Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

IO Schedulers (Elevater) concept and its affection on database performance

•
•
Alireza Kamrani
Alireza Kamrani

I/O schedulers in Linux reorder and group I/O requests to improve throughput while balancing latency. Different schedulers take different approaches, and there is no single best scheduler for all situations. For Oracle databases on Linux, Oracle recommends using the Deadline scheduler for HDD storage to prioritize I/O requests, while the none scheduler may be best for SSD/NVMe storage. When selecting a scheduler, it is important to consider the storage media and I/O characteristics of the workload.

1 of 5
I/O Schedulers (Elevater) concept and its a ff ection on database performance Linux I/O schedulers in Linux I/O schedulers attempt to improve throughput by reordering request access into a linear order based on the logical addresses of the data and trying to group these together. While this may increase overall throughput it may lead to some I/O requests waiting for too long, causing latency issues. I/O schedulers attempt to balance the need for high throughput while trying to fairly share I/O requests amongst processes. Di ff erent approaches have been taken for various I/O schedulers and each has their own set of strengths and weaknesses and the general rule is that there is no perfect default I/O scheduler for all the range of I/O demands a system may experience. I/O Elevater or Scheduler Types in Debian distribution as Ubuntu Multiqueue I/O schedulers Note: These are the only I/O schedulers available in Ubuntu Eoan Ermine 19.10 and onwards. The following I/O schedulers are designed for multiqueue devices. These map I/O requests to multiple queues and these are handled by kernel threads that are distributed across multiple CPUs. bfq (Budget Fair Queuing) (Multiqueue) Designed to provide good interactive response, especially for slower I/O devices. This is a complex I/O scheduler and has a relatively high per-operation overhead so it is not ideal for devices with slow CPUs or high throughput I/O devices. Fair sharing is based on the number of sectors requested and heuristics rather than a time slice. Desktop users may like to experiment with this I/O scheduler as it can be advantageous when loading large applications. kyber (Multiqueue) Designed for fast multi-queue devices and is relatively simple. Has two request queues: • Synchronous requests (e.g. blocked reads) • Asynchronous requests (e.g. writes) There are strict limits on the number of request operations sent to the queues. In theory this limits the time waiting for requests to be dispatched, and hence should provide quick completion time for requests that are high priority. none (Multiqueue) The multi-queue no-op I/O scheduler. Does no reordering of requests, minimal overhead. Ideal for fast random I/O devices such as NVME. mq-deadline (Multiqueue) This is an adaption of the deadline I/O scheduler but designed for Multiqueue devices. A good all- rounder with fairly low CPU overhead.
Non-multiqueue I/O schedulers NOTE: Non-multiqueue have been deprecated in Ubuntu Eoan Ermine 19.10 onwards as they are no longer supported in the Linux 5.3 kernel. deadline This fi xes starvation issues seen in other schedulers. It uses 3 queues for I/O requests: • Sorted • Read FIFO - read requests stored chronologically • Write FIFO - write requests stored chronologically Requests are issued from the sorted queue inless a read from the head of a read or write FIFO expires. Read requests are preferred over write requests. Read requests have a 500ms expiration time, write requests have a 5s expiration time. cfq (Completely Fair Queueing) • Per-process sorted queues for synchronous I/O requests. • Fewer queues for asynchronous I/O requests. • Priorities from ionice are taken into account. Each queue is allocated a time slice for fair queuing. There may be wasteful idle time if a time slice quantum has not expired. noop (No-operation) Performs merging of I/O requests but no sorting. Good for random access devices ( fl ash, ramdisk, etc) and for devices that sort I/O requests such as advanced storage controllers. Selecting I/O Schedulers Prior to Ubuntu 19.04 with Linux 5.0 or Ubuntu 18.04.3 with Linux 4.15, the multiqueue I/O scheduling was not enabled by default and just the deadline, cfq and noop I/O schedulers were available by default. For Ubuntu 19.10 with Linux 5.0 or Ubuntu 18.04.3 with Linux 5.0 onwards, multiqueue is enabled by default providing the bfq, kyber, mq-deadline and none I/O schedulers. For Ubuntu 19.10 with Linux 5.3 the deadline, cfq and noop I/O schedulers are deprecated. Disk I/O Scheduler on Linux Redhat distribution as Oracle Linux The following are the common starting recommendations for an I/O scheduler for a RHEL based virtual guest based upon kernel version and disk type. • RHEL 8,9 : mq-deadline is default I/O scheduler unless otherwise changed [FN.1] ◦ Virtual disks: keep current io scheduler setting (mq-deadline) ◦ Physical disks: keep current io scheduler setting (mq-deadline, or for NVMe none) • RHEL 7.5+ : deadline is default io scheduler unless otherwise changed ◦ Virtual disks: keep current io scheduler setting (deadline) ✦ 6.2. I/O Scheduling with Red Hat Enterprise Linux as a Virtualization Guest ◦ Physical disks: keep current io scheduler setting (deadline) • RHEL 4,5,6,(7.0-7.4) : cfq is default I/O scheduler unless otherwise changed ◦ Virtual disks: change to noop scheduler [FN.2] ✦ 6.3. I/O Scheduling with Red Hat Enterprise Linux as a Virtualization Guest ◦ Physical disks: keep current io scheduler setting
Red Hat Enterprise Linux: Disk I/O Scheduler and Orcale Linux There are several choices for the type of disk scheduler used by the operating system for Red Hat Enterprise Linux 8. The function of the disk scheduler is for ordering, delaying, or merging I/O requests to storage to achieve better throughput and latency. Each disk scheduler provides a di ff erent method for managing storage requests: None: Implements a fi rst-in fi rst-out (FIFO) scheduling algorithm. The none scheduler is recommended for systems that have high performance storage like Solid State Drives (SSD) or Non-volatile Memory Express (NVMe) drives. The default schedule for NVME is none. Mq-deadline: This scheduler groups queued I/O requests into batches. The I/O in the batches are organized by logical block addressing so that writes are organized by location to storage. The mq- deadline is suited for traditional HDD storage. Bfq: This scheduler prioritizes latency rather than maximum throughput. The bfq disk scheduler is recommended for desktop or interactive tasks and traditional HDD storage. Kyber: This scheduler tunes itself by analyzing I/O requests. With each I/O request a calculation is done to determine of the I/O can be satis fi ed with the least amount of latency. Kyber is recommended for high performance storage like SSDs and NVMe drives. In this best practice, we followed Red Hat’s recommendation of using the “none” disk I/O scheduler as the PowerStore contains NVMe drives and the database access the storage volumes as virtual devices through virtual guest OS and through host bus adapters (HBAs). Best I/O Scheduler for Oracle Database? For best performance for Oracle ASM, Oracle recommends that you use the Deadline I/O Scheduler when using Baremetal servers with SAS or Sata HDD on storage or local disk. When using Nvme or enterprise SSD disks may be good options be none. Note: there are many models of SSD disks in market with di ff erent characteristics and price. Disk I/O schedulers reorder, delay, or merge requests for disk I/O to achieve better throughput and lower latency.
Orcale Linux has multiple disk I/O schedulers available, including Deadline, Noop, Anticipatory, and Completely Fair Queuing (CFQ). On each cluster node, enter the following command to verify that the Deadline disk I/O scheduler is con fi gured for use: # cat /sys/block/${ASM_DISK}/queue/scheduler noop [deadline] cfq Note: On newer kernel 'deadline' chnages to 'mq-deadline'. In this example, the default disk I/O scheduler is Deadline and ASM_DISK is the Oracle Automatic Storage Management (Oracle ASM) disk device. On some virtual environments (VM) and special devices such as fast storage devices, the output of the above command may be none. The operating system or VM bypasses the kernel I/O scheduling and submits all I/O requests directly to the device. Do not change the I/O Scheduler settings on such environments. If the default disk I/O scheduler is not Deadline, then set it using a rules fi le: Note:You can modify balue of scheduler fi le manually by echo command but following role is better to maintain. 1. Using a text editor, create a UDEV rules fi le for the Oracle ASM devices: # vi /etc/udev/rules.d/60-oracle-schedulers.rules 2. Add the following line to the rules fi le and save it: ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/ scheduler}="deadline" 3. On clustered systems, copy the rules fi le to all other nodes on the cluster. For example: $ scp 60-oracle-schedulers.rules root@node2:/etc/udev/rules.d/ 4. Load the rules fi le and restart the UDEV service. For example: ◦ Oracle Linux and Red Hat Enterprise Linux # udevadm control --reload-rules ◦ SUSE Linux Enterprise Server # /etc/init.d boot.udev restart 1. Verify that the disk I/O scheduler is set as Deadline. When use ASMLib: Get the details of the disk: # /etc/init.d/oracleasm querydisk -p DISK1
Disk "DISK1" is a valid ASM disk /dev/sdb1: LABEL="DISK1" TYPE="oracleasm" Check the deadline scheduler is con fi gured at the disk level: # more /sys/block/sdb/queue/scheduler noop [deadline] cfq <<<< Here in this case deadline is con fi gured. Likewise you check for the disks of the ASM. It is con fi rmed that the parent disks which are part of ASM diskgroup are using the "deadline" I/O scheduler. This can be con fi rmed by running latest orachk utility. Which Linux scheduler is recommended on Linux, 'CFQ Scheduler' or 'DEADLINE Scheduler'? The CFQ scheduler is suitable for a wide variety of applications and provides a good compromise between throughput and latency, hence it is the default. In comparison to the CFQ algorithm, the Deadline scheduler caps per request and maintains a good disk throughput which is best for disk-intensive database applications, so if you application is IO intensive, then you would want to use the Deadline scheduler, but for all other applications, the Completely Fair Queuing (CFQ) scheduler is better. On newer kernel 'deadline' chnages to 'mq-deadline'. Earlier CFQ scheduler is exposed to an issue (5041764 / bugzilla#151368) if OCFS2 is used, the DEADLINE scheduler did not encounter the problem. Regards Alireza Kamrani Senior RDBMS Consultant

More Related Content

IO Schedulers (Elevater) concept and its affection on database performance

  • 1. I/O Schedulers (Elevater) concept and its a ff ection on database performance Linux I/O schedulers in Linux I/O schedulers attempt to improve throughput by reordering request access into a linear order based on the logical addresses of the data and trying to group these together. While this may increase overall throughput it may lead to some I/O requests waiting for too long, causing latency issues. I/O schedulers attempt to balance the need for high throughput while trying to fairly share I/O requests amongst processes. Di ff erent approaches have been taken for various I/O schedulers and each has their own set of strengths and weaknesses and the general rule is that there is no perfect default I/O scheduler for all the range of I/O demands a system may experience. I/O Elevater or Scheduler Types in Debian distribution as Ubuntu Multiqueue I/O schedulers Note: These are the only I/O schedulers available in Ubuntu Eoan Ermine 19.10 and onwards. The following I/O schedulers are designed for multiqueue devices. These map I/O requests to multiple queues and these are handled by kernel threads that are distributed across multiple CPUs. bfq (Budget Fair Queuing) (Multiqueue) Designed to provide good interactive response, especially for slower I/O devices. This is a complex I/O scheduler and has a relatively high per-operation overhead so it is not ideal for devices with slow CPUs or high throughput I/O devices. Fair sharing is based on the number of sectors requested and heuristics rather than a time slice. Desktop users may like to experiment with this I/O scheduler as it can be advantageous when loading large applications. kyber (Multiqueue) Designed for fast multi-queue devices and is relatively simple. Has two request queues: • Synchronous requests (e.g. blocked reads) • Asynchronous requests (e.g. writes) There are strict limits on the number of request operations sent to the queues. In theory this limits the time waiting for requests to be dispatched, and hence should provide quick completion time for requests that are high priority. none (Multiqueue) The multi-queue no-op I/O scheduler. Does no reordering of requests, minimal overhead. Ideal for fast random I/O devices such as NVME. mq-deadline (Multiqueue) This is an adaption of the deadline I/O scheduler but designed for Multiqueue devices. A good all- rounder with fairly low CPU overhead.
  • 2. Non-multiqueue I/O schedulers NOTE: Non-multiqueue have been deprecated in Ubuntu Eoan Ermine 19.10 onwards as they are no longer supported in the Linux 5.3 kernel. deadline This fi xes starvation issues seen in other schedulers. It uses 3 queues for I/O requests: • Sorted • Read FIFO - read requests stored chronologically • Write FIFO - write requests stored chronologically Requests are issued from the sorted queue inless a read from the head of a read or write FIFO expires. Read requests are preferred over write requests. Read requests have a 500ms expiration time, write requests have a 5s expiration time. cfq (Completely Fair Queueing) • Per-process sorted queues for synchronous I/O requests. • Fewer queues for asynchronous I/O requests. • Priorities from ionice are taken into account. Each queue is allocated a time slice for fair queuing. There may be wasteful idle time if a time slice quantum has not expired. noop (No-operation) Performs merging of I/O requests but no sorting. Good for random access devices ( fl ash, ramdisk, etc) and for devices that sort I/O requests such as advanced storage controllers. Selecting I/O Schedulers Prior to Ubuntu 19.04 with Linux 5.0 or Ubuntu 18.04.3 with Linux 4.15, the multiqueue I/O scheduling was not enabled by default and just the deadline, cfq and noop I/O schedulers were available by default. For Ubuntu 19.10 with Linux 5.0 or Ubuntu 18.04.3 with Linux 5.0 onwards, multiqueue is enabled by default providing the bfq, kyber, mq-deadline and none I/O schedulers. For Ubuntu 19.10 with Linux 5.3 the deadline, cfq and noop I/O schedulers are deprecated. Disk I/O Scheduler on Linux Redhat distribution as Oracle Linux The following are the common starting recommendations for an I/O scheduler for a RHEL based virtual guest based upon kernel version and disk type. • RHEL 8,9 : mq-deadline is default I/O scheduler unless otherwise changed [FN.1] ◦ Virtual disks: keep current io scheduler setting (mq-deadline) ◦ Physical disks: keep current io scheduler setting (mq-deadline, or for NVMe none) • RHEL 7.5+ : deadline is default io scheduler unless otherwise changed ◦ Virtual disks: keep current io scheduler setting (deadline) ✦ 6.2. I/O Scheduling with Red Hat Enterprise Linux as a Virtualization Guest ◦ Physical disks: keep current io scheduler setting (deadline) • RHEL 4,5,6,(7.0-7.4) : cfq is default I/O scheduler unless otherwise changed ◦ Virtual disks: change to noop scheduler [FN.2] ✦ 6.3. I/O Scheduling with Red Hat Enterprise Linux as a Virtualization Guest ◦ Physical disks: keep current io scheduler setting
  • 3. Red Hat Enterprise Linux: Disk I/O Scheduler and Orcale Linux There are several choices for the type of disk scheduler used by the operating system for Red Hat Enterprise Linux 8. The function of the disk scheduler is for ordering, delaying, or merging I/O requests to storage to achieve better throughput and latency. Each disk scheduler provides a di ff erent method for managing storage requests: None: Implements a fi rst-in fi rst-out (FIFO) scheduling algorithm. The none scheduler is recommended for systems that have high performance storage like Solid State Drives (SSD) or Non-volatile Memory Express (NVMe) drives. The default schedule for NVME is none. Mq-deadline: This scheduler groups queued I/O requests into batches. The I/O in the batches are organized by logical block addressing so that writes are organized by location to storage. The mq- deadline is suited for traditional HDD storage. Bfq: This scheduler prioritizes latency rather than maximum throughput. The bfq disk scheduler is recommended for desktop or interactive tasks and traditional HDD storage. Kyber: This scheduler tunes itself by analyzing I/O requests. With each I/O request a calculation is done to determine of the I/O can be satis fi ed with the least amount of latency. Kyber is recommended for high performance storage like SSDs and NVMe drives. In this best practice, we followed Red Hat’s recommendation of using the “none” disk I/O scheduler as the PowerStore contains NVMe drives and the database access the storage volumes as virtual devices through virtual guest OS and through host bus adapters (HBAs). Best I/O Scheduler for Oracle Database? For best performance for Oracle ASM, Oracle recommends that you use the Deadline I/O Scheduler when using Baremetal servers with SAS or Sata HDD on storage or local disk. When using Nvme or enterprise SSD disks may be good options be none. Note: there are many models of SSD disks in market with di ff erent characteristics and price. Disk I/O schedulers reorder, delay, or merge requests for disk I/O to achieve better throughput and lower latency.
  • 4. Orcale Linux has multiple disk I/O schedulers available, including Deadline, Noop, Anticipatory, and Completely Fair Queuing (CFQ). On each cluster node, enter the following command to verify that the Deadline disk I/O scheduler is con fi gured for use: # cat /sys/block/${ASM_DISK}/queue/scheduler noop [deadline] cfq Note: On newer kernel 'deadline' chnages to 'mq-deadline'. In this example, the default disk I/O scheduler is Deadline and ASM_DISK is the Oracle Automatic Storage Management (Oracle ASM) disk device. On some virtual environments (VM) and special devices such as fast storage devices, the output of the above command may be none. The operating system or VM bypasses the kernel I/O scheduling and submits all I/O requests directly to the device. Do not change the I/O Scheduler settings on such environments. If the default disk I/O scheduler is not Deadline, then set it using a rules fi le: Note:You can modify balue of scheduler fi le manually by echo command but following role is better to maintain. 1. Using a text editor, create a UDEV rules fi le for the Oracle ASM devices: # vi /etc/udev/rules.d/60-oracle-schedulers.rules 2. Add the following line to the rules fi le and save it: ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/ scheduler}="deadline" 3. On clustered systems, copy the rules fi le to all other nodes on the cluster. For example: $ scp 60-oracle-schedulers.rules root@node2:/etc/udev/rules.d/ 4. Load the rules fi le and restart the UDEV service. For example: ◦ Oracle Linux and Red Hat Enterprise Linux # udevadm control --reload-rules ◦ SUSE Linux Enterprise Server # /etc/init.d boot.udev restart 1. Verify that the disk I/O scheduler is set as Deadline. When use ASMLib: Get the details of the disk: # /etc/init.d/oracleasm querydisk -p DISK1
  • 5. Disk "DISK1" is a valid ASM disk /dev/sdb1: LABEL="DISK1" TYPE="oracleasm" Check the deadline scheduler is con fi gured at the disk level: # more /sys/block/sdb/queue/scheduler noop [deadline] cfq <<<< Here in this case deadline is con fi gured. Likewise you check for the disks of the ASM. It is con fi rmed that the parent disks which are part of ASM diskgroup are using the "deadline" I/O scheduler. This can be con fi rmed by running latest orachk utility. Which Linux scheduler is recommended on Linux, 'CFQ Scheduler' or 'DEADLINE Scheduler'? The CFQ scheduler is suitable for a wide variety of applications and provides a good compromise between throughput and latency, hence it is the default. In comparison to the CFQ algorithm, the Deadline scheduler caps per request and maintains a good disk throughput which is best for disk-intensive database applications, so if you application is IO intensive, then you would want to use the Deadline scheduler, but for all other applications, the Completely Fair Queuing (CFQ) scheduler is better. On newer kernel 'deadline' chnages to 'mq-deadline'. Earlier CFQ scheduler is exposed to an issue (5041764 / bugzilla#151368) if OCFS2 is used, the DEADLINE scheduler did not encounter the problem. Regards Alireza Kamrani Senior RDBMS Consultant