About this series
This three-part series (see Resources) on the AIX® disk and I/O subsystem focuses on the challenges of optimizing disk I/O performance. While disk tuning is arguably less exciting than CPU or memory tuning, it is a crucial component in optimizing server performance. In fact, partly because disk I/O is your weakest subsystem link, there is more you can do to improve disk I/O performance than any other subsystem.
Unlike the tuning of other subsystems, tuning disk I/O should actually be started during
the architectural phase of building your systems. While there are virtual memory
equivalents of I/O tuning parameters (
lvmo), the best
way to increase disk I/O performance is by properly configuring your systems and not
tuning parameters. Unlike virtual memory tuning, it is much more complex to change the
way in which you structure your logical volumes after they have been created and running,
so you usually get only one chance to do this right. This article discusses the ways that
you can configure your logical volumes and where to actually place them with respect to
the physical disk, and it also addresses the tools used to monitor your logical volumes.
Most of these tools are not meant to be used for long-term trending and are specific AIX
tools that provide information as to how the logical volumes are configured and if they
have been optimized for your environment.
Part 1 (see Resources) of this series introduced
iostat, but it did not address using the tool outside of viewing
Asynchronous I/O servers. Part 2 uses iostat to monitor your disks and shows you what it
can do to help quickly determine your I/O bottleneck. While
iostat is one of
those generic UNIX® utilities that was not developed specifically for AIX, it is
actually very useful for quickly determining what is going on in your system. The more
specific AIX logical volume commands help drill down deeper into your logical volumes to
help you really analyze what your problems are, if any. It's important that you clearly
understand what you're looking for before using these tools. This article describes the
tools and also shows you how to analyze their output, which helps in analyzing your disk
Logical volume and disk placement overview
This section defines the Logical Volume Manager (LVM) and introduces some of its features. Let's drill down into logical volume concepts, examine how they relate to improving disk I/O utilization, and talk about logical volume placement as it relates to the physical disk, by defining and discussing both intra-policy and inter-policy disk practices.
Conceptually, the logical volume layer sits between the application and physical layers. In the context of disk I/O, the application layers are the file system or raw logical volumes. The physical layer consists of the actual disk. LVM is an AIX disk management system that maps the data between logical and physical storage. This allows data to reside on multiple physical platters and to be managed and analyzed using specialized LVM commands. LVM actually controls all the physical disk resources on your system and helps provide a logical view of your storage subsystem. Understanding that it sits between the application layer and the physical layer should help you understand why it is arguably the most important of all the layers. Even your physical volumes themselves are part of the logical layer, as the physical layer only encompasses the actual disks, device drivers, and any arrays that you might have already configured. Figure 1 illustrates the concepts and shows how tightly integrated the logical I/O components relate to the physical disk and its application layer.
Figure 1. Logical volume diagram
Let's now quickly introduce the elements that are part of LVM, from the bottom up. Each of the drives is named as a physical volume. Multiple physical volumes make up a volume group. Within the volume groups, logical volumes are defined. The LVM enables the data to be on multiple physical drives, though they might be configured to be on a single volume group. These logical volumes can be either one or multiple logical partitions. Each of the logical partitions has a physical partition that correlates to it. Here is where you can have multiple copies of the physical portions for purposes such as disk mirroring.
Let's take a quick look at how logical volume creation correlates with physical volumes. Figure 2 illustrates the actual storage position on the physical disk platter.
Figure 2. Actual storage position on the physical disk platter
As a general rule, data that is written toward its center has faster seek times than data written on the outer edge. This has to do with the density of data. Because it is more dense as it moves toward its center, there is actually less movement of the head. The inner edge usually has the slowest seek times. As a best practice, the more intensive I/O applications should be brought closer to the center of the physical volumes. Note that there are exceptions to this. Disks hold more data per track on the edges of the disk, not on the center. That being said, logical volumes being accessed sequentially should actually be placed on the edge for better performance. The same holds true for logical volumes that have Mirror Write Consistency Check (MWCC) turned on. This is because the MWCC sector is on the edge of the disk and not at the center of it, which relates to the intra-disk policy of logical volumes.
Let's discuss another important concept referred to as the inter-disk policy of logical volumes. The inter-disk policy defines the number of disks on which the physical partitions of a logical volume actually resides. The general rule is that the minimum policy provides the greatest reliably and availability, and the maximum policy improves performance. Simply put, the more drives that data is spread on, the better the performance. Some other best practices include: allocating intensive logical volumes to separate physical volumes, defining the logical volumes to the maximum size you need, and placing logical volumes that are frequently used close together. This is why it is so important to know your data prior to configuring your systems so that you can create policies that make sense from the start.
You can define your polices when creating the logical volumes themselves using the
following command or smit fastpath:
# mklv or
# smitty mklv.
Monitoring logical volumes and analyzing results
This section provides instructions on how to monitor your logical volumes and analyze the results. Various commands are introduced along with the purposes for which they are used, and you examine the output as well.
A ticket has just been opened up with the service desk that relates to slow performance on
some database server. You suspect that there might be an I/O issue, so you start with
iostat. If you recall, this command was introduced in the first installment
of the series (see Resources), though only for the purposes of
viewing asynchronous I/O servers. Now, let's look at
iostat in more detail.
iostat, the equivalent of using
vmstat for virtual memory, is
arguably the most effective way to get a first glance of what is happening with your I/O
Listing 1. Using iostat
# iostat 1 System configuration: lcpu=4 disk=4 tty: tin tout avg-cpu: % user % sys % idle % iowait 0.0 392.0 5.2 5.5 88.3 1.1 Disks: % tm_act Kbps tps Kb_read Kb_wrtn hdisk1 0.5 19.5 1.4 53437739 21482563 hdisk0 0.7 29.7 3.0 93086751 21482563 hdisk4 1.7 278.2 6.2 238584732 832883320 hdisk3 2.1 294.3 8.0 300653060 832883320
What are you seeing here and what does this all mean?
- % tm_act: Reports back the percentage of time that the physical disk was active or the total time of disk requests.
- Kbps: Reports back the amount of data transferred to the drive in kilobytes.
- tps: Reports back the number of transfers per second issued to the physical disk.
- Kb_read: Reports back the total data (kilobytes) from your measured interval that is read from the physical volumes.
- Kb_wrtn: Reports back the amount of data (kilobytes) from your measured interval that is written to the physical volumes.
You need to watch % tm_act very carefully, because when its utilization exceeds roughly 60 to 70 percent, it usually is indicative that processes are starting to wait for I/O. This might be your first clue of impending I/O problems. Moving data to less busy drives can obviously help ease this burden. Generally speaking, the more drives that your data hits, the better. Just like anything else, too much of a good thing can also be bad, as you have to make sure you don't have too many drives hitting any one adapter. One way to determine if an adapter is saturated is to sum the Kbps amounts for all disks attached to one adapter. The total should be below the disk adapter throughput rating, usually less than 70 percent.
-a flag (see Listing 2) helps you drill down
further to examine adapter utilization.
Listing 2. Using iostat with the
# iostat -a Adapter: Kbps tps Kb_read Kb_wrtn scsi0 0.0 0.0 0 0 Paths/Disk: % tm_act Kbps tps Kb_read Kb_wrtn hdisk1_Path0 37.0 89.0 0.0 0 0 hdisk0_Path0 67.0 47.0 0.0 0 0 hdisk4_Path0 0.0 0.0 0.0 0 0 hdisk3_Path0 0.0 0.0 0.0 0 0 Adapter: Kbps tps Kb_read Kb_wrtn ide0 0.0 0.0 0 0 Paths/Disk: % tm_act Kbps tps Kb_read Kb_wrtn cd0 0.0 0.0 0.0 0 0
Clearly, there are no bottlenecks here. Using the
-d flag allows you to drill
down to one specific disk (see Listing 3).
Listing 3. Using iostat with the
# iostat -d hdisk1 1 System configuration: lcpu=4 disk=5 Disks: % tm_act Kbps tps Kb_read Kb_wrtn hdisk1 0.5 19.4 1.4 53437743 21490480 hdisk1 5.0 78.0 23.6 3633 3564 hdisk1 0.0 0.0 0.0 0 0 hdisk1 0.0 0.0 0.0 0 0 hdisk1 0.0 0.0 0.0 0 0 hdisk1 0.0 0.0 0.0 0 0
Let's look at some specific AIX LVM commands. You examined disk placement earlier and the importance of architecting your systems correctly from the beginning. Unfortunately, you don't always have that option. As system administrators, you sometimes inherit systems that must be fixed. Let's look at the layout of the logical volumes on disks to determine if you need to change definitions or re-arrange your data.
Let's look first at a volume group and find the logical volumes that are a part of it.
lsvg is the command that provides volume group information (see Listing 4).
Listing 4. Using lsvg
# lsvg -l data2vg Data2vg: LV NAME TYPE LPs PPs PVs LV STATE MOUNT POINT data2lv jfs 128 256 2 open/syncd /data2 loglv00 jfslog 1 2 2 open/syncd N/A appdatalv jfs 128 256 2 open/syncd /appdata
Now, let's use
lslv, which provides for specific data on logical volumes (see
Listing 5. Using lslv
# lslv data2lv LOGICAL VOLUME: data2lv VOLUME GROUP: data2vg LV IDENTIFIER: 0003a0ec00004c00000000fb076f3f41.1 PERMISSION: read/write VG STATE: active/complete LV STATE: opened/syncd TYPE: jfs WRITE VERIFY: off MAX LPs: 512 PP SIZE: 64 megabyte(s) COPIES: 2 SCHED POLICY: parallel LPs: 128 PPs: 256 STALE PPs: 0 BB POLICY: relocatable INTER-POLICY: minimum RELOCATABLE: yes INTRA-POLICY: center UPPER BOUND: 32 MOUNT POINT: /data LABEL: /data MIRROR WRITE CONSISTENCY: on/ACTIVE EACH LP COPY ON A SEPARATE PV ?: yes Serialize IO ?: NO
This view provides a detailed description of your logical volume attributes. What do you have here? The intra-policy is at the center, which is normally the best policy to have for I/O-intensive logical volumes. As you recall from an earlier discussion, there are exceptions to this rule. Unfortunately, you've just hit one of them. Because Mirror Write Consistency (MWC) is on, the volume would have been better served if it were placed on the edge. Let's look at its inter-policy. The inter-policy is minimum, which is usually the best policy to have if availability is more important then performance. Further, there are double the number of physical partitions than logical partitions, which signify that you are mirroring your systems. In this case, you were told that raw performance was the most important objective, so the logical volume was not configured in such a way as to the reality of how the volume is being utilized. Further, if you are mirroring your system and using an external storage array, this would even be worse, as you're already providing mirroring at the hardware layer, which is actually more effective then using AIX mirroring.
Let's drill down even further in Listing 6.
Listing 6. lslv with the
# lslv -l data2lv data2lv:/data2 PV COPIES IN BAND DISTRIBUTION hdisk2 128:000:000 100% 000:108:020:000:000 hdisk3 128:000:000 100% 000:108:020:000:000
-l flag of
lslv lists all the physical volumes associated
with the logical volumes and distribution for each logical volume. You can then determine
that 100 percent of the physical partitions on the disk are allocated to this logical
volume. The distribution sections show the actual number of physical partitions within
each physical volume. From here, you can detail its intra-disk policy. The order of these
fields are as follows:
The reports show that most of the data is in the middle and some at the center.
Let's keep going and find out which logical volumes are associated with the one physical
volume. This is done with the
lspv command (see Listing
Listing 7. Using the lspv command
# lspv -l hdisk2 hdisk2: LV NAME LPs PPs DISTRIBUTION MOUNT POINT loglv01 1 1 01..00..00..00..00 N/A data2lv 128 128 00..108..20..00..00 /data2 appdatalv 128 128 00..00..88..40..00 /appdata
Now you can actually identify which of the logical volumes on this disk are geared up for maximum performance.
You can drill down even further to get more specific (see Listing 8).
Listing 8. lspv with the
# lspv -p hdisk2 hdisk2: PP RANGE STATE REGION LV ID TYPE MOUNT POINT 1-108 free outer edge 109-109 used outer edge loglv00 jfslog N/A 110-217 used outer middle data2lv jfs /data2 218-237 used center appdatalv jfs /appdata 238-325 used center testdatalv jfs /testdata 326-365 used inner middle stagingdatalv jfs /staging 366-433 free inner middle 434-542 free inner edge
This view tells you what is free on the physical volume, what has been used, and which partitions are used where. This is a nice view.
One of the best tools to look at LVM usage is with
lvmstat (see Listing 9).
Listing 9. Using lvmstat
# lvmstat -v data2vg 0516-1309 lvmstat: Statistics collection is not enabled for this logical device. Use -e option to enable.
As you can see by the output here, it is not enabled (by default), so you need to actually
enable it prior to running the tool using
# lvmstat -v data2vg -e. The
following command takes a snapshot of LVM information every second for 10 intervals:
# lvmstat -v data2vg 1 10
This view shows the most utilized logical volumes on your system since you started the data collection tool. This is very helpful when drilling down to the logical volume layer when tuning your systems (see Listing 10).
Listing 10. lvmstat with the
# lvmstat -v data2vg Logical Volume iocnt Kb_read Kb_wrtn Kbps appdatalv 306653 47493022 383822 103.2 loglv00 34 0 3340 2.8 data2lv 453 234543 234343 89.3
What are you looking at here?
- % iocnt: Reports back the number of read and write requests.
- Kb_read: Reports back the total data (kilobytes) from your measured interval that is read.
- Kb_wrtn: Reports back the amount of data (kilobytes) from your measured interval that is written.
- Kbps: Reports back the amount of data transferred in kilobytes.
Look at the man pages for all the commands discussed before you start to add them to your repertoire.
Tuning with lvmo
This section goes over using a specific logical volume tuning command. The
lvmo is used to set and display your pbuf tuning parameters. It is also used
to display blocked I/O statistics.
lvmo is one of those new commands first introduced in AIX Version 5.3. It's
important to note that the usage of the
lvmo command allows changes for LVM
pbuf tunables only that are dedicated to specific volume groups. The ioo utility is still
the only way to manage pbufs on a system-wide basis. This is because prior to AIX Version
5.3, the pbuf pool parameter was a system-wide resource. With the introduction of AIX
Version 5.3, LVM manages one pbuf pool for each volume group. What is a pbuf? A pbuf is
best defined as a pinned memory buffer. LVM uses these pbufs to control pending disk I/O
Let's display your
lvmo tunables for the data2vg volume group (see Listing 11).
Listing 11. Displaying lvmo tunables
# lvmo -v data2vg -a vgname = data2vg pv_pbuf_count = 1024 total_vg_pbubs = 1024 mag_vg_pbuf_count = 8192 perv_blocked_io_count = 7455 global_pbuf_count = 1024 global_blocked_io_count = 7455
What are the tunables here?
- pv_pbuf_count: Reports back the number of pbufs added when a physical volume is added to the volume group.
- Max_vg_pbuf_count: Reports back the max amount of pbufs that can be allocated for a volume group.
- Global_pbuf_count: Reports back the number of pbufs that are added when a physical volume is added to any volume group.
Let's increase the pbuf count for this volume group:
# lvmo -v redvg -o pv_pbuf_count=2048
Quite honestly, I usually stay away from
lvmo and use
more used to tuning the global parameters. It's important to note that if you increase
the pbuf value too much, you can actually see a degradation in performance.
This article focused on logical volumes and how they relate to the disk I/O subsystem. It defined logical volumes at a high level and illustrated how it relates to the application and physical layers. It also defined and discussed some best practices for inter-disk and intra-disk polices as they relate to creating and maintaining logical volumes. You looked at ways to monitor I/O usage for your logical volumes, and you analyzed the data that was captured from the commands that were used to help determine what your problems were. Finally, you actually tuned your logical volumes by determining and increasing the amount of pbufs used in a specific volume group. Part 3 of this series focuses on the application layer as you move on to file systems, using various commands to monitor and tune your file systems and disk I/O subsystems.
- Check out other articles in this series: Optimizing AIX 5L™ performance: Tuning disk performance
- Check out other parts in this series:Optimizing AIX 5L performance: Monitoring your CPU
- Improving database performance with AIX concurrent I/O: Read this white paper for more information on how to improve database performance.
- IBM Redbooks: Database Performance Tuning on AIX is designed to help system designers, system administrators, and database administrators design, size, implement, maintain, monitor, and tune a Relational Database Management System (RDMBS) for optimal performance on AIX.
- "Processor Affinity on AIX" (developerWorks, November 2006): Using process affinity settings to bind or unbind threads can help you find the root cause of troublesome hang or deadlock problems. Read this article to learn how to use processor affinity to restrict a process and run it only on a specified central processing unit (CPU).
- IBM Redbooks: The AIX 5L Differences Guide Version 5.3 Edition focuses on the differences introduced in AIX 5L Version 5.3 when compared to AIX 5L Version 5.2.
- AIX and UNIX: The AIX and UNIX developerWorks zone provides a wealth of information relating to all aspects of AIX systems administration and expanding your UNIX skills.
- AIX 5L Wiki: Discover a collaborative environment for technical information related to AIX.
- developerWorks technical events and webcasts: Stay current with developerWorks technical events and webcasts.
- Podcasts: Tune in and catch up with IBM technical experts.
Get products and technologies
- IBM trial software: Build your next development project with software for download directly from developerWorks.
- Participate in the developerWorks blogs and get involved in the developerWorks community.
- Participate in the AIX and UNIX forums