Quantum Marine Sanitation System 6 01376 05 User Guide

File System Tuning Guide File System Tuning Guide File System Tuning Guide  
StorNext 3.0®  
6-01376-05  
Download from Www.Somanuals.com. All Manuals Search And Download.  
Contents  
Mount Command Options...................................................................... 16  
The Distributed LAN (Disk Proxy) Networks............................................. 18  
Network Configuration and Topology ................................................. 20  
Distributed LAN Servers................................................................................ 22  
Windows Memory Requirements ................................................................. 22  
Sample FSM Configuration File..................................................................... 25  
StorNext File System Tuning Guide  
i
Download from Www.Somanuals.com. All Manuals Search And Download.  
0StorNext File System Tuning  
The StorNext File System (SNFS) provides extremely high performance  
for widely varying scenarios. Many factors determine the level of  
performance you will realize. In particular, the performance  
characteristics of the underlying storage system are the most critical  
factors. However, other components such as the Metadata Network and  
MDC systems also have a significant effect on performance.  
Furthermore, file size mix and application I/O characteristics may also  
present specific performance requirements, so SNFS provides a wide  
variety of tunable settings to achieve optimal performance. It is usually  
best to use the default SNFS settings, because these are designed to  
provide optimal performance under most scenarios. However, this guide  
discusses circumstances in which special settings may offer a  
performance benefit.  
The Underlying Storage System  
The performance characteristics of the underlying storage system are the  
most critical factors for file system performance. Typically, RAID storage  
systems provide many tuning options for cache settings, RAID level,  
segment size, stripe size, and so on.  
StorNext File System Tuning Guide  
1
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
The Underlying Storage System  
RAID Cache  
Configuration  
The single most important RAID tuning component is the cache  
0
configuration. This is particularly true for small I/O operations.  
Contemporary RAID systems such as the EMC CX series and the various  
Engenio systems provide excellent small I/O performance with properly  
tuned caching. So, for the best general purpose performance  
characteristics, it is crucial to utilize the RAID system caching as fully as  
possible.  
For example, write-back caching is absolutely essential for metadata  
stripe groups to achieve high metadata operations throughput.  
However, there are a few drawbacks to consider as well. For example,  
read-ahead caching improves sequential read performance but might  
reduce random performance. Write-back caching is critical for small write  
performance but may limit peak large I/O throughput. Some RAID  
systems cannot safely support write-back caching without risk of data  
loss, which is not suitable for critical data such as file system metadata.  
Consequently, this is an area that requires an understanding of  
application I/O requirements. As a general rule, RAID system caching is  
critically important for most applications, so it is the first place to focus  
tuning attention.  
RAID Write-Back  
Caching  
Write-back caching dramatically reduces latency in small write  
operations. This is accomplished by returning a successful reply as soon  
as data is written into cache, and then deferring the operation of actually  
writing the data to the physical disks. This results in a great performance  
improvement for small I/O operations.  
0
Many contemporary RAID systems protect against write-back cache data  
loss due to power or component failure. This is accomplished through  
various techniques including redundancy, battery backup, battery-  
backed memory, and controller mirroring. To prevent data corruption, it  
is important to ensure that these systems are working properly. It is  
particularly catastrophic if file system metadata is corrupted, because  
complete file system loss could result. Check with your RAID vendor to  
make sure that write-back caching is safe to use.  
Minimal I/O latency is critically important for metadata stripe groups to  
achieve high metadata operations throughput. This is because metadata  
operations involve a very high rate of small writes to the metadata disk,  
so disk latency is the critical performance factor. Write-back caching can  
be an effective approach to minimizing I/O latency and optimizing  
StorNext File System Tuning Guide  
2
Download from Www.Somanuals.com. All Manuals Search And Download.  
   
StorNext File System Tuning  
The Underlying Storage System  
metadata operations throughput. This is easily observed in the hourly  
File System Manager (FSM) statistics reports in the cvlog file. For  
example, here is a message line from the cvlog file:  
PIO HiPriWr SUMMARY SnmsMetaDisk0 sysavg/350 sysmin/333 sysmax/367  
This statistics message reports average, minimum, and maximum write  
latency (in microseconds) for the reporting period. If the observed  
average latency exceeds 500 microseconds, peak metadata operation  
throughput will be degraded. For example, create operations may be  
around 2000 per second when metadata disk latency is below 500  
microseconds. However, if metadata disk latency is around 5  
milliseconds, create operations per second may be degraded to 200 or  
worse.  
Another typical write caching approach is a “write-through.” This  
approach involves synchronous writes to the physical disk before  
returning a successful reply for the I/O operation. The write-through  
approach exhibits much worse latency than write-back caching; therefore,  
small I/O performance (such as metadata operations) is severely  
impacted. It is important to determine which write caching approach is  
employed, because the performance observed will differ greatly for small  
write I/O operations.  
In some cases, large write I/O operations can also benefit from caching.  
However, some SNFS customers observe maximum large I/O  
throughput by disabling caching. While this may be beneficial for special  
large I/O scenarios, it severely degrades small I/O performance;  
therefore, it is suboptimal for general-purpose file system performance.  
RAID Read-Ahead  
Caching  
RAID read-ahead caching is a very effective way to improve sequential  
read performance for both small (buffered) and large (DMA) I/O  
operations. When this setting is utilized, the RAID controller pre-fetches  
disk blocks for sequential read operations. Therefore, subsequent  
application read operations benefit from cache speed throughput, which  
is faster than the physical disk throughput.  
0
This is particularly important for concurrent file streams and mixed I/O  
streams, because read-ahead significantly reduces disk head movement  
that otherwise severely impacts performance.  
While read-ahead caching improves sequential read performance, it does  
not help random performance. Furthermore, some SNFS customers  
actually observe maximum large sequential read throughput by disabling  
caching. While disabling read-ahead is beneficial in these unusual cases,  
StorNext File System Tuning Guide  
3
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
The Underlying Storage System  
it severely degrades typical scenarios. Therefore, it is unsuitable for most  
environments.  
RAID Level, Segment  
Size, and Stripe Size  
Configuration settings such as RAID level, segment size, and stripe size  
are very important and cannot be changed after put into production, so it  
is critical to determine appropriate settings during initial configuration.  
0
The best RAID level to use for high I/O throughput is usually RAID5.  
The stripe size is determined by the product of the number of disks in the  
RAID group and the segment size. For example, a 4+1 RAID5 group with  
64K segment size results in a 256K stripe size. The stripe size is a very  
critical factor for write performance because I/Os smaller than the stripe  
size may incur a read/modify/write penalty. It is best to configure  
RAID5 settings with no more than 512K stripe size to avoid the read/  
modify/write penalty. The read/modify/write penalty is most  
noticeable in the absence of “write-back” caching being performed by the  
RAID controller.  
The RAID stripe size configuration should typically match the SNFS  
StripeBreadth configuration setting when multiple LUNs are utilized in a  
stripe group. However, in some cases it might be optimal to configure the  
SNFS StripeBreadth as a multiple of the RAID stripe size, such as when  
the RAID stripe size is small but the user's I/O sizes are very large.  
However, this will be suboptimal for small I/O performance, so may not  
be suitable for general purpose usage.  
RAID1 mirroring is the best RAID level for metadata and journal storage  
because it is most optimal for very small I/O sizes. It is also very  
important to allocate entire physical disks for the Metadata and Journal  
LUNs in order to avoid bandwidth contention with other I/O traffic.  
Metadata and Journal storage requires very high IOPS rates (low latency)  
for optimal performance, so contention can severely impact IOPS (and  
latency) and thus overall performance. If Journal I/O exceeds 1ms  
average latency, you will observe significant performance degradation.  
It can be useful to use a tool such as lmdd to help determine the storage  
system performance characteristics and choose optimal settings. For  
example, varying the stripe size and running lmdd with a range of I/O  
sizes might be useful to determine an optimal stripe size multiple to  
configure the SNFS StripeBreadth.  
StorNext File System Tuning Guide  
4
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
File Size Mix and Application I/O Characteristics  
File Size Mix and Application I/O Characteristics  
It is always valuable to understand the file size mix of the target dataset  
as well as the application I/O characteristics. This includes the number of  
concurrent streams, proportion of read versus write streams, I/O size,  
sequential versus random, Network File System (NFS) or Common  
Internet File System (CIFS) access, and so on.  
For example, if the dataset is dominated by small or large files, various  
settings can be optimized for the target size range.  
Similarly, it might be beneficial to optimize for particular application I/O  
characteristics. For example, to optimize for sequential 1MB I/O size it  
LUNs with 256K stripe size.  
However, optimizing for random I/O performance can incur a  
performance trade-off with sequential I/O.  
Furthermore, NFS and CIFS access have special requirements to consider  
as described in the Direct Memory Access (DMA) I/O Transfer section.  
Direct Memory Access  
(DMA) I/O Transfer  
To achieve the highest possible large sequential I/O transfer throughput,  
SNFS provides DMA-based I/O. To utilize DMA I/O, the application  
must issue its reads and writes of sufficient size and alignment. This is  
called well-formed I/O. See the mount command settings  
auto_dma_read_length and auto_dma_write_length, described in the  
Mount Command Options on page 16.  
0
0
Reads and writes that aren't well-formed utilize the SNFS buffer cache.  
This also includes NFS or CIFS-based traffic because the NFS and CIFS  
daemons defeat well-formed I/Os issued by the application.  
Buffer Cache  
There are several configuration parameters that affect buffer cache  
performance. The most critical is the RAID cache configuration because  
buffered I/O is usually smaller than the RAID stripe size, and therefore  
incurs a read/modify/write penalty. It might also be possible to match  
the RAID stripe size to the buffer cache I/O size. However, kernel  
memory fragmentation can defeat attempts to increase the SNFS buffer  
cache I/O size (see the cachebufsize setting described in the Mount  
StorNext File System Tuning Guide  
5
Download from Www.Somanuals.com. All Manuals Search And Download.  
   
StorNext File System Tuning  
File Size Mix and Application I/O Characteristics  
Command Options on page 16). So, it is typically most important to  
optimize the RAID cache configuration settings described earlier in this  
document.  
It is usually best to configure the RAID stripe size no greater than 256K  
for optimal small file buffer cache performance.  
For more buffer cache configuration settings, see Mount Command  
Options on page 16.  
It is best to isolate NFS and/or CIFS traffic off of the metadata network to  
eliminate contention that will impact performance. For optimal  
performance it is necessary to use 1000BaseT instead of 100BaseT. When  
possible, it is also best to utilize TCP Offload capabilities as well as jumbo  
frames.  
NFS / CIFS  
0
It is best practice to have clients directly attached to the same network  
switch as the NFS or CIFS server. Any routing required for NFS or CIFS  
traffic incurs additional latency that impacts performance.  
It is critical to make sure the speed/duplex settings are correct, because  
this severely impacts performance. Most of the time auto-detect is the  
correct setting. Some managed switches allow setting speed/duplex (for  
example 1000Mb/full,) which disables auto-detect and requires the host to  
be set exactly the same. However, if the settings do not match between  
switch and host, it severely impacts performance. For example, if the  
switch is set to auto-detect but the host is set to 1000Mb/full, you will  
observe a high error rate along with extremely poor performance. On  
Linux, the mii-diag tool can be very useful to investigate and adjust speed/  
duplex settings.  
A higher performance alternative to NFS and CIFS is ISCSI. If  
performance requirements cannot be achieved with NFS or CIFS, SCSI  
should be considered.  
It can be useful to use a tool such as netperf to help verify network  
performance characteristics.  
StorNext File System Tuning Guide  
6
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Network  
The Metadata Network  
As with any client/server protocol, SNFS performance is subject to the  
limitations of the underlying network. Therefore, it is recommended that  
you use a dedicated Metadata Network to avoid contention with other  
network traffic. Either 100BaseT or 1000BaseT is required, but for a  
dedicated Metadata Network there is usually no benefit from using  
1000BaseT over 100BaseT. Neither TCP offload nor are jumbo frames  
required.  
It is best practice to have all SNFS clients directly attached to the same  
network switch as the MDC systems. Any routing required for metadata  
traffic will incur additional latency that impacts performance.  
It is critical to ensure that speed/duplex settings are correct, as this will  
severely impact performance. Most of the time auto-detect is the correct  
setting. Some managed switches allow setting speed/duplex, such as  
100Mb/full, which disables auto-detect and requires the host to be set  
exactly the same. However, performance is severely impacted if the  
settings do not match between switch and host. For example, if the switch  
is set to auto-detect but the host is set to 100Mb/full, you will observe a  
high error rate and extremely poor performance. On Linux the mii-diag  
tool can be very useful to investigate and adjust speed/duplex settings.  
It can be useful to use a tool like netperf to help verify the Metadata  
Network performance characteristics. For example, if netperf -t TCP_RR  
reports less than 15,000 transactions per second capacity, a performance  
penalty may be incurred.  
The Metadata Controller System  
The CPU and memory power of the MDC System are important  
performance factors, as well as the number of file systems hosted per  
system. In order to ensure fast response time it is necessary to use  
dedicated systems, limit the number of file systems hosted per system  
(maximum 8), and have an adequate CPU and memory.  
StorNext File System Tuning Guide  
7
Download from Www.Somanuals.com. All Manuals Search And Download.  
   
StorNext File System Tuning  
The Metadata Controller System  
Some metadata operations such as file creation can be CPU intensive, and  
benefit from increased CPU power. The MDC platform is important in  
these scenarios because lower clock- speed CPUs such as Sparc and Mips  
degrade performance.  
Other operations can benefit greatly from increased memory, such as  
directory traversal. SNFS provides three config file settings that can be  
used to realize performance gains from increased memory:  
BufferCacheSize, InodeCacheSize, and ThreadPoolSize.  
However, it is critical that the MDC system have enough physical  
memory available to ensure that the FSM process doesn’t get swapped  
out. Otherwise, severe performance degradation and system instability  
can result.  
FSM Configuration File  
Settings  
The following FSM configuration file settings are explained in greater  
detail in the cvfs_config man page. For a sample FSM configuration file,  
see Sample FSM Configuration File on page 25.  
0
The examples in the following sections are excerpted from the sample  
configuration file from Sample FSM Configuration File on page 25.  
Stripe Groups  
0
Splitting apart data, metadata, and journal into separate stripe groups is  
usually the most important performance tactic. The create, remove, and  
allocate (e.g., write) operations are very sensitive to I/O latency of the  
journal stripe group. Configuring a separate stripe group for journal  
greatly benefits the speed of these operations because disk seek latency is  
minimized. However, if create, remove, and allocate performance aren't  
critical, it is okay to share a stripe group for both metadata and journal,  
but be sure to set the exclusive property on the stripe group so it doesn't  
get allocated for data as well. It is recommended that you assign only a  
single LUN for each journal or metadata stripe group. Multiple metadata  
stripe groups can be utilized to increase metadata I/O throughput  
through concurrency. RAID1 mirroring is optimal for metadata and  
journal storage. Utilizing the write-back caching feature of the RAID  
system (as described previously) is critical to optimizing performance of  
the journal and metadata stripe groups.  
StorNext File System Tuning Guide  
8
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
The Metadata Controller System  
Example:  
[stripeGroup RegularFiles]  
Status UP  
Exclusive No  
Read Enabled  
Write Enabled  
StripeBreadth 256K  
##Non-Exclusive stripeGroup for all Files##  
MultiPathMethod Rotate  
Node CvfsDisk6 0  
Node CvfsDisk7 1  
Affinities  
Affinities are another stripe group feature that can be very beneficial.  
0
Affinities can direct file allocation to appropriate stripe groups according  
to performance requirements. For example, stripe groups can be set up  
with unique hardware characteristics such as fast disk versus slow disk,  
or wide stripe versus narrow stripe. Affinities can then be employed to  
steer files to the appropriate stripe group. For optimal performance, files  
that are accessed using large DMA-based  
I/O could be steered to wide-stripe stripe groups. Less performance-  
critical files could be steered to slow disk stripe groups. Small files could  
be steered to narrow-stripe stripe groups.  
Example:  
[stripeGroup VideoFiles]  
Status UP  
Exclusive Yes  
##These Two lines set Exclusive stripeGroup##  
Affinity VideoFiles ##for Video Files Only##  
Read Enabled  
Write Enabled  
StripeBreadth 4M  
MultiPathMethod Rotate  
Node CvfsDisk2 0  
Node CvfsDisk3 1  
StorNext File System Tuning Guide  
9
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
StripeBreadth  
0
This setting must match the RAID stripe size or be a multiple of the RAID  
stripe size. Matching the RAID stripe size is usually the most optimal  
setting. However, depending on the RAID performance characteristics  
and application I/O size, it might be beneficial to use a multiple of the  
RAID stripe size. For example, if the RAID stripe size is 256K, the stripe  
group contains 4 LUNs, and the application to be optimized uses DMA I/  
O with 8MB block size, a StripeBreadth setting of 2MB might be optimal.  
In this example the 8MB application I/O is issued as 4 concurrent 2MB I/  
Os to the RAID. This concurrency can provide up to a 4X performance  
increase. This typically requires some experimentation to determine the  
RAID characteristics. The lmdd utility can be very helpful. Note that this  
setting is not adjustable after initial file system creation.  
Example:  
[stripeGroup AudioFiles]  
Status UP  
Exclusive Yes  
##These two lines set Exclusive stripeGroup ##  
Affinity AudioFiles ##for Audio Files Only##  
Read Enabled  
Write Enabled  
StripeBreadth 1M  
MultiPathMethod Rotate  
Node CvfsDisk4 0  
Node CvfsDisk5 1  
BufferCacheSize  
0
This setting consumes up to 2X bytes of memory times the number  
specified. Increasing this value can reduce latency of any metadata  
operation by performing a hot cache access to directory blocks, inode  
information, and other metadata info. This is about 10 to 1000 times faster  
than I/O. It is especially important to increase this setting if metadata I/O  
latency is high, (for example, more than 2ms average latency). We  
recommend sizing this according to how much memory is available;  
more is better.  
Example: # BufferCacheSize  
64M # default 32MB  
StorNext File System Tuning Guide  
10  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
InodeCacheSize  
0
This setting consumes about 800-1000 bytes of memory times the number  
specified. Increasing this value can reduce latency of any metadata  
operation by performing a hot cache access to inode information instead  
of an I/O to get inode info from disk, about 100 to 1000 times faster. It is  
especially important to increase this setting if metadata I/O latency is  
high, (for example, more than 2ms average latency). You should try to  
size this according to the sum number of working set files for all clients.  
The recommended range is 16K to 64K.  
Example: InodeCacheSize  
16K # 800-1000 bytes each, default 8K  
ThreadPoolSize  
0
This setting consumes 512 KB memory times the number specified.  
Increasing this value can improve concurrency of metadata operations.  
For example, if many client processes are executing concurrently, the  
thread pool can become exhausted by I/O wait time. Increasing the  
thread pool size permits hot cache operations to be processed that would  
otherwise be backed up behind the I/O-bound operations. There are  
various O/S limits to the number of threads that can cause fatal problems  
for the FSM daemon, so it's not a good idea to set this setting too high. A  
range from 32 to 128 is recommended, depending on the amount of  
available memory. It is recommended to size it according to the max  
threads FSM hourly statistic reported in the cvlog file.  
Example: ThreadPoolSize  
32  
# default 16, 512 KB memory per thread  
ForcestripeAlignment  
0
This setting should always be set to Yes. This is critical if the largest  
StripeBreadth defined is greater than 1MB. Note that this setting is not  
adjustable after initial file system creation.  
Example: ForcestripeAlignment Yes  
FsBlockSize  
0
The optimal settings are in the range of 4K, 8K, 16K, or 32K. Settings  
greater than 32K can adversely impact performance because metadata I/  
O operations are performed less efficiently. Any value greater than 4K  
StorNext File System Tuning Guide  
11  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
severely consumes metadata space in cases where the file-to-directory  
ratio is less than 100 to 1. However, startup and failover time can be  
minimized by increasing FsBlockSize. This is very important for multi-  
terabyte file systems, and especially when the metadata servers have  
slow CPU clock speed (such as Sparc and Mips). A good rule of thumb is  
to use 16K unless other requirements such as directory ratio dictate  
otherwise. Note that this setting is not adjustable after initial file system  
creation, so it is very important to give it careful consideration during  
initial configuration.  
Example: FsBlockSize  
16K  
JournalSize  
0
The optimal settings are in the range between 16M and 64M. Avoid  
values greater than 64M due to potentially severe impacts on startup and  
failover times. Values at the higher end of the 16M-64M range may  
improve performance of metadata operations in some cases, although at  
the cost of slower startup and failover time. A good rule of thumb is to  
use 16M unless another requirement dictates differently. This setting is  
adjustable using the cvupdatefs utility. For more information, see the  
cvupdatefs man page.  
Example: JournalSize  
16M  
The snfsdefrag tool is very useful to identify and correct file extent  
fragmentation. Reducing extent fragmentation can be very beneficial for  
performance. You can use this utility to determine whether files are  
fragmented, and if so, fix them. If your files are prone to fragmentation  
you should also use the FSM config file tuning options to minimize  
fragmentation. These global configuration settings are InodeExpandMin,  
InodeExpandInc, and InodeExpandMax. (For more information, see the  
man cvfs_config page.) The snfsdefrag man page explains the command  
options in greater detail.  
SNFS Tools  
0
FSM hourly statistics reporting is another very useful tool. This can show  
you the mix of metadata operations being invoked by client processes, as  
well as latency information for metadata operations and metadata and  
journal I/O. This information is easily accessed in the cvlog log files. All  
of the latency oriented stats are reported in microsecond units.  
StorNext File System Tuning Guide  
12  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
It also possible to trigger an instant FSM statistics report by setting the  
Once Only debug flag using cvadmin. For example:  
cvadmin -F snfs1 -e ‘debug 0x01000000’ ; tail -100 /usr/cvfs/data/snfs1/log/cvlog  
The following items are a few things to watch out for:  
• A non-zero value for FSM wait SUMMARY journal waits indicates  
insufficient IOPS performance of the disks assigned to the metadata  
stripe group. This usually requires reducing the metadata I/O  
latency time by adjusting RAID cache settings or reducing  
bandwidth contention for the metadata LUN. Another possible  
solution is to add another metadata stripe group to the file system.  
This will improve metadata ops performance through I/O  
concurrency.  
• Non-zero value for FSM wait SUMMARY free buffer waits or low hit  
ratio for FSM cache SUMMARY buffer lookups indicates the FSM  
configuration setting BufferCacheSize is insufficient.  
• Non-zero value for FSM wait SUMMARY free inode waits or low hit  
ratio for FSM cache SUMMARY inode lookups indicates the FSM  
configuration setting InodeCacheSize is insufficient.  
• Large value for FSM threads SUMMARY max busy indicates the FSM  
configuration setting ThreadPoolSize is insufficient.  
• Extremely high values for FSM cache SUMMARY inode lookups, TKN  
SUMMARY TokenRequestV3, or TKN SUMMARY TokenReqAlloc might  
indicate excessive file fragmentation. If so, the snfsdefrag utility can  
be used to fix the fragmented files.  
• The VOP and TKN summary statistics indicate the count and avg/  
min/max microsecond latency for the various metadata operations.  
This shows what type of metadata operations are most prevalent and  
most costly. These are also broken out per client, which can be useful  
to identify a client that is disproportionately loading the FSM.  
SNFS supports the Windows Perfmon utility. This provides many useful  
statistics counters for the SNFS client component. To install, obtain a copy  
of cvfsperf.dll from the SCM team in Denver and copy it into the c:/  
winnt/system32 directory on the SNFS client system. Then run  
rmperfreg.exe and instperfreg.exe to set up the required registry settings.  
After these steps, the SNFS counters should be visible to the Windows  
Perfmon utility. If not, check the Windows Application Event log for  
errors.  
StorNext File System Tuning Guide  
13  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
The cvcp utility is a higher performance alternative to commands such as  
cp and tar. The cvcp utility achieves high performance by using threads,  
large I/O buffers, preallocation, stripe alignment, DMA I/O transfer, and  
Bulk Create. Also, the cvcp utility uses the SNFS External API for  
preallocation and stripe alignment. In the directory-to-directory copy  
mode (for example, cvcp source_dir destination_dir,) cvcp conditionally  
uses the Bulk Create API to provide a dramatic small file copy  
performance boost. However, it will not use Bulk Create in some  
scenarios, such as non-root invocation, managed file systems, quotas, or  
Windows security. Hopefully, these limitations will be removed in a  
future release. When Bulk Create is utilized, it significantly boosts  
performance by reducing the number of metadata operations issued. For  
example, up to 20 files can be created all with a single metadata  
operation. For more information, see the cvcp man page.  
The cvmkfile utility provides a command line tool to utilize valuable  
SNFS performance features. These features include preallocation, stripe  
alignment, and affinities. See the cvmkfile man page.  
The Lmdd utility is very useful to measure raw LUN performance as well  
as varied I/O transfer sizes.  
The cvdbset utility has a special “Perf” trace flag that is very useful to  
analyze I/O performance. For example: cvdbset perf  
Then, you can use cvdb -g to collect trace information such as this:  
PERF: Device Write 41 MB/s IOs 2 exts 1 offs 0x0 len 0x400000 mics 95589 ino  
0x5  
PERF: VFS Write EofDmaAlgn 41 MB/s offs 0x0 len 0x400000 mics 95618 ino  
0x5  
The “PERF: Device” trace shows throughput measured for the device I/  
O. It also shows the number of I/Os into which it was broken, and the  
number of extents (sequence of consecutive filesystem blocks).  
The “PERF: VFS” trace shows throughput measured for the read or write  
system call and significant aspects of the I/O, including:  
• Dma: DMA  
• Buf: Buffered  
• Eof: File extended  
• Algn: Well-formed DMA I/O  
• Shr: File is shared by another client  
• Rt: File is real time  
StorNext File System Tuning Guide  
14  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
• Zr: Hole in file was zeroed  
Both traces also report file offset, I/O size, latency (mics), and inode  
number.  
Sample use cases:  
• Verify that I/O properties are as expected.  
You can use the VFS trace to ensure that the displayed properties are  
consistent with expectations, such as being well formed; buffered  
versus DMA; shared/non-shared; or I/O size. If a small I/O is being  
performed DMA, performance will be poor. If DMA I/O is not well  
formed, it requires an extra data copy and may even be broken into  
small chunks. Zeroing holes in files has a performance impact.  
• Determine if metadata operations are impacting performance.  
If VFS throughput is inconsistent or significantly less than Device  
throughput, it might be caused by metadata operations. In that case,  
it would be useful to display “fsmtoken,” “fsmvnops,” and  
“fsmdmig” traces in addition to “perf.”  
• Identify disk performance issues.  
If Device throughput is inconsistent or less than expected, it might  
indicate a slow disk in a stripe group, or that RAID tuning is  
necessary.  
• Identify file fragmentation.  
If the extent count “exts” is high, it might indicate a fragmentation  
problem.This causes the device I/Os to be broken into smaller  
chunks, which can significantly impact throughput.  
• Identify read/modify/write condition.  
If buffered VFS writes are causing Device reads, it might be beneficial  
to match I/O request size to a multiple of the “cachebufsize” (default  
64KB; see mount_cvfs man page). Another way to avoid this is by  
truncating the file before writing.  
The cvadmin command includes a latency-test utility for measuring the  
latency between an FSM and one or more SNFS clients. This utility causes  
small messages to be exchanged between the FSM and clients as quickly  
as possible for a brief period of time, and reports the average time it took  
for each message to receive a response.  
StorNext File System Tuning Guide  
15  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Metadata Controller System  
The latency-test command has the following syntax:  
latency-test index-number [seconds]  
latency-test all [seconds]  
If an index-number is specified, the test is run between the currently-  
selected FSM and the specified client. (Client index numbers are  
displayed by the cvadmin who command). If all is specified, the test is  
run against each client in turn.  
The test is run for 2 seconds, unless a value for seconds is specified.  
Here is a sample run:  
snadmin (lsi) > latency-test  
Test started on client 1 (bigsky-node2)... latency 55us  
Test started on client 2 (k4)... latency 163us  
There is no rule-of-thumb for “good” or “bad” latency values. Latency  
can be affected by CPU load or SNFS load on either system, by unrelated  
Ethernet traffic, or other factors. However, for otherwise idle systems,  
differences in latency between different systems can indicate differences  
in hardware performance. (In the example above, the difference is a  
Gigabit Ethernet and faster CPU versus a 100BaseT Ethernet and a slower  
CPU.) Differences in latency over time for the same system can indicate  
new hardware problems, such as a network interface going bad.  
If a latency test has been run for a particular client, the cvadmin who  
long command includes the test results in its output, along with  
information about when the test was last run.  
The following SNFS mount command settings are explained in greater  
detail in the mount_cvfs man page.  
MountCommandOptions0  
By default, the size of the buffer cache is 32MB and each buffer is 64K, so  
there is a total of 512 buffers. In general, increasing the size of the buffer  
cache will not improve performance for streaming reads and writes.  
However, a large cache helps greatly in cases of multiple concurrent  
streams, and where files are being written and subsequently read. Buffer  
cache size is adjusted with the buffercachemin and buffercachemax  
settings.  
StorNext File System Tuning Guide  
16  
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
The Metadata Controller System  
The buffer cache I/O size is adjusted using the cachebufsize setting. The  
default setting is usually optimal; however, sometimes performance can  
be improved by increasing this setting to match the RAID5 stripe size.  
Unfortunately, this is often not possible on Linux due to kernel memory  
fragmentation. In this case performance may degrade severely because  
the full amount of buffer cache cannot be allocated. Using a large  
cachebufsize setting also decreases random I/O performance when the  
amount of data being read is smaller than the cache buffer size.  
Buffer cache read-ahead can be adjusted with the buffercache_readahead  
setting. When the system detects that a file is being read in its entirety,  
several buffer cache I/O daemons pre-fetch data from the file in the  
background for improved performance. The default setting is optimal in  
most scenarios.  
The buffer flusher can be tuned with the buffercache_iods setting. A  
single flusher daemon is responsible for flushing dirty buffers. Instead of  
synchronously writing each buffer, the flusher places the buffers in an I/  
O queue that is processed by multiple daemons. By default there are eight  
I/O daemons that simultaneously perform disk I/O. Provided that the  
system supports SCSI Command Tagged Queuing, this concurrency  
dramatically improves throughput.  
RAID systems such as the EMC CX series and Engenio that provide  
excellent concurrent small I/O performance can usually benefit from  
increasing the buffercache_iods setting.  
The auto_dma_read_length and auto_dma_write_length settings determine  
the minimum transfer size where direct DMA I/O is performed instead  
of using the buffer cache for well-formed I/O. These settings can be  
useful when performance degradation is observed for small DMA I/O  
sizes compared to buffer cache.  
For example, if buffer cache I/O throughput is 200 MB/sec but 512K  
DMA I/O size observes only 100MB/sec, it would be useful to determine  
which DMA I/O size matches the buffer cache performance and adjust  
auto_dma_read_length and auto_dma_write_length accordingly. The lmdd  
utility is handy here.  
The dircachesize option sets the size of the directory information cache on  
the client. This cache can dramatically improve the speed of readdir  
operations by reducing metadata network message traffic between the  
SNFS client and FSM. Increasing this value improves performance in  
scenarios where very large directories are not observing the benefit of the  
client directory cache.  
StorNext File System Tuning Guide  
17  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Distributed LAN (Disk Proxy) Networks  
The SNFS External API might be useful in some scenarios because it  
offers programmatic use of special SNFS performance capabilities such as  
affinities, preallocation, and quality of service. For more information, see  
the Quality of Service chapter of the StorNext User’s Guide API Guide.  
SNFS External API  
0
The Distributed LAN (Disk Proxy) Networks  
As with any client/server protocol, SNFS Distributed LAN performance  
is subject to the limitations of the underlying network. Therefore, it is  
strongly recommended that you use dedicated networks for Distributed  
LAN traffic, to avoid contention with other network traffic. Gigabit  
(1000BaseT) Ethernet is recommended. Neither TCP offload nor jumbo  
frames are required.  
Dedicated networks can help provide the most predictable, deterministic  
level of performance. However, if shared network service is required  
(perhaps for peak NIC bandwidth utilization,) network contention could  
result. Network contention can lead to poor or inconsistent Distributed  
LAN performance, poor or inconsistent performance of other distributed  
applications, and, depending on their resiliency, intermittent failures of  
those applications.  
Hardware Configuration 0  
SNFS Distributed LAN can easily fill several Gigabit Ethernets with data,  
so take special care when selecting and configuring the switches used to  
interconnect SNFS Distributed LAN clients and servers. Ensure that your  
network switches have enough internal bandwidth to handle all of the  
anticipated traffic between all Distributed LAN clients and servers  
connected to them.  
A network switch that is dropping packets will cause TCP  
retransmissions. This can be easily observed on both Linux and Windows  
platforms by using the netstat -s command while proxy I/O is in  
progress.  
Reducing the TCP window size used by Distributed LAN might also help  
with an oversubscribed network switch. The Windows client Disk Proxy  
tab and the Linux dpserver file contain the tuning parameter for the TCP  
window size. Note that Distributed LAN server remounts are required  
after changing this parameter.  
StorNext File System Tuning Guide  
18  
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
The Distributed LAN (Disk Proxy) Networks  
Within each Distributed LAN network, it is best practice to have all SNFS  
Distributed LAN clients and servers directly attached to the same  
network switch. A router between a Distributed LAN client and server  
could be easily overwhelmed by the data rates required.  
It is critical to ensure that speed/duplex settings are correct, as this will  
severely impact performance. Most of the time auto-detect is the correct  
setting. Some managed switches allow setting speed/duplex, such as  
1000Mb/full, which disables auto-detect and requires the host to be set  
exactly the same. However, performance is severely impacted if the  
settings do not match between switch and host. For example, if the switch  
is set to auto-detect but the host is set to 1000Mb/full, you will observe a  
high error rate and extremely poor performance. On Linux the mii-diag  
tool can be very useful to investigate and adjust speed/duplex settings.  
In some cases, TCP offload seems to cause problems with Distributed  
LAN by miscalculating checksums under heavy loads. This is indicated  
by bad segments indicated in the output of netstat -s. On Linux, the TCP  
offload state can be queried by running ethtool -k, and modified by  
running ethtool -K. On Windows it is configured through the Advanced  
tab of the configuration properties for a network interface.  
The internal bus bandwidth of a Distributed LAN client or server can also  
place a limit on performance. A basic PCI- or PCI-X-based workstation  
might not have enough bus bandwidth to run multiple Gigabit Ethernet  
NICs at full speed; PCI Express is recommended but not required.  
Similarly, the performance characteristics of NICs can vary widely and  
ultimately limit the performance of Distributed LAN. For example, some  
NICs might be able to transmit or receive each packet at Gigabit speeds,  
but not be able to sustain the maximum needed packet rate. An  
inexpensive 32-bit NIC plugged into a 64-bit PCI-X slot is incapable of  
fully utilizing the host's bus bandwidth.  
It can be useful to use a tool like netperf to help verify the performance  
characteristics of each Distributed LAN network. (When using netperf,  
take care to specify the right IP addresses in order to ensure the network  
being tested is a Distributed LAN network, not the SNFS dedicated  
Metadata Network or another network.) For example, if netperf -t  
TCP_RR reports less than 15,000 transactions per second capacity, a  
performance penalty might be incurred. Multiple copies of netperf can  
also be run in parallel to determine the performance characteristics of  
multiple NICs.  
StorNext File System Tuning Guide  
19  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
The Distributed LAN (Disk Proxy) Networks  
Network Configuration  
and Topology  
A common source of difficult-to-diagnose issues with SNFS is improper  
IP network configuration. Many incorrect IP configurations might appear  
to work when tested with particular applications or particular kinds of  
hosts, but fail when used with SNFS or when a different kind of host is  
added to the cluster.  
0
SNFS Distributed LAN should be run over one or more dedicated IP  
subnetworks, using dedicated NICs for each subnetwork. These  
subnetworks are typically configured using site-local addresses, such as  
10.a.b.c or 192.168.x.y.  
For example, if SNFS Distributed LAN is set up to use two Gigabit  
Ethernet NICs (for a total bandwidth of 2 Gbits/s for Distributed LAN  
I/O), each Distributed LAN Server and Distributed LAN Client should  
have at least 4 NICs: 2 dedicated NICs for Distributed LAN, 1 dedicated  
NIC for SNFS Metadata traffic, and 1 NIC for administrative access and  
other traffic. Each of the NICs must be configured for a separate IP  
subnet. So for example, if the subnets chosen for Distributed LAN are  
192.168.1.0/24 and 192.168.2.0/24, each Distributed LAN Server and  
each Distributed LAN Client must have one (and only one) NIC with an  
IP address of 192.168.1.x, plus one (and only one) NIC with an IP  
address of 192.168.2.y.  
It is best practice to have all of the NICs for the same Distributed LAN  
subnetwork directly connected to the same network switch. A router  
between a Distributed LAN client and server could be easily  
overwhelmed by the data rates required.  
By contrast, Quantum recommends having the NICs for different  
Distributed LAN subnetworks connected to different network switches, to  
avoid overwhelming any one network switch.  
All Distributed LAN subnetworks must be completely connected. This  
typically means all Distributed LAN clients and all Distributed LAN  
servers have dedicated NICs for each of the Distributed LAN  
subnetworks. If a Distributed LAN client does not have a dedicated NIC  
for a particular Distributed LAN subnetwork, it must be configured with  
static routes to give it connectivity to each of the IP addresses advertised  
by the Distributed LAN servers. Note that if a Distributed LAN client is  
unable to connect to a particular server’s NIC, it will continue forever  
trying to connect to it, potentially causing slow file system mounts,  
repeated messages in error logs, and other problems.  
StorNext File System Tuning Guide  
20  
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
The Distributed LAN (Disk Proxy) Networks  
Figure 1 Multi-NIC Hardware  
and IP Configuration Diagram  
10.0.0.42  
Other  
Host  
10.0.0.1  
SNFS  
MDC  
172.16.47.1  
10.0.0.12  
SNFS  
SAN  
Client  
172.16.47.12  
10.0.0.27  
172.16.47.27  
192.168.1.27  
192.168.2.27  
Public  
Net  
Switch  
Distributed  
LAN  
Server  
A
SAN  
Meta-  
Data  
Switch  
10.0.0.28  
172.16.47.28  
192.168.1.28  
192.168.2.8  
Distributed  
LAN  
Server  
B
Dist-  
LAN 1  
Switch  
10.0.0.91  
172.16.47.91  
192.168.1.91  
192.168.2.91  
Distributed  
LAN  
Client  
X
Dist-  
LAN 2  
Switch  
10.0.0.92  
172.16.47.92  
192.168.1.92  
192.168.2.92  
Distributed  
LAN  
Client  
Y
StorNext File System Tuning Guide  
21  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
Distributed LAN Servers  
Distributed LAN Servers  
Distributed LAN Servers must have sufficient memory. When a  
Distributed LAN Server does not have sufficient memory, its  
performance in servicing Distributed LAN I/O requests might suffer. In  
some cases (particularly on Windows,) it might hang.  
Refer to the StorNext Release Notes for this release’s memory  
requirements.  
Distributed LAN Servers must also have sufficient bus bandwidth. As  
discussed above, a Distributed LAN Server must have sufficient bus  
bandwidth to operate the NICs used for Distributed LAN I/O at full  
speed, while at the same time operating their Fibre Channel HBAs. Thus,  
Quantum strong recommends using PCI Express for Distributed LAN  
Servers.  
Note that a single slow Distributed LAN server can reduce overall  
throughput for all clients. The slow server will get some of the traffic  
from all clients and tend to reduce throughput across the board. An  
example would be a server with less disk bandwidth than the other  
servers, or a server with a slow network connection.  
Windows Memory Requirements  
Beginning in version 2.6.1, StorNext includes a number of performance  
enhancements that enable it to better react to changing customer load.  
However, these enhancements come with a price: memory requirement.  
When running in the Windows environment on a machine with  
minimum memory (256 MB) the tuning parameters need to be adjusted  
or the machine will run out of memory. This can be seen by bringing up  
task manager and watching the Nonpaged tag in the Kernel Memory pane  
in the lower right hand corner. On a machine with 256 MB of memory  
this value must be less than 96 MB.  
The problem will manifest itself by commands failing, messages being  
sent to the system log about insufficient memory, the fsmpm  
StorNext File System Tuning Guide  
22  
Download from Www.Somanuals.com. All Manuals Search And Download.  
   
StorNext File System Tuning  
Windows Memory Requirements  
mysteriously dying, repeated FSM reconnect attempts, and messages  
being sent to the application log and cvlog.txt about socket failures with  
the status code (10555) which is ENOBUFS.  
The solution is to adjust a few parameters on the Cache Parameters tab in  
the SNFS control panel (cvntclnt). These parameters control how much  
memory is consumed by the directory cache, the buffer cache, and the  
local file cache.  
As always, an understanding of the customers’ workload aids in  
determining the correct values. Tuning is not an exact science, and  
requires some trial-and-error (and the unfortunate reboots) to come up  
with values that work best in the customer’s environment.  
The first is the Directory Cache Size. The default is 10 (MB). If you do not  
have large directories, or do not perform lots of directory scans, this  
number can be reduced to 1 or 2 MB. The impact will be slightly slower  
directory lookups in directories that are frequently accessed.  
Also, in the Mount Option panel, you should set the Paged DirCache  
option.  
The next parameter is the Buffer Cache NonPaged Pool Usage. This value  
is in percent (%) and represents the percentage of available non-paged  
pool that the buffer cache will consume. By default, this value is 75%.  
This should be set to 25 or at most 50. The minimum value is 10 and the  
maximum value is 90.  
The next parameters control how many file structures are cached on the  
client. These are controlled by the Meta-data Cache low water mark, Meta-  
data Cache high water mark and Meta-data Cache Max water mark. Each file  
structure is represented internally by a data structure called the  
“cvnode.” The cvnode represents all the state about a file or directory.  
The more cvnodes that there are encached on the client, the fewer trips  
the client has to make over the wire to contact the FSM.  
Each cvnode is approximately 1462 bytes in size and is allocated from the  
non-paged pool. The cvnode cache is periodically purged so that unused  
entries are freed. The decision to purge the cache is made based on the  
Low, High, and Max water mark values. The 'Low' default is 1024, the  
'High' default is 3072, and the 'Max' default is 4096.  
These values should be adjusted so that the cache does not bloat and  
consume more memory than it should. These values are highly  
dependent on the customers work load and access patterns. Values of 512  
for the High water mark will cause the cvnode cache to be purged when  
more than 512 entries are present. The cache will be purged until the low  
StorNext File System Tuning Guide  
23  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
Windows Memory Requirements  
water mark is reached, for example 128. The Max water mark is for  
situations where memory is very tight. The normal purge algorithms  
takes access time into account when determining a candidate to evict  
from the cache; in tight memory situations (when there are more than  
'max' entries in the cache), these constraints are relaxed so that memory  
can be released. A value of 1024 in a tight memory situation should work.  
StorNext File System Tuning Guide  
24  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
Sample FSM Configuration File  
Sample FSM Configuration File  
This sample configuration file is located in the SNFS install directory  
under the examples subdirectory named example.cfg.  
#****************************************************************************  
# A global section for defining file system-wide parameters.  
#
# For Explanations of Values in this file see the following:  
#
# UNIX Users:  
man cvfs_config  
# Windows Users: Start > Programs > StorNext File System > Help >  
# Configuration File Internal Format  
#
#****************************************************************************  
GlobalSuperUser  
Systems ##  
Yes  
## Must be set to Yes for SNMS Managed File  
WindowsSecurity  
Quotas  
Yes  
No  
No  
No  
32K  
128K  
8M  
16K  
16M  
Round  
32  
FileLocks  
DataMigration  
InodeExpandMin  
InodeExpandInc  
InodeExpandMax  
FsBlockSize  
## SNMS Managed File Systems Only ##  
JournalSize  
AllocationStrategy  
MaxConnections  
ForceStripeAlignment  
ThreadPoolSize  
InodeCacheSize  
Yes  
32  
16K  
# default 16, 512 KB memory per thread  
# 800-1000 bytes each, default 8K  
Debug  
MaxLogSize  
MaxLogs  
0x0  
16M  
4
#
# Globals Defaulted  
#
# BufferCacheSize  
# StripeAlignSize  
64M  
2M  
# default 32MB  
# auto alignment threshold, default  
StorNext File System Tuning Guide  
25  
Download from Www.Somanuals.com. All Manuals Search And Download.  
 
StorNext File System Tuning  
Sample FSM Configuration File  
MAX(StripeBreadth)  
# MaxMBPerClientReserve  
# OpHangLimitSecs  
50  
# in MBs, default 100 MB reserved per client  
# default 180 secs  
300  
# DataMigrationThreadPoolSize 128  
# Managed only, default 8  
#****************************************************************************  
# A disktype section for defining disk hardware parameters.  
#****************************************************************************  
[DiskType MetaDrive] ##1+1 Raid 1 Mirrored Pair##  
Sectors XXXXXXXX  
SectorSize 512  
## Sectors Per Disk From Command "cvlabel -l" ##  
[DiskType JournalDrive] ##1+1 Raid 1 Mirrored Pair##  
Sectors XXXXXXXX  
SectorSize 512  
[DiskType VideoDrive] ##8+1 Raid 5 Lun for Video##  
Sectors XXXXXXXX  
SectorSize 512  
[DiskType AudioDrive] ##4+1 Raid 3 Lun for Audio##  
Sectors XXXXXXXX  
SectorSize 512  
[DiskType DataDrive] ##4+1 Raid 5 Lun for Regular Data##  
Sectors XXXXXXXX  
SectorSize 512  
#****************************************************************************  
# A disk section for defining disks in the hardware configuration.  
#****************************************************************************  
[Disk CvfsDisk0]  
Status UP  
Type MetaDrive  
[Disk CvfsDisk1]  
Status UP  
Type JournalDrive  
[Disk CvfsDisk2]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk3]  
Status UP  
Type VideoDrive  
StorNext File System Tuning Guide  
26  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
Sample FSM Configuration File  
[Disk CvfsDisk4]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk5]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk6]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk7]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk8]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk9]  
Status UP  
Type VideoDrive  
[Disk CvfsDisk10]  
Status UP  
Type AudioDrive  
[Disk CvfsDisk11]  
Status UP  
Type AudioDrive  
[Disk CvfsDisk12]  
Status UP  
Type AudioDrive  
[Disk CvfsDisk13]  
Status UP  
Type AudioDrive  
[Disk CvfsDisk14]  
Status UP  
Type DataDrive  
[Disk CvfsDisk15]  
Status UP  
Type DataDrive  
[Disk CvfsDisk16]  
Status UP  
Type DataDrive  
StorNext File System Tuning Guide  
27  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
Sample FSM Configuration File  
[Disk CvfsDisk17]  
Status UP  
Type DataDrive  
#****************************************************************************  
# A stripe section for defining stripe groups.  
#****************************************************************************  
[StripeGroup MetaFiles]  
Status UP  
MetaData Yes  
Journal No  
Exclusive Yes  
Read Enabled  
Write Enabled  
StripeBreadth 256K  
MultiPathMethod Rotate  
Node CvfsDisk0 0  
[StripeGroup JournFiles]  
Status UP  
Journal Yes  
MetaData No  
Exclusive Yes  
Read Enabled  
Write Enabled  
StripeBreadth 256K  
MultiPathMethod Rotate  
Node CvfsDisk1 0  
[StripeGroup VideoFiles]  
Status UP  
Exclusive Yes  
##Exclusive StripeGroup for Video Files Only##  
Affinity VideoFiles  
Read Enabled  
Write Enabled  
StripeBreadth 4M  
MultiPathMethod Rotate  
Node CvfsDisk2 0  
Node CvfsDisk3 1  
Node CvfsDisk4 2  
Node CvfsDisk5 3  
Node CvfsDisk6 4  
Node CvfsDisk7 5  
Node CvfsDisk8 6  
Node CvfsDisk9 7  
StorNext File System Tuning Guide  
28  
Download from Www.Somanuals.com. All Manuals Search And Download.  
StorNext File System Tuning  
Sample FSM Configuration File  
[StripeGroup AudioFiles]  
Status UP  
Exclusive Yes  
##Exclusive StripeGroup for Audio File Only##  
Affinity AudioFiles  
Read Enabled  
Write Enabled  
StripeBreadth 1M  
MultiPathMethod Rotate  
Node CvfsDisk10 0  
Node CvfsDisk11 1  
Node CvfsDisk12 2  
Node CvfsDisk13 3  
[StripeGroup RegularFiles]  
Status UP  
Exclusive No  
##Non-Exclusive StripeGroup for all Files##  
Read Enabled  
Write Enabled  
StripeBreadth 256K  
MultiPathMethod Rotate  
Node CvfsDisk14 0  
Node CvfsDisk15 1  
Node CvfsDisk16 2  
Node CvfsDisk17 3  
StorNext File System Tuning Guide  
29  
Download from Www.Somanuals.com. All Manuals Search And Download.  

Q Logic Computer Hardware QLA2300 User Guide
Quadra Fire Indoor Fireplace QVI 30FB S User Guide
ResMed Sleep Apnea Machine Mirage User Guide
Rollei Digital Camera 24611 User Guide
Samsung MP3 Player YP 910 User Guide
Sansui Flat Panel Television HDLCD185W User Guide
Sansui Flat Panel Television SLED4650 User Guide
Savin All in One Printer 9940DP User Guide
Seagate Computer Hardware ST19171FC User Guide
Sennheiser Headphones 502399 User Guide