Category Archives: VHDX

SMB 3 NAS is preferable to DAS in a Windows environment

Microsoft is investing heavily in the Network Attached Storage (NAS) protocol SMB 3 and is clearly laying out a road map that suggests NAS is the future as opposed to Direct Attached Storage (DAS). Consider:

  • SQL Server 2012 system d/b and user d/bs, as well as Hyper-V 2012 workloads can be placed on NAS provided the NAS is SMB 3!
  • Microsoft made significant speed improvements in the SMB 3 client and server to have NAS achieve 97% of the speed of DAS, and this is without hardware acceleration.
  • Microsoft invested in SMB 3 Multi Channel by aggregating the bandwidth using parallel TCP channels using multiple NICs at the SMB 3 protocol layer. Multi Channel is all about speed AND reliability where failed I/Os are seamlessly moved to a different TCP channel when one channel fails.
  • Continuing on the speed theme, Microsoft invested in RDMA support via SMB Direct, which requires not just SMB 3, but also SMB 3 Multi Channel. The maximum IOPS on a Windows system is achieved when using SMB 3 NAS with SMB Direct support, NOT with DAS!
  • Going back to the reliability theme, SMB 3 includes support for Persistent Handles, which combined with the Witness Protocol, ensure applications such as SQL, Exchange, and Hyper-V never see an I/O failure, and the I/O is seamlessly moved to a different node as needed. This only works with SMB 3 NAS, and does NOT work with DAS!
  • I have been asked numerous times “But Microsoft has invested in Storage Spaces and Tiering where data is moved between SSD and spinning media to optimize performance. Does that not indicate Microsoft advocates DAS?” And my answer has always been “Storage Spaces is even more valuable when used as the storage backing a Windows Server 2012/R2 NAS!” Using Storage Spaces does not mean one has to abandon NAS.
  • Microsoft supports deduplication of VDI VMs, but the only supported configuration is with the VDI VM files residing on an SMB 3 based Windows Server 2012 R2 based NAS! (and not with DAS!)
  • To provide examples of other Microsoft efforts leveraging SMB3 , consider the simple “copy” or “xcopy” command to say copy a GBs large file. Microsoft changed the CopyFileEx API to leverage all SMB 3 features including SMB 3 credits, SMB 3 Multi Channel, and SMB Direct (RDMA) to ensure the file copy is as fast as possible.
  • The Microsoft Hyper-V team re-wrote live migration in Hyper-V 2012 R2 to leverage SMB 3. While migrating a VM, Hyper-V 2012 setup its own TCP channel to copy the VM RAM. Hyper-V 2012 R2 uses SMB 3, and thereby gets the speed/reliability improvements of SMB 3 while doing the same copy.

Tiered Storage and write back caching in Windows Server 2012 R2

With Windows Server 2012 R2, Microsoft introduces support for tiered storage and write back caching. With only rudimentary details available, this blog examines some highlights and also asks a few questions that I hope to make the content of a future blog.

Tiered storage and write back caching with Windows Server 2012 R2 requires:

  • A Storage Spaces capable set of rotating hard disks i.e. SAS, SATA, or USB hard disks. Obviously USB disks have their limitations in terms of IOPS.
  • A Storage Spaces (set) of flash storage – the word “flash” is used to loosely include SLC, MLC, and other kinds of SSD; again these must be SAS, SATA, or USB
  • Creation of a Storage Space that includes both rotating hard disk and flash disks

Tiered storage in Windows Server 2012 R2 provides just two levels. At any given time, a particular file may be

  • Fully on SSD because that is how Windows decided the file should be
  • Fully on HDD because that is how Windows decided the file should be
  • Partly on SSD and partly on HDD because that is how Windows decided the file should be
  • Pinned fully to either HDD or SSD by the administrator

With tiered storage, Windows tracks access to file ranges with a granularity of 1MB ranges. By default, a scheduled job runs at 1AM and moves the often accessed ranges of the file to SSD and the less often accessed parts of the file to HDD. The retiering can also be run on demand by the system administrator.

Windows Server 2012 R2 also introduces write back caching along with tiered storage. When writes happen, some (or all) of them end up with the new data on SSD tier. Presumably, at a later time when the scheduled optimization job runs, the data is moved to HDD.

The pros of the Windows Server 2012 tiered storage and write back caching:

  • Built into the operating system and free
  • If the understanding that write back caching simply places data on SSD and uses regular file structures is correct, the likelihood of  data corruption due to cache coherency and cache corruption is minimized

The cons of Windows server 2012 tiered storage and write back caching:

  • Only works with Storage spaces which requires SAS, SATA or USB, and in addition requires all storage to be non-RAID
  • Does not work with dedicated SSD designated as cache or in other words, the likelihood of the SSD becoming full and then write back caching being turned off is higher
  • Is not “real time” in the sense that potentially all writes go into the SSD. I could be wrong here since not enough details are available. But certainly, the process of moving often accessed file ranges to SSD, and less often accessed file ranges to HDD is not real time. It does allow the file to still be used during this retiering process, but it is still only periodic and by default only once per day.
  • The retiering process and also the process of monitoring and logging statistics as to what file ranges are actively accessed and what file ranges are less actively accessed may be resource intensive.

Hyper-V 2012 operations and the importance of SMB 3.0 Multichannel

Jose Baretto from Microsoft  has put out numerous blogs and talks, including some on SMB 3.0 and Multichannel. Some examples include The basics of SMB 3.0 Multichannel and Windows Server 2012 NIC Teaming and Multichannel. While it is difficult to add to the voluminous material Jose has contributed, this blog highlights the Hyper-V 2012 operational scenarios where SMB 3.0 Multi-Channel is useful and also points out an often overlooked fact of a configuration that is SMB 3.0 Multichannel capable. The goal of the blog is to draw attention to things that Jose has adequately already explained, but people have missed for some reason.

The Hyper-V 2012 team rewrote pieces of Hyper-V to support placing Hyper-V workloads on an SMB 3.0 share. Significantly, Hyper-V uses SMB 3.0 as a transport to move large amount of data/files from one SMB 3.0 NAS to another during live migration. But SMB 3.0 Multichannel is useful in scenarios other than live migration as well.

With Windows Server 2012, Microsoft also rewrote APIs such as the CopyFile API to leverage SMB 3.0 and its performance. CopyFile ensures there are multiple 1MB I/Os in flight and using the SMB 2/3 credit algorithm, the number of I/Os in flight can be increased. Also, these multiple I/Os in flight can flow on parallel TCP connections using SMB 3.0 Multi-Channel. Deploying a VM would typically involve copying a large vhdx file from say the test server to the production server. Another example would be Microsoft System Center copying a large (10s of GBs) vhdx file from a System Center Library Server to the production Hyper-V server. These file copies would benefit from SMB Multichannel.

And that brings me to the last point in this blog. Many folks I have talked to miss the fact that there is a case that allows for Multi-Channel without requiring multiple NICs. Here are the cases where SMB 3.0 Multichannel can come into play:

  • Either client OR server has multiple 1GB NICs. Note that it is not necessary for BOTH client and server to have multiple NICs.
  • Either client OR server has multiple 10GB NICs. Note that it is not necessary for BOTH client and server to have multiple NICs.
  • Both client AND server have a single 10 GB NIC and both the client and server NICs are RSS capable

As Jose Baretto has pointed out multiple times, even when both client and server each have only a single 10GB RSS capable NIC, that is sufficient to enable Multichannel. Microsoft observed that without RSS, TCP send/receive completion interrupts get serviced by a single CPU and that CPU becomes a bottleneck and prevents a typical Windows system from pumping a full 10Gbps through the NIC. A single TCP channel using RSS can alleviate the single CPU bottleneck, but it cannot use the full capability of the 10Gbps NIC. But a Multichannel TCP using RSS does have a chance of saturating the full 10Gbps capability.

Given that backup and vhdx file copy scenarios will occur often in a typical Hyper-V 2012 environment, SMB Multichannel thus not only plays an important role, but is very likely to be enabled with the hardware used for commercial deployment of Hyper-V 2012 VM workloads.

Windows Write Caching – Part 3 – An Overview For System Administrators

The Windows Cache Manager (also referred to as System Cache) acts as a single system-wide cache that contains driver code, application code, data for both, user mode applications as well as driver data. While an application can make API calls in a manner that guarantees the application data bypasses this cache, there is no way for an application to guarantee that its data WILL be cached. Because the behavior of the cache depends upon a number of factors and is very often non repeatable, the application and system administrator can only increase the likelihood that the application data will be cached. In other words, executing the same program multiple times is very likely to result in slightly different cache behavior each time. This is part of the reason why applications such as Microsoft SQL and Microsoft Exchange bypass the System Cache.

To illustrate the complexities involved, consider the seemingly simple act of copying a file from one volume to another. Some, but not all, of these have been originally described in References 8 and 9.

  • Either the source or the destination volume may be a local volume or a network volume
  • The access speed for the source and destination volumes may either be the same, or one may be significantly slower than the other. Further, the access speed can change depending upon a variety of factors such as network load, system load in terms of other application execution, resource usage e.g. I/O may switch from being cached to non cached and vice versa.
  • The optimum I/O size for the source and destination volumes may be either the same or significantly different
  • If both the source and destination volumes are on a Windows system, then the System Cache is involved in both reads and writes
  • The Windows team has spent a considerable amount of resources fine tuning the CopyFile and CopyFileEx APIs. Details are described in Reference 8, but the lesson to take away is the complexity of the issue and that further changes are probably forthcoming

Applications may

  • Use the CopyFile or CopyFileEx APIs and utilize the system cache
  • Use the CreateFile, ReadFile, WriteFile APIs and utilize the system cache for the source only, destination only, or both, or none.

Once you combine all of the various permutations and combinations offered by the above mentioned elements, the following situations can and do occur, when large files are being copied:

  • The system cache on the computer hosting the source file gets filled to a large extent with data from the source file. At the very least, this will affect other programs executing on that system.
  • The system cache on the computer system hosting the destination file gets filled with data for the destination file. This occurs fairly often since in the beginning, all of the destination file data is cached and thus writes appear to complete quickly. Once the destination file system cache hits a limit, disk writes (for flushing that cache) may occur slowly because the disk subsystem may be relatively slow
  • To complicate matters further, even when the data is flushed from system cache, it may be cached inside the block storage device (storage array)

When suspecting problems that may involve the System Cache, an administrator can

  • Inspect the application being used and switch to using a different application that explicitly does not use the System Cache. The Microsoft Server Performance Team Blog (Reference 7) explicitly suggests using Microsoft Exchange EseUtil as a file copy tool. The legal implications of using software shipped with Microsoft Exchange on a regular file server are beyond the scope of this document and best decided by your legal department
  • Use some other means to affect the System Cache e.g. use some other application that will consume up the System Cache, but not otherwise unduly load the system.
  • Attempt to administer the System Cache behavior utilizing in built utilities and/or registry keys

System caching can be controlled using administrative utilities and or a registry key.

To change the setting by editing the registry – as always beware of making registry changes and do so at your own risk – edit the registry key

HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management\LargeSystemCache

By default this DWORD is set to one (enabled) on Server SKUs and to zero (disabled)  on desktop SKUs.

On Windows XP, Microsoft provides a GUI to make the same changes which is preferable to making these changes via registry edits. Figure 2 shows the GUI that results from launching the Control Panel System Applet and then clicking the Advanced Tab

wincachefig2

Figure 2 Windows XP Control Panel System Applet Advanced Tab

Figure 3 shows the resulting System Cache size adjustment GUI when the advanced tab is clicked in Figure 2 on a Windows XP system

WinCacheFig3

Figure 3 Windows XP Control Panel System Applet Advanced Tab to adjust System Cache Size

Note that this GUI to change the System Cache size has been removed in Windows Vista.

Windows Server 2003, Windows Vista, and Windows Server 2008 Block Storage Cache Administration

Recall the earlier explanation of the bug in previous versions of Windows that ignored application requests to ensure data/metadata got committed to storage media and the subsequent fix made in Windows Server 2000 SP3 and also Windows XP SP2. To allow system administrators an informed choice, Microsoft made available a cache administration utility called DskCache.exe. This utility was only available by calling Microsoft PSS and could be obtained without incurring any monetary charge. To make it very clear that the DskCache utility should only be used in rare circumstances, Microsoft labeled it the “Power Protected Write Cache” and shipped it natively with Windows Server 2003 and higher versions of Windows. The new utility name emphasizes that it should be used only when the administrator is sure that the disk storage cache has a battery backup to ensure data integrity.

For Windows Server 2003 and higher versions of Windows, Microsoft has provided the equivalent of the DskCache.exe tool built into Windows. To use this feature:

  • Start Device Manager
  • Select the drive for which you wish to administer the caching policy
  • Select Properties
  • Click on Policies tab
  • Look for the option  “Enable write caching on the disk”  and make sure it is selected
  • And just below that, look for an option “Enable advanced performance”. This  option favors throughput/speed at the potential risk of data corruption.

The resulting GUI from following these steps is shown in Figure 4.

wincachefig4

 

Figure 4 – Windows Server 2003, Windows Vista & Windows Server 2008 disk caching policy administration

For Windows Server 2012, here is what the disk caching policy GUI looks like

WinCacheFig5

Figure 5 – Windows Server 2012 disk caching policy administration

Conclusion

This article described means by which application programmers can

  • Ensure that their file level data does not get cached in the Windows System Cache
  • Ensure that their file data does not get cached in the block storage layer and does get committed to storage media, given the correct hardware
  • Attempt to ensure, with no guarantee of success, that their file data does indeed get cached in the Windows System Cache

This article also describes means by which system administrators can attempt to ensure that data gets committed to storage media and does not get cached at either the System Cache or any block storage cache.

References

  1. Microsoft KB 241374 (http://support.microsoft.com/kb/241374/EN-US/) : Read and Write Access Required for SCSI Pass Through Request
  2. Microsoft KB 8373314: About Cache Manager in Windows Server 2003
  3. Microsoft KB 332023 Slow Disk Performance When Write Caching Is Disabled
  4. Nuances of Windows NT and SCSI disk performance article by Dilip Naik
  5. Force Unit Access Proposal
  6. Microsoft KB  870894 You receive a “Delayed Write Failed” error message in Windows XP Service Pack 2 or Windows XP Tablet PC Edition 2005
  7. Slow Large File Copy Issues – Microsoft Server Technical Support Performance Team Blog http://blogs.technet.com/askperf/archive/2007/05/08/slow-large-file-copy-issues.aspx
  8. Inside Vista SP1 File Copy Improvements – Mark Russinovich Blog http://blogs.technet.com/markrussinovich/
  9. Server Generates Delayed Errors Copying Very Large Files http://www.eggheadcafe.com/software/aspnet/32252624/server-generates-delayed.aspx
  10. Microsoft KB 920739  http://support.microsoft.com/kb/920739  Decreased Performance when copying files larger than 500 MB
  11. Serial ATA Program Revision 1.2 http://www.sata-io.org/documents/Interop_UnifiedTest_Rev1_2_v10_091707_000.pdf
  12. Disks, Lies, and damn disks http://perspectives.mvdirona.com/2008/04/17/DisksLiesAndDamnDisks.aspx
  13. Serial ATA in the Microsoft operating system environment http://www.microsoft.com/whdc/device/storage/serialATA_FAQ.mspx
  14. Enforcing Database Recoverability on Disks that lack Write-Through ftp://ftp.research.microsoft.com/pub/tr/TR-2008-36.pdf

 

 

 

Windows Write Caching – Part 2 An overview for Application Developers

Part1 of this blog presented an overview of the Windows storage stack.

Application programmers may use a number of interfaces to control the way their application data is cached or if they prefer, not cached and committed to disk media.

CreateFile API

Applications open a handle to a resource such as a file or a volume using the CreateFile API. One of the parameters to the CreateFile API is the dwFlagsAndAttributes parameter, which can be any valid combination of file attributes and file flags.

To avoid caching at the file system layer, an application should specify a valid combination that includes FILE_FLAG_NO_BUFFERING in this dwFlagsAndAttributes parameter. Some notable consequences of specifying this flag include:

  • Applications are expected to perform I/O that is in integer units of the volume sector size.
  • FILE_FLAG_NO_BUFFERING only applies to application data – the file system may still cache file metadata. Data and metadata may be flushed to disk by using the FlushFileBuffers API

Some well known applications such as Microsoft SQL and the Microsoft JET database (ships with Windows Server SKUs) specify FILE_FLAG_NO_BUFFERING with the CreateFile API.

Alternatively, applications may call the CreateFile API making sure that the FILE_FLAG_NO_BUFFERING flag is cleared in the parameter dwFlagsAndAttributes. In this case, it is likely that the file system and cache manager will cooperate to cache the application data, but there is no guarantee the caching will actually occur. A number of factors including the file/volume open mode specified in CreateFile, the data access pattern, and the load on the system will affect whether a particular application I/O is cached or not.

Note: The FILE_FLAG_NO_BUFFERING only affects file/volume data and does not apply to file/volume metadata, which may be cached even though FILE_FLAG_NO_BUFFERING is set.

Another relevant flag in the CreateFile API is FILE_FLAG_WRITE_THROUGH. Specified by itself, this flag ensures that both file/volume data and file/volume metadata are immediately flushed to storage media. Note that this does not mean the data does not traverse the cache. Referring to Figure 1, FILE_FLAG_NO_BUFFERING may mean that the data is written to Cache Manager and then immediately flushed from there. So the I/O may still be a buffered I/O.

  • Application developers favoring data integrity at the cost of reduced throughput should specify both FILE_FLAG_NO_BUFFERING and FILE_FLAG_WRITE_THROUGH are set when invoking the CreateFile API.
  • Application developers favoring speed should make sure these flags are cleared when invoking CreateFile.
  • Application developers favoring a balance of throughput and data integrity need to read the section on FlushFileBuffers API.

Hardware Considerations

The main issues with hardware are the FILE_FLAG_WRITE_THROUGH parameter and how an operating system can handle it.

SCSI protocols define a Force Unit Access (FUA) flag in the SCSI Request Block (SRB). Early versions of the SCSI protocol (circa 1997) defined FUA as optional, while later specifications have made it mandatory. In situations where it is imperative that data gets committed to media, ensure that the deployed  storage hardware does support the Force Unit Access semantics and that this feature is not disabled.

In the enterprise world, NTFS has been deployed on a lot of non SCSI hardware which increasingly so includes SATA aka Serial ATA storage. The implementation of FUA in these devices is, at best, inconsistent. Even when implemented, the default is to turn it off, because of severe performance penalties.

Lower end PCs continue to use what is loosely termed IDE/ATAPI (ATA Parallel Interface which is ATA retroactively renamed) storage drives. Strictly speaking, this is more a family of protocols, rather than a single protocol. The ATAPI-4 specification is implemented in Windows 2000 and the ATAPI-5 specification in Windows Server 2003. Neither of these have any equivalent to the Force Unit Access semantic of SCSI. This is a long understood problem and Microsoft proposed in 2002 that the relevant standard be revised.

Note: The conclusion is that there is a possibility of data corruption due to drives caching data, especially so in ATA, IDE, ATAPI, and SATA devices, to name a few. This problem exists for other popular operating systems such as the Apple OS X and Linux as well. See the “FlushFileBuffers” section within this document. Also be sure to read the “NTFS” section further ahead in this blog.

FlushFileBuffers API

The FlushFileBuffers API can be used to flush all the outstanding data and metadata on a single file or a whole volume. However, frequent use of this API can cause reduced throughput. Internally, Windows uses the SCSI Synchronize Cache or the IDE/ATAPI Flush cache commands.

  • Application developers desiring a combination of speed and data integrity can
    • Specify ~FILE_FLAG_NO_BUFFERING and ~FILE_FLAG_WRITE_THROUGH when invoking CreateFile
    • Write data as needed using an appropriate API such as WriteFile
    • Periodically call FlushFileBuffers to commit the data and meatadata to storage media – the exact period at which this occurs is application specific
    • Application developers favoring data integrity at the cost of reduced throughput should make judicious use of this API, especially so when they specify only FILE_FLAG_NO_BUFFERING while invoking the CreateFile API. In other words, these applications are using FILE_FLAG_NO_BUFFERING to ensure data is committed to storage media and using FlushFileBuffers to ensure metadata is committed to storage media.
    • Application developers favoring pure throughput at the risk of potential data corruption should never use FlushFileBuffers API.
    • As reference 14 describes, FlushFileBuffers can be used to mitigate the hardware not supporting write-Through

Liberal use of the FlushFileBuffers API can severely affect system throughput. This is because at the file system layer, it is quite clear what data blocks belong to what file. So when FlushFileBuffers is invoked, it is also apparent what data buffers need to be written out to media. However, at the block storage layer – shown as “Sector I/O” in Figure 1, it is difficult to track what blocks are associated with what files. Consequently the only way to honor any FlushFileBuffers call is to make sure all data is flushed to media. Therefore, not only is more data written out than originally intended, but the larger amount of data can affect other I/O optimizations such as queuing of the writes.

There is also a bright side to this picture. While it is true that FlushFileBuffers, if handled properly by all involved layers, will flush all data and cause performance degradation, it can also help in preserving data integrity. An application that “forgets” to invoke FlushFileBuffers will still have its data committed to media due to other applications invoking this API.

FlushFileBuffers should be judiciously used as needed.

Hardware Considerations

The SCSI-3 protocol defines a command SYNCHRONIZE_CACHE that commits data from the storage cache to media. Hence SCSI devices are good candidates for applications that are highly sensitive to data being committed to media. However, it is always good practice to verify that  a particular SCSI hardware does implement the SYNCHRONIZE_CACHE command.

The relevant ATA/IDE specifications define an equivalent command FLUSH_CACHE. ATAPI-4 (relevant circa 2000) defines the FLUSH_CACHE command as optional; ATAPI-5 makes this command mandatory. As always, verify the exact functionality of a particular hardware and do not assume it implements the relevant standard precisely.

Client/Server application considerations

Both the Common Internet File Systems (CIFS) and the new SMB 2.0 protocols define a flush SMB command. Therefore the CIFS/SMB redirector built into Windows can easily propagate a FlushFileBuffers command over to a file server.

DeviceIOControl API

While data may be cached at the file system layer, it can (and does) occur at other layers such as the disk block level. For the purposes of this discussion, caching done within a disk or a disk controller is referred to as disk block caching.

Storage devices cache data to enhance throughput, but sometimes at the cost of data integrity. Some storage devices include their own battery backup to enhance data integrity.

Further increasing ambiguity is the fact that there is no standard manner to determine whether such caching is indeed occurring or not. Some storage systems have their own battery backed cache and hence under many circumstances, it may be OK to leave the data in this battery backed cached and treat it as being committed to media. Some of these storage systems ignore the command to commit data to media.

The DeviceIOControl API can be used to inspect block storage caching configurations and also potentially set these configurations. Both the caching policy inspection and setting have limitations as well.

Windows offers a number of ways for applications to programmatically affect the caching at the sector I/O level. Many of these interfaces consist of calling the DeviceIOControl API with some different function code. Windows NT 4.0 SP4 and higher versions of Windows require administrator privileges to submit SCSI pass-through requests. SCSI pass-through requests are submitted via the WIN32 DeviceIOControl API, which requires a handle to a file or a volume as a parameter. This handle is obtained via the CreateFile API and starting with Windows NT 4.0 SP4, the CreateFile function requires GENERIC_READ and GENERIC_WRITE to be specified in the dwDesiredAccess parameter of the CreateFile API.

Windows defines a number of IOCTL codes that may be used to inspect and control the disk block caching functionality. These include:

IOCTL_DISK_GET_CACHE

Returns information about whether the disk cache is enabled or not. The function works only where the disk returns correct information via a SCSI mode page.

IOCTL_DISK_SET_CACHE

-Sets disk caching functionality to be enabled or disabled. The function works only where the disk implements SCSI mode pages.

IOCTL_STORAGE_QUERY_PROPERTY

Windows Vista and higher versions of Windows support another interface to retrieve disk cache properties.

Pros

  • A single interface that applications can code to – in the absence of this single interface, applications would need to understand the nuances of various devices e.g. for 1394 devices, obtain page 6 of the mode data, but for SCSI compliant devices, obtain page 8 of the mode data, etc.
  • Windows implements the necessary details to retrieve  the information in a transport dependent manner (SCSI/IDE/etc) to surface the information
  • A richer interface that can provide information not just whether the disk implements a cache, but also what kind of cache
  • Time will tell whether this interface continues to be evolve and be populated with more data

Cons

  • Yet another interface that applications need to code to
  • Does not work for RAID devices
  • Does not work for Flash drives
  • Not yet widely implemented by storage vendors
  • Only available in the newer versions of Windows
  • Does not (yet at least) allow setting of caching properties

For an application to retrieve information about a device’s write cache property, use the STORAGE_WRITE_CACHE_PROPERTY structure with the IOCTL_STORAGE_QUERY_PROPERTY request.

NTFS and flushing behavior

NTFS, when introduced with Windows NT 3.X depended upon the SCSI Forced Unit Access behavior to ensure its meta data is flushed to media. As described above, NTFS has been deployed on a fair number of non SCSI devices, all of which have an inconsistent implementation of FUA. Perhaps the saving grace may have been that the consumer devices do not have a cache, and hence even without FUA, the data hits the media. In any case, what is worth noting is that all NTFS versions upto and including Windows 7 and Windows 2008 R2 depend upon FUA to ensure that NTFS. NTFS in Windows 8 has switched to using the FlushFileBuffers API instead of depending upon the Forced Unit Access behavior.

This concludes an overview of the knobs an application developer can twist and turn to influence application data write caching. Part 3 of this blog will present an overview of the knobs a system administrator can twist and turn to influence write data caching.

Windows Write caching – Part 1 Overview


Certain Windows applications such as database applications need to ensure their I/O is committed to media, even at the cost of reduced throughput. However, at times an administrator has faith in the hardware and is willing to accept a small risk of data corruption in favor of a higher throughput by allowing caching to occur .

This is a 3 part blog that concentrates on the write caching behavior in the Windows storage stack.

  1. Part 1 presents an overview of the Windows storage stack with specific reference to write caching
  2. Part 2 presents the “knobs” an application programmer can twist and turn to affect write caching
  3. Part 3 presents the “knobs” a system administrator can twist and turn to affect write caching

Windows Storage Stack

Figure 1 Windows Storage I/O Stack

Figure 1 shows a simplified overview of the Windows Storage I/O stack. Starting from the top of Figure 1,

  • Applications make I/O requests. Figure 1 concentrates on write requests and hence the unidirectional arrows from the application towards the disk media.
  • Depending upon the nature of the I/O (decided partly by the way the application opens a file or volume), some I/O requests completely bypass the Windows Cache Manager and go straight from the file system to the Volume Manager layer. This is labeled Unbuffered I/O in Figure 1. As will be explained later, applications can ensure that their I/O is Unbuffered.
  • Alternatively, Application I/O may traverse the buffered path labeled in Figure 1. While applications may strive to ensure their I/O is buffered, in reality, there is no way to ensure this. I/O is buffered depending upon a number of factors such as nature of file open, the type of I/O, the history of the application I/O, the load on the system, etc.
  • The Volume Manager performs sector I/O. While the application may strive to ensure that there is no caching at the sector I/O level, the reality is that applications have limited success in some cases. This is discussed in more detail within the document.

Different types of Write Caching

Irrespective of whether the data is written at the file system level or at the disk block level, write caching can be broadly classified into two categories:

  • Write-through caching:  where data is written to cache AND also written to non volatile media. The data integrity is high, but write performance is slower whereas read performance is enhanced
  • Writeback caching: where data is written to cache, the operating system write request is completed, and the data is lazily written to media at a later point in time. Writeback Caching emphasizes write performance, but at the possible loss of data integrity.

Part 2 of this blog will describe the APIs an application developer can use to control write caching behavior.

The perils of alignment for memory access and disk I/O

In my earlier blog, I described how Visual Studio (VS) 2012 is now a requirement for writing kernel mode drivers on both the x86/x64 Intel/AMD, and also the ARM version of Windows 8. So I installed VS 2012 RC on two different laptops and was unhappy with the installation time. I must place on my record my appreciation for the Visual Studio team, which has been very diligent in following up and looking into the issue. Of course, I will acknowledge that my belief of “it takes too long” could be incorrect, and I may be encountering unusual circumstances on both my systems. So with that caveat that perhaps “I am encountering a one off situation”, here we go with my analysis.
First, a couple of references are in order.

  1.  To quote from MSDN “In this document we explain why you should care about data alignment, the costs if you do not, how to get your data aligned, and what to do when you cannot. You will never look at your data access the same way again.” The point is; aligned memory access in Windows is very important.
  2. It is equally important to ensure that writes are aligned as well. Most current disks write data in 512 chunks called sectors. So if you write 512 bytes at offset zero, a single write suffices. But if you write 512 bytes at offset 1, the I/O spans 2 disk sectors. So a single write becomes read 2 disk sectors, copy over the new 512 bytes of data, and issue 2 sector writes, each of 512 bytes. So a 512 byte write becomes a 1024 byte read and a 1024 byte write. Here is an MSDN blog explaining among other things, the importance of aligned I/Os for SQL. And here is a another MSDN SQL blog explaining the importance of aligned I/O

Now back to the topic at hand – installing Visual Studio 2012 RC and analyzing possible causes for why it takes as long as it does. So I decided to investigate further, by tracing the I/Os using Sysinternals (now part of MSDN) tool Process Monitor.

Here is a screen shot showing a small part of the I/O of the installation. Note that I randomly located this I/O pattern. I also cursorily checked that other files have similar behavior; in particular, write an odd number of bytes at offset zero, and then proceed to write the rest of the file.

Image

For file DataCollection.dll, please notice

  1. The write at offset zero for 32,447 bytes
  2. The write at offset 32,447 for 32,768 bytes
  3. The write at offset 62,215 for 16761 bytes
  4. The total file size is 81,976 bytes and 32,447 + 32,768 + 16,761 = 81976

Now apply the logic of the references quoted – in particular, the importance of aligned memory access, and aligned disk I/O access.

At the very least, the each of the 3 I/Os will consist of a 1 or 3 byte copy, a copy of some N DWORDs, followed by a 1 or 3 byte copy. This could have been completely avoided by doing 3 I/Os, each consisting of an even number of bytes. There is a penalty to be paid for the 1 byte and 3 byte memory access.

I must admit that this trace is at the file system layer. It is certain that before the I/O hits the disk, which is a block mode I/O, the Windows Cache Manager and I/O subsystem will have intervened to make the I/O aligned and an integral number of sectors. There will still be some disk I/O penalties however, when some writes get split across 2 adjacent sectors. This could be avoided. Consider the case where say part of the file has been written, and is in cache. And the I/O pattern guarantees that there will be an odd number of bytes cached, until the final odd length write arrives. Now imagine that for some reason, the cache gets flushed before the last write arrives. This could be because the file is very large, or there is memory pressure. This means that the cache manager will zero fill a buffer until the end of a sector (an odd number of bytes) and then write out that sector. When the next write arrives, this just flushed sector needs to be read, the zero filled bytes are copied over with the newly arrived data, and then the same sector is written – again!

There is no perceivable advantage in making the I/O nonaligned – and significant potential harm. It is difficult to estimate how much VS 2012 installation will speed up, were the writes to be aligned.
There are other oddities as well in the trace, but I will write about those in future blogs.

I invite reader comments on whether they believe this I/O pattern is within acceptable bounds. For readers willing to trace their VS 2012 install, I would also welcome feedback as to whether they observe this pattern.

Windows 8 VHDX file instant dedupe wish list

I have been testing the Windows 8 dedupe feature, especially so for large VHDX files. But the testing has revealed a major “wish”. Hopefully somebody from the right department at Microsoft reads this and at least puts it on a feature list for the future.

Here is a scenario I exercised – and it seems to be a very common scenario

  1. Create a Windows Server VM inside a 40GB VHDX file- call it VM1.vhdx
  2. Xcopy – (and yes – xcopy /J –see my previous blog “Tips for copying VHD and VHDX files”)  the VM1.vhdx  file to say VM2.vhdx. That’s 40 GB of reads and 40 GB of writes.  
  3. Repeat the xcopy to a different destination file – Xcopy /J VM1.vhdx to VM3.vhdx and that’s 40 GB more reads and 40GB more writes.
  4. Fire up VM1, enter license info, assign computer name, assign IP address, etc. Turn into a file server
  5. Fire up VM2, enter license info, etc, install Microsoft Exchange into a second VM and turn into an Exchange Server
  6. Fire up VM3, enter license info, etc and install SQL Server into a third VM and turn it into a SQL Server
  7. Now let the system idle, make sure it does not hibernate, wait for dedupe to read all 3 VHDX files ( 3 X 40 GB worth of reads, etc) and dedupe the files.

Instead, here is an alternative sequence that would be really useful

  1. Create a Windows Server VM inside a 40GB VHDX file
  2. Run a PS script that creates an instantly deduped second copy of this  VHDX file – with all the associated dedupe metadata. So now I have 2 VHDX files that are identical and have been deduped. The PS script would have to invoke some custom dedupe code Microsoft could ship. Create a new file entry for say VM2.vhdx and create the dedupe metadata for VM1.vhdx and VM2.vhdx.
  3. Repeat the same PS script with different parameters and now I have 3 identical VHDX files, all deduped
  4. Repeat steps 4 through 6 from the first sequence – step 7 – the dedupe step is not needed

This would save 100s of GBs of reads and writes, and administrator time, increasing productivity. Whether you call this instant dedupe or not is up to you.

In the interest of keeping the focus on the instant dedupe scenario, I have deliberately avoided the details of requiring Sysprep’ed installations. But the audience I am targeting with this blog will certainly understand the nuances of requiring Sysprep.

If you are a Microsoft MVP reading this blog, and you agree, please comment on the blog, and email your MVP lead asking for this feature.