SlideShare a Scribd company logo
1 of 28
Download to read offline
Tuning Storage
Subsystem for
Databases
Angelo Rajadurai
Agenda

Performance issues in Storage
Hybrid Storage (Disks, SSDs, Memory)
ZFS - Not Just Another File System
Tuning for databases (General principles)
Tuning for MySQL
Tuning for PostgreSQL
Tuning for Oracle
Why?
• Some very practical advice based on
  > recent test results
     > Improved pgbench results from 70 tps for pure disk to 5003 tps with
       SSD and tuning
     > Improved sysbench results from 425 tps to 1811 tps with SSD and
       tuning for read/write.
     > Improved sysbench results from 786 tps to 3085 tps with SSD and
       tuning for read.
  > collection of tuning knowledge from Sun performance
    engineers and the community
• Some very good resources at the end of the talk for
  further study
Storage Performance

                                                                       Cache

                                                                 Memory




                                                0,0 00 X       t ial
                                             10              en
                                                         fer
                                                   e dif
                                                nc
                           Disk            r ma
                   High   Cache        r fo
            Performance Disks     Pe

  Large Capacity
      Disks
Latency Comparison
Bridging the DRAM to HDD Gap




     1S

  100mS

    10m
      S
    1mS

   100uS

    10uS                                      TAPE

      1u                                HDD
       S
   100nS                       FLASH/
                                SSD
    10nS
                   DRAM
      1n
       S   CPU
Storage Technology
                Price, Performance & Capacity


                Capacity     Latency             Cost/IOPS    Cost/GB
Technologies                            IOPs
                 (GB)       (microS)                ($)         ($)

   Cloud
  Storage
                Unlimited   60,000       20       17c/GB     0.15/month

 Capacity
  HDDs           2,500      12,000      250        1.67         0.15

Performance
   HDDs           300        7,000      500        1.52         1.30

   SSDs
  (write)          64         300       5000       0.20          13

    SSDs
 (read only)       64         45       30,000      0.03          13

  DRAM             8         0.005     500,000    0.001         52
Incorporating Flash
     Storage Hierarchy
Hybrid Storage
                 Flash as Cache


                   Application


         DRAM
       Level 1 Cache




                                                  Write
Read




                                    Flash
                                 Write side Log
         Flash
       Level 2 Cache




              Disk Primary Storage
ZFS - Last Word in Filesystem
     Pooled Storage Vs Traditional Volumes
Data Management Unit
  Smarts Built Right Into the Filesystem
Administering ZFS in two slides
                            As easy as pie


• zpool commands
 > create a single disk pool:
        # zpool create newpool diskname
 > create a pool with a mirror
        # zpool create newpool mirror disk1name disk2name
 > Add device to a pool:
        # zpool add poolname diskname
 > Replace a bad disk
        # zpool replace poolname baddiskname newdiskname
 > History of commands on the pool:
        # zpool history poolname
 > How is my pool performing:
        # zpool iostat poolname
 No format command, No fdisk partitions, No volumes
Administering ZFS in two slides
                            As easy as pie


• zfs commands
 > create a filesystem:
          # zfs create poolname/fs-name
 > set filesystem property:
          # zfs set quota=size poolname/fs-name
          # zfs set compression=on poolname/fs-name
          # zfs set nfsshare=on poolname/fs-name
          # zfs set recordsize=16k poolname/fs-name
 > get filesystem property:
          # zfs get compressratio poolname/fs-name
          # zfs get all poolname/fs-name
 > snapshot the filesystem:
          # zfs snapshot poolname/fs-name@snapshotname
 No newfs, No mkfs, No /etc/vfstab, No fsck
ZFS and Hybrid Storage
                          As easy as pie


• Read side
  > Add ssd as a read side cache
  > # zpool add poolname cache ssd-device
• Write side
  > Add SSD as a ZFS Intent Log device
  > # zpool add poolname log ssd-device
ZFS Performance Features
• Copy-on-write
  > Turns Random writes to Sequential writes
• Dynamic Striping across all devices
  > Maximize throughput
• Multiple Block Sizes
  > Automatically chosen to match workload
• IO Pipelining
  > Priority/Deadline scheduling, sorting, aggregation
• Intelligent prefetch
• Compression - Improves performance & Capacity
• Can safely use write cache on disks
Databases
                      Not Just Another Application
• Most Databases do their own buffering
  > Filesystem caching can get in the way
  > “double buffer” problem
• Most Databases do “prefetch”
  > Filesystems prefetch can cause extra IO
  > “directio” gets filesystem out of the way
• Have their own “log” mechanism.
  > Interesting interaction with a transaction based filesystem
• Multiple blocks sizes
  > Database & Transaction log, block sizes are normally different
Tuning ZFS for Databases
                        Tuning is Evil - Long live Tuning


• In general tuning is evil. Let ZFS do it for you.
• A few fine tuning tips for databases
  > Get to the latest update of OS
  > Set the recordsize to match database
      block size
  >   Separate Transaction logs and data
      onto separate zpools
  >   [Note: This will be addressed with the ZIL bypass property fix]
  >   Reduce the impact of double buffering by changing the caching
      method to “metadata only”
  >   Use separate ZIL (ZFS Intent Log) preferably SSD
  >   Use SSD as secondary cache - L2ARC (Level 2. Adaptive
      Replacement Cache)
ZFS tuning for MySQL
• Many tuning depends on storage engine
• For Innodb
  > Prefer to cache in Innodb rather than ARC
     zfs set primarycache=metadata poolname/database
  > Set recordsize to 16k for data and 128k for log
     zfs set recordsize=16k poolname/database
     (Note: do this before you load any data)
  > Turn off prefetch
     set zfs:zfs_prefetch_disable = 1 (in /etc/system)
     (File level prefetch not triggered if you change record size to 16k)
  > Use raid0 or mirror over raidz
     raidz is no suitable for random IO
  > Add SSDs for either read side or write side based on workload
     zpool add datapool cache ssd-disk
     zpool create logpool ssd-disk3
        In my.cnf set innodb_data_home_dir & innodb_log_group_home_dir
ZFS tuning for MySQL

• More tuning for Innodb
  > Some device vendors flush cache even
    when not needed. (eg. battery backed cache)
    set zfs:zfs_nocacheflush = 1
  > Turn on compression
    zfs set compression=on poolname/database
     ZFS does not turn on compression if less than 12.5% saving.
     IO reduction may offset the extra cpu cost
  > Disable double writes
    innodb_doublewrite=0 (in my.cnf)
    ZFS does not allow any partial writes so no need to guard against it.
ZFS tuning for PostgreSQL

• Postgres tuning hints
  > Set recordsize to 8k
     zfs set recordsize=8k poolname/database
  > Turn down ARC cache.
     set zfs:zfs_arc_max in /etc/system
  > Add SSDs for either read side or write side based on workload
     zpool add poolname cache ssd-name
     zpool add poolname log ssd-name
  > Use separate pool for log (preferably one with SSD) & data
     initdb -X log_directory_name
     create tablespace datatbs location 'database_directory_name'
     create database mydb with  tablespace datatbs
  > Don’t forget to basic Postgres tuning on Solaris - (huge gains)
     Set shared_buffers, temp_buffers, work_mem, maintenance_work_mem,
     wal_sync_method, synchronous_commits etc
      see: http://blogs.sun.com/jkshah/entry/best_practices_with_postgresql_8
ZFS tuning for Oracle

• Oracle tuning hints
  > Set recordsize to match db_block_size (default 8k)
     zfs set recordsize=8k poolname/database
  > Use separate pool for Oracle logs
     make sure record size of the log filesystem is left to the 128k default
  > Add SSDs for either read side or write side based on workload
     zpool add poolname cache ssd-name
     zpool add poolname log ssd-name
Benchmark results

• Hardware
  > Sun x4150 2 x Quad core 2.3 GHz Xeon
     12 GB ram
      3 x 10000 rpm drives
      3 x 32 GB SSDs
• Software
  > OpenSolaris 2009.06
  > Postgres 8.3.7
  > MySQL 5.4 beta
Benchmark results

• pgbench & Postgres
  > command line: pgbench -c 10 -s 10 -t 10000 pgbench
                Description              TPS
 Single disk ZFS                          72 tps
 2 Raid 0 disk + SSD as level 2 cache    241 tps
 Above + general postgres optimization   2026 tps
 + all the data on SSD                   2603 tps
 + data on hdd & log on SSD              4372 tps
 + primarycache=metadata                 5003 tps
Benchmark results

• sysbench & mysql 5.4
  > read/write test: sysbench --max-time=300 --max-requests=0 --test=oltp --
     oltp-dist-type=special --oltp-table-size=10000000 --num-threads=20 run


                 Description                        TPS
 Single disk ZFS                                    425 tps
 raid0 ZFS                                          670 tps
   + SSD cache                                      788 tps
   + Separate intent log                            1352 tps
   + With optimization                              1809 tps
Benchmark results

• sysbench & mysql 5.4
  > read test: sysbench --max-time=300 --max-requests=0 --test=oltp --oltp-
     dist-type=special --oltp-table-size=10000000 --num-threads=20 --oltp-read-
     only=on run

                  Description                         TPS
  Single disk ZFS                                     786 tps
  2 disk raid0 ZFS                                   1501 tps
    + SSD cache                                      1981 tps
    + Separate intent log on SSD                     2567 tps
    + optimization                                   3065 tps
!"#$!%&'()*$'&(+,(-$.//012.$


Sun Unified Storage
                        *#%'345*6*5$!%(#+(5&#*
         :$;<$=*='&-$>$?@A9BB;<$!6!$1)CDC

                                               78)74+*#!8%3$!9(5(:5*
                      E($+'$F@;<$=*='&-G$142A?H<$!6H6$1)CDC
                           E($+'$IA?:$;< write-optimized SSDs
                                                                        *#%'345*6*5$95"!%*'49(-(:5*
                                                         E($+'$F@;<$=*='&-G$ZFA?H<$!6H6$1)CDC
                                                        E($+'$FA?BB;<$&*,1.?FA?:;<$J&)+*$!!5C
                                                                  6K+)L*M6K+)L*$/2"C+*&)#N$0'&$O.6
                                                                                                                   !9(5(:5*$95"!%*'49(-(:5*
          !%(#+('+$@*(%"'*!$>(55$,&+*5!?
  622$5,+,$%&'+'K'2C$,#1$5,+,$!*&L)K*C$V#K2"1*1                                                     E($+'$?I:;<$=*='&-G$I::A?H<$!6H6$1)CDC
S6#,2-+)KCG$!#,(CR'+CG$7*(2)K,+)'#G$/'=(&*CC)'#G$4U!G$/VU!G$)!/!VWT                                  E($+'$FA?BB;<$&*,1.?FA?:;<$J&)+*$!!5C
           <")2+M)#$@A?$;P$Q+R*&#*+$('&+C                                                                      6K+)L*M6K+)L*$/2"C+*&)#N$0'&$O.6
         7*='+*$/'#C'2*$S!*&),2$'&$Q+R*&#*+T
              X)NR+CM'"+$Y,#,N*=*#+

                                          &-%8&#(5$#*%;&'<8#)$=$9&##*9%868%3$>(55$,&+*5!?
                                                      IA?B$;P$Q+R*&#*+$S'(+)K,2T
                                           @A?$;P$Q+R*&#*+$SK'((*&T$>$IA?;P$Q+R*&#*+$S'(+)K,2T
                                                   U/$'&$!/!V$O<6$0'&$+,(*$P,KD"(


                                                         !"#$ %&'(&)*+,&-./'#0)1*#+),23$ 456$ 7*8")&*1                                            9
Getting these systems at a discount
Sun Startup Essentials
                         •   Exclusive program for startups
                         •   Eligibility <6 yrs. Old, <150
 sun.com/startup             employees
                         •   Co-marketing opportunities
                         •   Funding assistance
                         •   Deeply discounted storage and
                             servers certified for Linux,
                             Windows, and Solaris
                         •   Hosting starting at $40
                         •   Open source software, and
                             discounted MySQL
                         •   Free email based tech support
                         •   Free and discounted training on
                             Sun technologies
                         •   Member-only webinars
Resources
• ZFS info: http://www.opensolaris.org/os/community/zfs/
• ZFS Best Practices Guide:
  http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide
• ZFS Evil Tuning Guide:
  http://www.solarisinternals.com/wiki/index.php/ZFS_Evil_Tuning_Guide
• Blogs of note:
   > All things performance tuning:
     http://blogs.sun.com/realneel
     http://blogs.sun.com/roch
   > Postgres tuning - Jignesh’s Blog
     http://blogs.sun.com/jkshah
   > Angelo’s blog
     http://blogs.sun.com/angelo
Tuning Storage
Subsystem for
Databases
Angelo Rajadurai
angelo@sun.com
http://blogs.sun.com/angelo
twitter: rajadurai

More Related Content

What's hot

Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu
Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu
Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu 康志強 大人
 
openSUSE storage workshop 2016
openSUSE storage workshop 2016openSUSE storage workshop 2016
openSUSE storage workshop 2016Alex Lau
 
Uptime Database Appliance - Technology Preview
Uptime Database Appliance - Technology PreviewUptime Database Appliance - Technology Preview
Uptime Database Appliance - Technology PreviewUptime Technologies LLC
 
Azure VM 101 - HomeGen by CloudGen Verona - Marco Obinu
Azure VM 101 - HomeGen by CloudGen Verona - Marco ObinuAzure VM 101 - HomeGen by CloudGen Verona - Marco Obinu
Azure VM 101 - HomeGen by CloudGen Verona - Marco ObinuMarco Obinu
 
JetStor 8 series 16G FC 12G SAS units
JetStor 8 series 16G FC 12G SAS unitsJetStor 8 series 16G FC 12G SAS units
JetStor 8 series 16G FC 12G SAS unitsGene Leyzarovich
 
ZFS Workshop
ZFS WorkshopZFS Workshop
ZFS WorkshopAPNIC
 
Comparison of-foss-distributed-storage
Comparison of-foss-distributed-storageComparison of-foss-distributed-storage
Comparison of-foss-distributed-storageMarian Marinov
 
Linux con europe_2014_f
Linux con europe_2014_fLinux con europe_2014_f
Linux con europe_2014_fsprdd
 
LizardFS-WhitePaper-Eng-v3.9.2-web
LizardFS-WhitePaper-Eng-v3.9.2-webLizardFS-WhitePaper-Eng-v3.9.2-web
LizardFS-WhitePaper-Eng-v3.9.2-webSzymon Haly
 
Performance comparison of Distributed File Systems on 1Gbit networks
Performance comparison of Distributed File Systems on 1Gbit networksPerformance comparison of Distributed File Systems on 1Gbit networks
Performance comparison of Distributed File Systems on 1Gbit networksMarian Marinov
 
Modern CPUs and Caches - A Starting Point for Programmers
Modern CPUs and Caches - A Starting Point for ProgrammersModern CPUs and Caches - A Starting Point for Programmers
Modern CPUs and Caches - A Starting Point for ProgrammersYaser Zhian
 
SSD Deployment Strategies for MySQL
SSD Deployment Strategies for MySQLSSD Deployment Strategies for MySQL
SSD Deployment Strategies for MySQLYoshinori Matsunobu
 
RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)
RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)
RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)Isabella789
 
MySQL Replication: Demo Réplica en Español
MySQL Replication: Demo Réplica en EspañolMySQL Replication: Demo Réplica en Español
MySQL Replication: Demo Réplica en EspañolKeith Hollman
 
Perf Vsphere Storage Protocols
Perf Vsphere Storage ProtocolsPerf Vsphere Storage Protocols
Perf Vsphere Storage ProtocolsYanghua Zhang
 
Collaborate vdb performance
Collaborate vdb performanceCollaborate vdb performance
Collaborate vdb performanceKyle Hailey
 

What's hot (20)

Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu
Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu
Hadoop 2.2.0 Multi-node cluster Installation on Ubuntu
 
openSUSE storage workshop 2016
openSUSE storage workshop 2016openSUSE storage workshop 2016
openSUSE storage workshop 2016
 
ZFS Talk Part 1
ZFS Talk Part 1ZFS Talk Part 1
ZFS Talk Part 1
 
Uptime Database Appliance - Technology Preview
Uptime Database Appliance - Technology PreviewUptime Database Appliance - Technology Preview
Uptime Database Appliance - Technology Preview
 
Azure VM 101 - HomeGen by CloudGen Verona - Marco Obinu
Azure VM 101 - HomeGen by CloudGen Verona - Marco ObinuAzure VM 101 - HomeGen by CloudGen Verona - Marco Obinu
Azure VM 101 - HomeGen by CloudGen Verona - Marco Obinu
 
Hadoop 3.1.1 single node
Hadoop 3.1.1 single nodeHadoop 3.1.1 single node
Hadoop 3.1.1 single node
 
JetStor 8 series 16G FC 12G SAS units
JetStor 8 series 16G FC 12G SAS unitsJetStor 8 series 16G FC 12G SAS units
JetStor 8 series 16G FC 12G SAS units
 
ZFS Workshop
ZFS WorkshopZFS Workshop
ZFS Workshop
 
Comparison of-foss-distributed-storage
Comparison of-foss-distributed-storageComparison of-foss-distributed-storage
Comparison of-foss-distributed-storage
 
Linux con europe_2014_f
Linux con europe_2014_fLinux con europe_2014_f
Linux con europe_2014_f
 
LizardFS-WhitePaper-Eng-v3.9.2-web
LizardFS-WhitePaper-Eng-v3.9.2-webLizardFS-WhitePaper-Eng-v3.9.2-web
LizardFS-WhitePaper-Eng-v3.9.2-web
 
Performance comparison of Distributed File Systems on 1Gbit networks
Performance comparison of Distributed File Systems on 1Gbit networksPerformance comparison of Distributed File Systems on 1Gbit networks
Performance comparison of Distributed File Systems on 1Gbit networks
 
Modern CPUs and Caches - A Starting Point for Programmers
Modern CPUs and Caches - A Starting Point for ProgrammersModern CPUs and Caches - A Starting Point for Programmers
Modern CPUs and Caches - A Starting Point for Programmers
 
SSD Deployment Strategies for MySQL
SSD Deployment Strategies for MySQLSSD Deployment Strategies for MySQL
SSD Deployment Strategies for MySQL
 
RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)
RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)
RH-302 Exam-Red Hat Certified Engineer on Redhat Enterprise Linux 4 (Labs)
 
MySQL Replication: Demo Réplica en Español
MySQL Replication: Demo Réplica en EspañolMySQL Replication: Demo Réplica en Español
MySQL Replication: Demo Réplica en Español
 
Bluestore
BluestoreBluestore
Bluestore
 
Perf Vsphere Storage Protocols
Perf Vsphere Storage ProtocolsPerf Vsphere Storage Protocols
Perf Vsphere Storage Protocols
 
How swift is your Swift - SD.pptx
How swift is your Swift - SD.pptxHow swift is your Swift - SD.pptx
How swift is your Swift - SD.pptx
 
Collaborate vdb performance
Collaborate vdb performanceCollaborate vdb performance
Collaborate vdb performance
 

Similar to Tuning Storage Subsystems for Optimal Database Performance

Open Source Data Deduplication
Open Source Data DeduplicationOpen Source Data Deduplication
Open Source Data DeduplicationRedWireServices
 
Exploiting Your File System to Build Robust & Efficient Workflows
Exploiting Your File System to Build Robust & Efficient WorkflowsExploiting Your File System to Build Robust & Efficient Workflows
Exploiting Your File System to Build Robust & Efficient Workflowsjasonajohnson
 
Ssd And Enteprise Storage
Ssd And Enteprise StorageSsd And Enteprise Storage
Ssd And Enteprise StorageFrank Zhao
 
Demystifying Storage - Building large SANs
Demystifying  Storage - Building large SANsDemystifying  Storage - Building large SANs
Demystifying Storage - Building large SANsDirecti Group
 
Ceph Performance and Sizing Guide
Ceph Performance and Sizing GuideCeph Performance and Sizing Guide
Ceph Performance and Sizing GuideJose De La Rosa
 
Disk IO Benchmarking in shared multi-tenant environments
Disk IO Benchmarking in shared multi-tenant environmentsDisk IO Benchmarking in shared multi-tenant environments
Disk IO Benchmarking in shared multi-tenant environmentsRodrigo Campos
 
Dumb Simple PostgreSQL Performance (NYCPUG)
Dumb Simple PostgreSQL Performance (NYCPUG)Dumb Simple PostgreSQL Performance (NYCPUG)
Dumb Simple PostgreSQL Performance (NYCPUG)Joshua Drake
 
Demystifying Storage
Demystifying  StorageDemystifying  Storage
Demystifying Storagebhavintu79
 
MySQL Oslayer performace optimization
MySQL  Oslayer performace optimizationMySQL  Oslayer performace optimization
MySQL Oslayer performace optimizationLouis liu
 
Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]
Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]
Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]Kyle Hailey
 
SOUG_SDM_OracleDB_V3
SOUG_SDM_OracleDB_V3SOUG_SDM_OracleDB_V3
SOUG_SDM_OracleDB_V3UniFabric
 
Linux and H/W optimizations for MySQL
Linux and H/W optimizations for MySQLLinux and H/W optimizations for MySQL
Linux and H/W optimizations for MySQLYoshinori Matsunobu
 
San presentation nov 2012 central pa
San presentation nov 2012 central paSan presentation nov 2012 central pa
San presentation nov 2012 central paJoseph D'Antoni
 
Comparison of foss distributed storage
Comparison of foss distributed storageComparison of foss distributed storage
Comparison of foss distributed storageMarian Marinov
 
PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016
PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016
PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016Tomas Vondra
 
What is the average rotational latency of this disk drive What seek.docx
 What is the average rotational latency of this disk drive  What seek.docx What is the average rotational latency of this disk drive  What seek.docx
What is the average rotational latency of this disk drive What seek.docxajoy21
 
Ceph Day San Jose - Red Hat Storage Acceleration Utlizing Flash Technology
Ceph Day San Jose - Red Hat Storage Acceleration Utlizing Flash TechnologyCeph Day San Jose - Red Hat Storage Acceleration Utlizing Flash Technology
Ceph Day San Jose - Red Hat Storage Acceleration Utlizing Flash TechnologyCeph Community
 

Similar to Tuning Storage Subsystems for Optimal Database Performance (20)

Open Source Data Deduplication
Open Source Data DeduplicationOpen Source Data Deduplication
Open Source Data Deduplication
 
Exploiting Your File System to Build Robust & Efficient Workflows
Exploiting Your File System to Build Robust & Efficient WorkflowsExploiting Your File System to Build Robust & Efficient Workflows
Exploiting Your File System to Build Robust & Efficient Workflows
 
Ssd And Enteprise Storage
Ssd And Enteprise StorageSsd And Enteprise Storage
Ssd And Enteprise Storage
 
Demystifying Storage - Building large SANs
Demystifying  Storage - Building large SANsDemystifying  Storage - Building large SANs
Demystifying Storage - Building large SANs
 
IO Dubi Lebel
IO Dubi LebelIO Dubi Lebel
IO Dubi Lebel
 
Ceph Performance and Sizing Guide
Ceph Performance and Sizing GuideCeph Performance and Sizing Guide
Ceph Performance and Sizing Guide
 
Disk IO Benchmarking in shared multi-tenant environments
Disk IO Benchmarking in shared multi-tenant environmentsDisk IO Benchmarking in shared multi-tenant environments
Disk IO Benchmarking in shared multi-tenant environments
 
Dumb Simple PostgreSQL Performance (NYCPUG)
Dumb Simple PostgreSQL Performance (NYCPUG)Dumb Simple PostgreSQL Performance (NYCPUG)
Dumb Simple PostgreSQL Performance (NYCPUG)
 
Demystifying Storage
Demystifying  StorageDemystifying  Storage
Demystifying Storage
 
MySQL Oslayer performace optimization
MySQL  Oslayer performace optimizationMySQL  Oslayer performace optimization
MySQL Oslayer performace optimization
 
JetStor NAS series 2016
JetStor NAS series 2016JetStor NAS series 2016
JetStor NAS series 2016
 
Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]
Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]
Oracle Open World 2014: Lies, Damned Lies, and I/O Statistics [ CON3671]
 
SOUG_SDM_OracleDB_V3
SOUG_SDM_OracleDB_V3SOUG_SDM_OracleDB_V3
SOUG_SDM_OracleDB_V3
 
Linux and H/W optimizations for MySQL
Linux and H/W optimizations for MySQLLinux and H/W optimizations for MySQL
Linux and H/W optimizations for MySQL
 
San presentation nov 2012 central pa
San presentation nov 2012 central paSan presentation nov 2012 central pa
San presentation nov 2012 central pa
 
SSD-Bondi.pptx
SSD-Bondi.pptxSSD-Bondi.pptx
SSD-Bondi.pptx
 
Comparison of foss distributed storage
Comparison of foss distributed storageComparison of foss distributed storage
Comparison of foss distributed storage
 
PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016
PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016
PostgreSQL na EXT4, XFS, BTRFS a ZFS / FOSDEM PgDay 2016
 
What is the average rotational latency of this disk drive What seek.docx
 What is the average rotational latency of this disk drive  What seek.docx What is the average rotational latency of this disk drive  What seek.docx
What is the average rotational latency of this disk drive What seek.docx
 
Ceph Day San Jose - Red Hat Storage Acceleration Utlizing Flash Technology
Ceph Day San Jose - Red Hat Storage Acceleration Utlizing Flash TechnologyCeph Day San Jose - Red Hat Storage Acceleration Utlizing Flash Technology
Ceph Day San Jose - Red Hat Storage Acceleration Utlizing Flash Technology
 

Recently uploaded

Search Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdfSearch Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdfRankYa
 
Human Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsHuman Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsMark Billinghurst
 
Artificial intelligence in cctv survelliance.pptx
Artificial intelligence in cctv survelliance.pptxArtificial intelligence in cctv survelliance.pptx
Artificial intelligence in cctv survelliance.pptxhariprasad279825
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationSlibray Presentation
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupFlorian Wilhelm
 
Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Scott Keck-Warren
 
Advanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionAdvanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionDilum Bandara
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii SoldatenkoFwdays
 
Take control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test SuiteTake control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test SuiteDianaGray10
 
How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.Curtis Poe
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfAddepto
 
Story boards and shot lists for my a level piece
Story boards and shot lists for my a level pieceStory boards and shot lists for my a level piece
Story boards and shot lists for my a level piececharlottematthew16
 
Vertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering TipsVertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering TipsMiki Katsuragi
 
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostLeverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostZilliz
 
Anypoint Exchange: It’s Not Just a Repo!
Anypoint Exchange: It’s Not Just a Repo!Anypoint Exchange: It’s Not Just a Repo!
Anypoint Exchange: It’s Not Just a Repo!Manik S Magar
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsRizwan Syed
 
The Ultimate Guide to Choosing WordPress Pros and Cons
The Ultimate Guide to Choosing WordPress Pros and ConsThe Ultimate Guide to Choosing WordPress Pros and Cons
The Ultimate Guide to Choosing WordPress Pros and ConsPixlogix Infotech
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Mark Simos
 
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek SchlawackFwdays
 
DSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine TuningDSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine TuningLars Bell
 

Recently uploaded (20)

Search Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdfSearch Engine Optimization SEO PDF for 2024.pdf
Search Engine Optimization SEO PDF for 2024.pdf
 
Human Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsHuman Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR Systems
 
Artificial intelligence in cctv survelliance.pptx
Artificial intelligence in cctv survelliance.pptxArtificial intelligence in cctv survelliance.pptx
Artificial intelligence in cctv survelliance.pptx
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck Presentation
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project Setup
 
Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024
 
Advanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionAdvanced Computer Architecture – An Introduction
Advanced Computer Architecture – An Introduction
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko
 
Take control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test SuiteTake control of your SAP testing with UiPath Test Suite
Take control of your SAP testing with UiPath Test Suite
 
How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdf
 
Story boards and shot lists for my a level piece
Story boards and shot lists for my a level pieceStory boards and shot lists for my a level piece
Story boards and shot lists for my a level piece
 
Vertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering TipsVertex AI Gemini Prompt Engineering Tips
Vertex AI Gemini Prompt Engineering Tips
 
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage CostLeverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
Leverage Zilliz Serverless - Up to 50X Saving for Your Vector Storage Cost
 
Anypoint Exchange: It’s Not Just a Repo!
Anypoint Exchange: It’s Not Just a Repo!Anypoint Exchange: It’s Not Just a Repo!
Anypoint Exchange: It’s Not Just a Repo!
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL Certs
 
The Ultimate Guide to Choosing WordPress Pros and Cons
The Ultimate Guide to Choosing WordPress Pros and ConsThe Ultimate Guide to Choosing WordPress Pros and Cons
The Ultimate Guide to Choosing WordPress Pros and Cons
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
 
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
"Subclassing and Composition – A Pythonic Tour of Trade-Offs", Hynek Schlawack
 
DSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine TuningDSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine Tuning
 

Tuning Storage Subsystems for Optimal Database Performance

  • 2. Agenda Performance issues in Storage Hybrid Storage (Disks, SSDs, Memory) ZFS - Not Just Another File System Tuning for databases (General principles) Tuning for MySQL Tuning for PostgreSQL Tuning for Oracle
  • 3. Why? • Some very practical advice based on > recent test results > Improved pgbench results from 70 tps for pure disk to 5003 tps with SSD and tuning > Improved sysbench results from 425 tps to 1811 tps with SSD and tuning for read/write. > Improved sysbench results from 786 tps to 3085 tps with SSD and tuning for read. > collection of tuning knowledge from Sun performance engineers and the community • Some very good resources at the end of the talk for further study
  • 4. Storage Performance Cache Memory 0,0 00 X t ial 10 en fer e dif nc Disk r ma High Cache r fo Performance Disks Pe Large Capacity Disks
  • 5. Latency Comparison Bridging the DRAM to HDD Gap 1S 100mS 10m S 1mS 100uS 10uS TAPE 1u HDD S 100nS FLASH/ SSD 10nS DRAM 1n S CPU
  • 6. Storage Technology Price, Performance & Capacity Capacity Latency Cost/IOPS Cost/GB Technologies IOPs (GB) (microS) ($) ($) Cloud Storage Unlimited 60,000 20 17c/GB 0.15/month Capacity HDDs 2,500 12,000 250 1.67 0.15 Performance HDDs 300 7,000 500 1.52 1.30 SSDs (write) 64 300 5000 0.20 13 SSDs (read only) 64 45 30,000 0.03 13 DRAM 8 0.005 500,000 0.001 52
  • 7. Incorporating Flash Storage Hierarchy
  • 8. Hybrid Storage Flash as Cache Application DRAM Level 1 Cache Write Read Flash Write side Log Flash Level 2 Cache Disk Primary Storage
  • 9. ZFS - Last Word in Filesystem Pooled Storage Vs Traditional Volumes
  • 10. Data Management Unit Smarts Built Right Into the Filesystem
  • 11. Administering ZFS in two slides As easy as pie • zpool commands > create a single disk pool: # zpool create newpool diskname > create a pool with a mirror # zpool create newpool mirror disk1name disk2name > Add device to a pool: # zpool add poolname diskname > Replace a bad disk # zpool replace poolname baddiskname newdiskname > History of commands on the pool: # zpool history poolname > How is my pool performing: # zpool iostat poolname No format command, No fdisk partitions, No volumes
  • 12. Administering ZFS in two slides As easy as pie • zfs commands > create a filesystem: # zfs create poolname/fs-name > set filesystem property: # zfs set quota=size poolname/fs-name # zfs set compression=on poolname/fs-name # zfs set nfsshare=on poolname/fs-name # zfs set recordsize=16k poolname/fs-name > get filesystem property: # zfs get compressratio poolname/fs-name # zfs get all poolname/fs-name > snapshot the filesystem: # zfs snapshot poolname/fs-name@snapshotname No newfs, No mkfs, No /etc/vfstab, No fsck
  • 13. ZFS and Hybrid Storage As easy as pie • Read side > Add ssd as a read side cache > # zpool add poolname cache ssd-device • Write side > Add SSD as a ZFS Intent Log device > # zpool add poolname log ssd-device
  • 14. ZFS Performance Features • Copy-on-write > Turns Random writes to Sequential writes • Dynamic Striping across all devices > Maximize throughput • Multiple Block Sizes > Automatically chosen to match workload • IO Pipelining > Priority/Deadline scheduling, sorting, aggregation • Intelligent prefetch • Compression - Improves performance & Capacity • Can safely use write cache on disks
  • 15. Databases Not Just Another Application • Most Databases do their own buffering > Filesystem caching can get in the way > “double buffer” problem • Most Databases do “prefetch” > Filesystems prefetch can cause extra IO > “directio” gets filesystem out of the way • Have their own “log” mechanism. > Interesting interaction with a transaction based filesystem • Multiple blocks sizes > Database & Transaction log, block sizes are normally different
  • 16. Tuning ZFS for Databases Tuning is Evil - Long live Tuning • In general tuning is evil. Let ZFS do it for you. • A few fine tuning tips for databases > Get to the latest update of OS > Set the recordsize to match database block size > Separate Transaction logs and data onto separate zpools > [Note: This will be addressed with the ZIL bypass property fix] > Reduce the impact of double buffering by changing the caching method to “metadata only” > Use separate ZIL (ZFS Intent Log) preferably SSD > Use SSD as secondary cache - L2ARC (Level 2. Adaptive Replacement Cache)
  • 17. ZFS tuning for MySQL • Many tuning depends on storage engine • For Innodb > Prefer to cache in Innodb rather than ARC zfs set primarycache=metadata poolname/database > Set recordsize to 16k for data and 128k for log zfs set recordsize=16k poolname/database (Note: do this before you load any data) > Turn off prefetch set zfs:zfs_prefetch_disable = 1 (in /etc/system) (File level prefetch not triggered if you change record size to 16k) > Use raid0 or mirror over raidz raidz is no suitable for random IO > Add SSDs for either read side or write side based on workload zpool add datapool cache ssd-disk zpool create logpool ssd-disk3 In my.cnf set innodb_data_home_dir & innodb_log_group_home_dir
  • 18. ZFS tuning for MySQL • More tuning for Innodb > Some device vendors flush cache even when not needed. (eg. battery backed cache) set zfs:zfs_nocacheflush = 1 > Turn on compression zfs set compression=on poolname/database ZFS does not turn on compression if less than 12.5% saving. IO reduction may offset the extra cpu cost > Disable double writes innodb_doublewrite=0 (in my.cnf) ZFS does not allow any partial writes so no need to guard against it.
  • 19. ZFS tuning for PostgreSQL • Postgres tuning hints > Set recordsize to 8k zfs set recordsize=8k poolname/database > Turn down ARC cache. set zfs:zfs_arc_max in /etc/system > Add SSDs for either read side or write side based on workload zpool add poolname cache ssd-name zpool add poolname log ssd-name > Use separate pool for log (preferably one with SSD) & data initdb -X log_directory_name create tablespace datatbs location 'database_directory_name' create database mydb with  tablespace datatbs > Don’t forget to basic Postgres tuning on Solaris - (huge gains) Set shared_buffers, temp_buffers, work_mem, maintenance_work_mem, wal_sync_method, synchronous_commits etc see: http://blogs.sun.com/jkshah/entry/best_practices_with_postgresql_8
  • 20. ZFS tuning for Oracle • Oracle tuning hints > Set recordsize to match db_block_size (default 8k) zfs set recordsize=8k poolname/database > Use separate pool for Oracle logs make sure record size of the log filesystem is left to the 128k default > Add SSDs for either read side or write side based on workload zpool add poolname cache ssd-name zpool add poolname log ssd-name
  • 21. Benchmark results • Hardware > Sun x4150 2 x Quad core 2.3 GHz Xeon 12 GB ram 3 x 10000 rpm drives 3 x 32 GB SSDs • Software > OpenSolaris 2009.06 > Postgres 8.3.7 > MySQL 5.4 beta
  • 22. Benchmark results • pgbench & Postgres > command line: pgbench -c 10 -s 10 -t 10000 pgbench Description TPS Single disk ZFS 72 tps 2 Raid 0 disk + SSD as level 2 cache 241 tps Above + general postgres optimization 2026 tps + all the data on SSD 2603 tps + data on hdd & log on SSD 4372 tps + primarycache=metadata 5003 tps
  • 23. Benchmark results • sysbench & mysql 5.4 > read/write test: sysbench --max-time=300 --max-requests=0 --test=oltp -- oltp-dist-type=special --oltp-table-size=10000000 --num-threads=20 run Description TPS Single disk ZFS 425 tps raid0 ZFS 670 tps + SSD cache 788 tps + Separate intent log 1352 tps + With optimization 1809 tps
  • 24. Benchmark results • sysbench & mysql 5.4 > read test: sysbench --max-time=300 --max-requests=0 --test=oltp --oltp- dist-type=special --oltp-table-size=10000000 --num-threads=20 --oltp-read- only=on run Description TPS Single disk ZFS 786 tps 2 disk raid0 ZFS 1501 tps + SSD cache 1981 tps + Separate intent log on SSD 2567 tps + optimization 3065 tps
  • 25. !"#$!%&'()*$'&(+,(-$.//012.$ Sun Unified Storage *#%'345*6*5$!%(#+(5&#* :$;<$=*='&-$>$?@A9BB;<$!6!$1)CDC 78)74+*#!8%3$!9(5(:5* E($+'$F@;<$=*='&-G$142A?H<$!6H6$1)CDC E($+'$IA?:$;< write-optimized SSDs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
  • 26. Getting these systems at a discount Sun Startup Essentials • Exclusive program for startups • Eligibility <6 yrs. Old, <150 sun.com/startup employees • Co-marketing opportunities • Funding assistance • Deeply discounted storage and servers certified for Linux, Windows, and Solaris • Hosting starting at $40 • Open source software, and discounted MySQL • Free email based tech support • Free and discounted training on Sun technologies • Member-only webinars
  • 27. Resources • ZFS info: http://www.opensolaris.org/os/community/zfs/ • ZFS Best Practices Guide: http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide • ZFS Evil Tuning Guide: http://www.solarisinternals.com/wiki/index.php/ZFS_Evil_Tuning_Guide • Blogs of note: > All things performance tuning: http://blogs.sun.com/realneel http://blogs.sun.com/roch > Postgres tuning - Jignesh’s Blog http://blogs.sun.com/jkshah > Angelo’s blog http://blogs.sun.com/angelo
  • 28. Tuning Storage Subsystem for Databases Angelo Rajadurai angelo@sun.com http://blogs.sun.com/angelo twitter: rajadurai