. With
many Aerospike use cases, this will happen with
IOPS before bandwidth.
© 2013 Aerospike. All rights reserved. Confidential Pg. 10
Improving Performance: Aerospike Node
There is rarely a single possible solution. The best
choice will vary. Here are some possible
remedies.
➤ Increase the number of nodes/clients
➤ Increase the number of SSDs
➤ Use faster network equipment
➤ Place servers in closer network proximity
➤ Change equipment (e.g. network, RAID controllers)
➤ Balance load across multiple cores
© 2013 Aerospike. All rights reserved. Confidential Pg. 11
Improving Performance: Remedies

More Related Content

PPTX
Configuring Aerospike - Part 2
PPTX
Configuring Aerospike - Part 1
PDF
Aerospike Hybrid Memory Architecture
PPTX
Redis vs Aerospike
PPTX
Distributing Data The Aerospike Way
PPTX
How to Get a Game Changing Performance Advantage with Intel SSDs and Aerospike
PPT
Aerospike: Key Value Data Access
PPTX
Getting The Most Out Of Your Flash/SSDs
Configuring Aerospike - Part 2
Configuring Aerospike - Part 1
Aerospike Hybrid Memory Architecture
Redis vs Aerospike
Distributing Data The Aerospike Way
How to Get a Game Changing Performance Advantage with Intel SSDs and Aerospike
Aerospike: Key Value Data Access
Getting The Most Out Of Your Flash/SSDs

What's hot (20)

PDF
Using Databases and Containers From Development to Deployment
PPTX
Aerospike Architecture
PDF
Aerospike DB and Storm for real-time analytics
PPTX
WEBINAR: Architectures for Digital Transformation and Next-Generation Systems...
PPT
Predictable Big Data Performance in Real-time
PPTX
Hadoop Storage in the Cloud Native Era
PDF
HBase Tales From the Trenches - Short stories about most common HBase operati...
PPTX
The role of NoSQL in the Next Generation of Financial Informatics
PDF
Troubleshooting redis
PPTX
Practical NoSQL: Accumulo's dirlist Example
PDF
Unlock Bigdata Analytic Efficiency with Ceph Data Lake - Zhang Jian, Fu Yong
PPTX
Ceph Day San Jose - Enable Fast Big Data Analytics on Ceph with Alluxio
PPTX
HBaseCon 2013: Apache HBase on Flash
PPTX
IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
PDF
Basic and Advanced Analysis of Ceph Volume Backend Driver in Cinder - John Haan
PDF
RADOS improvements and roadmap - Greg Farnum, Josh Durgin, Kefu Chai
PPTX
Scaling HDFS at Xiaomi
PDF
Ceph Day San Jose - Object Storage for Big Data
PPTX
Ceph: Low Fail Go Scale
PPTX
Redis on NVMe SSD - Zvika Guz, Samsung
Using Databases and Containers From Development to Deployment
Aerospike Architecture
Aerospike DB and Storm for real-time analytics
WEBINAR: Architectures for Digital Transformation and Next-Generation Systems...
Predictable Big Data Performance in Real-time
Hadoop Storage in the Cloud Native Era
HBase Tales From the Trenches - Short stories about most common HBase operati...
The role of NoSQL in the Next Generation of Financial Informatics
Troubleshooting redis
Practical NoSQL: Accumulo's dirlist Example
Unlock Bigdata Analytic Efficiency with Ceph Data Lake - Zhang Jian, Fu Yong
Ceph Day San Jose - Enable Fast Big Data Analytics on Ceph with Alluxio
HBaseCon 2013: Apache HBase on Flash
IMC Summit 2016 Breakout - Per Minoborg - Work with Multiple Hot Terabytes in...
Basic and Advanced Analysis of Ceph Volume Backend Driver in Cinder - John Haan
RADOS improvements and roadmap - Greg Farnum, Josh Durgin, Kefu Chai
Scaling HDFS at Xiaomi
Ceph Day San Jose - Object Storage for Big Data
Ceph: Low Fail Go Scale
Redis on NVMe SSD - Zvika Guz, Samsung
Ad

Viewers also liked (18)

PPTX
Tectonic Shift: A New Foundation for Data Driven Business
PPTX
There are 250 Database products, are you running the right one?
PPTX
Introduction to Aerospike
PPTX
Developing High Performance Application with Aerospike & Go
PPTX
Live Analytics with Go & Aerospike
PDF
Design and flow simulation of truncated aerospike nozzle
PPTX
Leveraging Big Data with Hadoop, NoSQL and RDBMS
PDF
Storm Persistence and Real-Time Analytics
PDF
2017 DB Trends for Powering Real-Time Systems of Engagement
PDF
Building ZingMe News Feed System
PDF
01282016 Aerospike-Docker webinar
PDF
Zing Me Real Time Web Chat Architect
PDF
Design a scalable social network: Problems and solutions
PPTX
Inside Zalo: Developing a mobile messenger for the audience of millions - VN ...
PDF
Riak Search 2.0を使ったデータ集計
PPTX
Running a High Performance NoSQL Database on Amazon EC2 for Just $1.68/Hour
PDF
What the Spark!? Intro and Use Cases
PPTX
The TCP/IP Stack in the Linux Kernel
Tectonic Shift: A New Foundation for Data Driven Business
There are 250 Database products, are you running the right one?
Introduction to Aerospike
Developing High Performance Application with Aerospike & Go
Live Analytics with Go & Aerospike
Design and flow simulation of truncated aerospike nozzle
Leveraging Big Data with Hadoop, NoSQL and RDBMS
Storm Persistence and Real-Time Analytics
2017 DB Trends for Powering Real-Time Systems of Engagement
Building ZingMe News Feed System
01282016 Aerospike-Docker webinar
Zing Me Real Time Web Chat Architect
Design a scalable social network: Problems and solutions
Inside Zalo: Developing a mobile messenger for the audience of millions - VN ...
Riak Search 2.0を使ったデータ集計
Running a High Performance NoSQL Database on Amazon EC2 for Just $1.68/Hour
What the Spark!? Intro and Use Cases
The TCP/IP Stack in the Linux Kernel
Ad

Similar to Aerospike: Maximizing Performance (20)

PDF
Aerospike AdTech Gets Hacked in Lower Manhattan
PDF
You Snooze You Lose or How to Win in Ad Tech?
PPTX
Scale and Throughput @ Clicktale with Akka
ODP
Nagios Conference 2011 - Nicholas Scott - Nagios Performance Tuning
PPT
fdocuments.in_aerospike-key-value-data-access.ppt
PPTX
Fastest Servlets in the West
PDF
LCA13: Web Server and Caching Technologies
PPTX
SQLintersection keynote a tale of two teams
PDF
Art of the Possible_Tim Faulkes.pdf
PDF
Webcenter application performance tuning guide
PDF
EVCache: Lowering Costs for a Low Latency Cache with RocksDB
PDF
Netflix Open Source Meetup Season 4 Episode 2
PDF
What a Modern Database Enables_Srini Srinivasan.pdf
PDF
Linux Admin optimizing drbd performance guide.pdf
PPTX
All about Azure workshop deck
PPTX
Breakout - Airheads Macau 2013 - Top 10 Tips from Aruba TAC
PPTX
Trouble shooting Storage Area Networks for virtualisation deployments
PDF
Acug datafiniti pellon_sept2013
PPTX
Технологии работы с дисковыми хранилищами и файловыми системами Windows Serve...
PDF
Performance Tuning Oracle Weblogic Server 12c
Aerospike AdTech Gets Hacked in Lower Manhattan
You Snooze You Lose or How to Win in Ad Tech?
Scale and Throughput @ Clicktale with Akka
Nagios Conference 2011 - Nicholas Scott - Nagios Performance Tuning
fdocuments.in_aerospike-key-value-data-access.ppt
Fastest Servlets in the West
LCA13: Web Server and Caching Technologies
SQLintersection keynote a tale of two teams
Art of the Possible_Tim Faulkes.pdf
Webcenter application performance tuning guide
EVCache: Lowering Costs for a Low Latency Cache with RocksDB
Netflix Open Source Meetup Season 4 Episode 2
What a Modern Database Enables_Srini Srinivasan.pdf
Linux Admin optimizing drbd performance guide.pdf
All about Azure workshop deck
Breakout - Airheads Macau 2013 - Top 10 Tips from Aruba TAC
Trouble shooting Storage Area Networks for virtualisation deployments
Acug datafiniti pellon_sept2013
Технологии работы с дисковыми хранилищами и файловыми системами Windows Serve...
Performance Tuning Oracle Weblogic Server 12c

Recently uploaded (20)

PDF
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
PDF
Per capita expenditure prediction using model stacking based on satellite ima...
PDF
Electronic commerce courselecture one. Pdf
PPTX
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
PDF
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
PDF
CIFDAQ's Market Insight: SEC Turns Pro Crypto
PPTX
VMware vSphere Foundation How to Sell Presentation-Ver1.4-2-14-2024.pptx
PDF
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
PDF
Approach and Philosophy of On baking technology
DOCX
The AUB Centre for AI in Media Proposal.docx
PDF
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
PDF
KodekX | Application Modernization Development
PDF
Encapsulation theory and applications.pdf
PDF
Dropbox Q2 2025 Financial Results & Investor Presentation
PPTX
Digital-Transformation-Roadmap-for-Companies.pptx
PPTX
20250228 LYD VKU AI Blended-Learning.pptx
PDF
Reach Out and Touch Someone: Haptics and Empathic Computing
PDF
NewMind AI Monthly Chronicles - July 2025
PDF
NewMind AI Weekly Chronicles - August'25 Week I
PPT
“AI and Expert System Decision Support & Business Intelligence Systems”
7 ChatGPT Prompts to Help You Define Your Ideal Customer Profile.pdf
Per capita expenditure prediction using model stacking based on satellite ima...
Electronic commerce courselecture one. Pdf
Detection-First SIEM: Rule Types, Dashboards, and Threat-Informed Strategy
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
CIFDAQ's Market Insight: SEC Turns Pro Crypto
VMware vSphere Foundation How to Sell Presentation-Ver1.4-2-14-2024.pptx
How UI/UX Design Impacts User Retention in Mobile Apps.pdf
Approach and Philosophy of On baking technology
The AUB Centre for AI in Media Proposal.docx
TokAI - TikTok AI Agent : The First AI Application That Analyzes 10,000+ Vira...
KodekX | Application Modernization Development
Encapsulation theory and applications.pdf
Dropbox Q2 2025 Financial Results & Investor Presentation
Digital-Transformation-Roadmap-for-Companies.pptx
20250228 LYD VKU AI Blended-Learning.pptx
Reach Out and Touch Someone: Haptics and Empathic Computing
NewMind AI Monthly Chronicles - July 2025
NewMind AI Weekly Chronicles - August'25 Week I
“AI and Expert System Decision Support & Business Intelligence Systems”

Aerospike: Maximizing Performance

  • 1. Aerospike aer . o . spike [air-oh- spahyk] noun, 1. tip of a rocket that enhances speed and stability Aerospike: Maximizing Performance
  • 2. Improving Performance © 2012 Aerospike. All rights reserved. Confidential Pg. 2 Network Aerospike Nodes Clients Clients: • CPU • Network • Storage Aerospike Nodes: • CPU • Network • Storage Network: • Network throughput
  • 3. Improving Performance: Client ➤ CPU  Loads placed on the CPU vary by language and application. Easily seen through the top command. For example, if your application creates random data, this can take a lot of CPU resources to create the data. © 2013 Aerospike. All rights reserved. Confidential Pg. 3
  • 4. Improving Performance: Client ➤ Network  Number of connections. Some languages do not relinquish connections easily. Use “netstat” to find how many connections are open.  Bandwidth. Use the sar command to find the bandwidth usage (e.g. “sar –n DEV 10 10”).  Transactions per second. This is more subtle, but most servers have a problem exceeding 90,000 transactions per second due to context switching limitations. This can most easily be seen by using “top” and hitting the “1” key. If the percentage of software interrupts is greater than around 30%, you are probably hitting this limit. © 2013 Aerospike. All rights reserved. Confidential Pg. 4
  • 5. Improving Performance: Client ➤ Storage  Depends on how data is generated, but disk is a common bottleneck.  If reading from file, check disk performance  If reading from database, check on how fast data can be pulled from the database. © 2013 Aerospike. All rights reserved. Confidential Pg. 5
  • 6. Improving Performance: Network ➤ Long delays due to multiple network hops between client and server. ➤ Saturated network device (switch, router) These generally require conversations with the network admin, but are rarely the issue. © 2013 Aerospike. All rights reserved. Confidential Pg. 6
  • 7. Improving Performance: Aerospike Node ➤ CPU  Easily seen through the top command. Very rarely the issue © 2013 Aerospike. All rights reserved. Confidential Pg. 7
  • 8. ➤ Network  Number of connections  Bandwidth  Transactions per second These are as with the client © 2013 Aerospike. All rights reserved. Confidential Pg. 8 Improving Performance: Aerospike Node
  • 9. Calculating Bandwidth Requirements for Replicated Writes One consideration in estimating network bandwidth requirements is that writes are replicated. So the bandwidth required for writes must be multiplied by the replication factor (usually 2). 1. Write sent to row master 2. Latch against simultaneous writes 3. Apply write to master memory 4. Apply write synchronously to replica(s) memory 5. Queue operations to disk 6. Signal completed transaction 7. Master applies conflict resolution policy – rollback / rollforward © 2012 Aerospike. All rights reserved. Pg. 9 master replica
  • 10. ➤ Storage  Check on disk utilization with iostat.  Check to see what the performance on the SSD is like with the ACT.  Sometimes RAID controllers will be saturated. With many Aerospike use cases, this will happen with IOPS before bandwidth. © 2013 Aerospike. All rights reserved. Confidential Pg. 10 Improving Performance: Aerospike Node
  • 11. There is rarely a single possible solution. The best choice will vary. Here are some possible remedies. ➤ Increase the number of nodes/clients ➤ Increase the number of SSDs ➤ Use faster network equipment ➤ Place servers in closer network proximity ➤ Change equipment (e.g. network, RAID controllers) ➤ Balance load across multiple cores © 2013 Aerospike. All rights reserved. Confidential Pg. 11 Improving Performance: Remedies

Editor's Notes

  • #2: FastestBest uptimePredictable performanceconsistency