ZFS on HW RAID with poor performance. Ideas?

Hi all,
I'm trying to figure out why my RAID performs so poorly using ZFS compared to even UFS. Any thoughts on the following would be helpful.
H/W Config:_
- SunFire v245 w/ 4GB RAM and LSI1064 mirrored 73GB SAS
- QLogic QLE2462
- Fibre Channel 4Gb Hardware RAID w/ 1GB ECC cache
- 16 x 1TB SATA II drives
- Configured 1 volume RAID-6
S/W Config:_
- Solaris 08/07 SPARC - 5.10 Generic_120011-14
- Latest 10_Recommended as of 1/08
- zpool create fileserv /dev/dsk/c3t0d0s6
- zfs set atime=off fileserv
Test Raw:_
To the "RAW" disk or volume I get descent numbers:
# dd if=/dev/zero of=/dev/rdsk/c3t0d0s6 bs=1048576 count=16384
# dd if=/dev/rdsk/c3t0d0s6 of=/dev/null bs=1048576 count=16384
Both c3t0d0 and c3t0d0s6 result in similar numbers:
5 x 16GB writes took 475.50 seconds, yielding an average of *172.28 MB/s*.
5 x 16GB reads took 384.68 seconds, yielding an average of *212.97 MB/s*.
Test ZFS:_
Creating and mounting ZFS to the same volume, I modified my scripts to use dd created files from /dev/null instead:
# dd if=/dev/zero of=./16G bs=16777216 count=4096
Then ran:
# dd if=/dev/zero of=./16G bs=1048576 count=16384
# dd if=./16G of=/dev/null bs=1048576
The result for ZFS on a 14.6TB volume is horrible:
5 x 16GB writes took 1013.28 seconds, yielding an average of *80.84 MB/s*.
5 x 16GB reads took 980.36 seconds, yielding an average of *83.56 MB/s*.
What would cause a delta difference of more than half my capable speed from the unit just by running ZFS? Thinking it might just be a dd issue, I ran iozone using the same parameters as "milek's blog" at http://milek.blogspot.com and got the same crappy results:
# iozone -ReM -b iozone-2G.wks -r 128k -s 2g -t 32 -F /test/*
(snippet:)
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
Throughput test with 32 processes
Each process writes a 2097152 Kbyte file in 128 Kbyte records
Children see throughput for 32 initial writers = 90584.21 KB/sec
Parent sees throughput for 32 initial writers = 79823.98 KB/sec
Min throughput per process = 2186.76 KB/sec
Max throughput per process = 3225.63 KB/sec
Avg throughput per process = 2830.76 KB/sec
Min xfer = 1427456.00 KB
Children see throughput for 32 rewriters = 88526.23 KB/sec
Parent sees throughput for 32 rewriters = 88380.95 KB/sec
Min throughput per process = 2264.00 KB/sec
Max throughput per process = 3152.58 KB/sec
Avg throughput per process = 2766.44 KB/sec
Min xfer = 1510656.00 KB
I can understand that ZFS wants to have the entire "disk" as in a JBOD, but that's just not a viable option in my Corporate America world. Does presenting a RAID Controller handled LUN Vs. a JBOD presented matrix of drives really make this much of a difference? I'm just at a loss here as to why a unit that is capable of 4Gbit per channel is working as it should via raw, but runs like a single disk in performance as a ZFS filesystem.
I would appreciate some assistance in figuring this one out.
Thanks to all in advance.
--Tom

Robert,
Thanks for the help. Unfortunately, using RAID-Z or Z2 is not an option here as I'd rather rely on my hardware RAID over using software. While doing testing, I set my RAID to pass-thru mode and allowed ZFS to inherit the RAW drives from the RAID Chassis, set up a pool, and did my testing. In all cases, write performance was not much better overall, CPU utilization with compression on was ~30% higher, and my RAM utilization shot up under load. I just don't see that as an option on a heavily hit CVS server where IOPS and CPU/RAM are important for compilations and the like.
Things Hardware RAID gives me that ZFS can't do are:
Ease of use - changing failed drives for example is just different. It's a complicated filesystem once you get into it.
Pre-write buffer performance - This is ZFS's bottleneck afterall.
Fastest I/O - RAID Controllers handle more drives better than ZFS and IOPS are much faster under a H/W RAIDThere's a whole line of reasons why I choose not to use ZFS RAID-Z/Z2. Why else would Sun continue selling both H/W RAID and JBOD? I will check into the Intent Logging though, that's a good idea. I appreciate the link to solarisinternals, I had not thought about using their wiki for this.
Thanks again!
--Tom

Similar Messages

Are there Issues with poor performance with Maverick OS,

Are there issues with slow and reduced performance with Mavericks OS

check this
http://apple.stackexchange.com/questions/126081/10-9-2-slows-down-processes
or
this:
https://discussions.apple.com/message/25341556#25341556
I am doing a lot of analyses with 10.9.2 on a late 2013 MBP, and these analyses generally last hours or days. I observed that Maverick is slowing them down considerably for some reasons after few hours of computations, making it impossible for me to work with this computer...

Poor performance of VL10A transaction

Hello Experts,
We are facing very poor performance in One of our production system in VL10A transaction
I checked for the response time and it is going up to 5 seconds
I checked for :
Table Space > it is fine
Pending updates ? it is fine
Available workprocesses > it is fine
Abap Dumps > no major issues.
Could any body suggest ??
Thank you
Regards,
Amber S

Hi
Try to get a working example of the transaction with poor performance. I would just run the transaction and monitor the work processes SM50 at same time. If you always see your request on a certain table, for example VEPVG, then I would take an SQL trace with transaction ST05.
If the process in SM50 is only on a table (-> accessing the database) for a short time, then you would investigate more on the Abap side.
Btw, if you did not modifiy anything, then you can always go to SAP support (-> look for SAP notes first!) as this is a standard functionality.
Best regards, Michael

Poor performance of Report Writer reports (Special Ledger Library)

Greetings - We are running into problems with poor performance of reports that are written with the SAP Report Writer. The problem appears to be caused when SAP is using the primary-key index in our Special Purpose ledger (where the reports are generated). The index contains object fields that cannot be added to the report library (COBJNR, SOBJNR, ROBJNR). We have created alternate indices, but they are not being picked up with the Report Writer reports.
Are there any configurable or technical settings that we can work with in order to force the use of a specific index for a report? It seems logical that SAP would find the most efficient index to use, but with the reports that we are looking at, this does not appear to be the case.
Any help that can be offered will be greatly appreciated...We are currently using version 4.6C, but are planning an upgrade to ECC 6.0 later this year.
Thanks in advance -

Arjun,
Where / which files contains these parameters we cannot find them all ???
Tomcat - Java _properties and try again ( You can tune below value as per your system memory)
-XX:PermSize=256m
-XX:MaxPermSize=256m
-XX:NewSize=171m
-XX:MaxNewSize=171m
-XX:SurvivorRatio=2
-XX:TargetSurvivorRatio=90
-XX:+DisableExplicitGC
-XX:+UseTLAB
As a general update it looks like we need to use the Monitoring tools that are installed by default, we are now in the process of installing the database etc
Cheers

Poor performance of facetime on macbook air while it works fine with other devices

Hi all,
I am experiencing poor performance when using facetime with my macbook air (image is poor and voice keeps breaking) while facetime works absolutely fine with my iphone or ipad. Skype works fie with the macbook as well. I have fiber optic connection with plenty of bandwidth. Any ideas?
Dimitri

Reboot and try again.

Poor Performance with 10.1

I'm on a windows xp sp 3 machine with an x800 xt pe graphics card using the latest catalyst drivers from Ati. Ever since I updated flash, I'm getting very poor performance during 720p with acceleration enabled. I'm also getting bad performance when it's disabled. Is there an issue with the 10.2 legacy (ati) drivers not enabling hardware acceleration with 10.1 flash?
BTW I went back to using 10.0.42.34 flash and everything is running fine again.

I've also noticed poor performance since installing the 10.1 plugin. Not just in HD video etc, but in general use. So much for "improved performance"...
No idea on solutions at this stage - In the process of downgrading to previous version to see if that fixes my problems

Zfs/nfs poor performance compared to zfs/smb

I setup zfs with sharenfs on my solaris box and i'm having a very poor performance with nfs (about 1-3MB/s).
When i share the same zfs volume with sharesmb performance is very good (about 40-50MB/s) for a single disk zfs volume on a gigabit network.
Both tests where made using MAC OSX 10.5.2 client using a single iso file transfert (about 4Gigs).
I also mount through nfs my MAC client (using apple nfs share) from the solaris box and copy the file to the ZFS volume, nfs performance was very good in this case (45-50MB/s).
Why my solaris zfs/nfs sharing is so slow compared to mac nfs sharing or solaris zfs/smb sharing ?

I'm guess your talking about write performance ie your copying the ISO from a local drive to the ZFS/NFS drive.
The write performance of NFS/ZFS has been known to suck for a long time.
First make sure your running the latest solaris version (10U4) or a recent recommended patch set.
If not, upgrade and see if that helps.
Unfortunately, the solaris developers have a (some say overly) strict interpretation of the coherence requirements of the NFS spec and what it implies about caching data in memory and flushing to disk.
This is discussed a little here
and
here
Their position (as I understand it) that their implemation provides the best performance that can be obtained while maintaining adherence with the relevant standards and the fact that every other NFS implementation has streets better performance just means that everyone else is non conforming.
Anyway maybe can try the solution that the "best practices" suggests ie using a local drive as an intent log.
If that doesnt work, try setting zil_disable (�set zfs:zil_disable=1′ to /etc/system).
You will have to weigh the significance of the purported reliability issues of that solution for yourself.
Basically what it comes down to is that if your nfs server crashes and the client doesnt, but keeps cheerfully writing away. Then some of what it wrote maybe lost. Even silently a part in the middle.
One other thing to check, the storage isn't coming off a SAN is it?
If it is you can try "set zfs:zfs_nocacheflush=1" instead of zil_disable. Its a less aggressive variant that can reduce issues with san storage.

Poor performance when dragging item within a list loaded with images - Flex 4

Hi,
I have a custom built List component that is using a TileLayout. I am using a custom itemRenderer to load images into this list from our server (in this test, 44 images rae loaded). I have enabled dragEnabled and dragMove so that I can move items around within the list. The problem comes when I start dragging an item. The dragging operation is very slow and clunky.
When I move the mouse to drag the item, the dropIndicator does not get refreshed for a few seconds and the movement feels like my PC is lagging pretty badly. I've also noticed that during this time my CPU usage is spiking up to around 25-40%. The funny part is I can scroll the list just fine without any lag at all, but the minute I drag the application starts to lag really bad. I do have some custom dragOver code that I used to override the dragOverHandler of the list control, but the problem persists even if I take that code out. I've tried with useVirtualLayout set to both true and false and neither setting made a difference.
Any ideas as to what could be causing the poor performance and/or how I can go abouts fixing it?
Thanks a lot in advance!

Ahh, good call about the Performance profiler. I'm pretty new to the profiling thing with Flex (haven't used Builder Pro before
the Flex 4 beta) so please forgive me
I found some interesting things running the performance profiler but I'm not sure I understand what to make of it all. I cleared the Performance Profile data when right before I loaded the images into the list. I then moved some images around and then captured the Profiling Data (If I understand Adobe correctly, this is the correct way to capture performance information for a set of actions).
What I found is there is a [mouseEvent] item that took 3101ms with 1 "Calls" (!!!!). When I drill down into that item to see the Method Statistics, I actually see three different Callees and no callers. The sum of the time it took the Callees to execute does not even come close to adding up to the 3101 ms (about 40ms). I'm not sure what I can make of those numbers, or if they are even meaningful. Any insight into these?
The only other items that stand out to me are [pre-render] which has 863ms (Cumulative Time) / 639ms (Self Time), [enterFrameEvent] which has 746ms / 6ms (?!), and [tincan] (what the heck is tincan?) which has times of 521ms for both Cumulative and Self.
Can anyone offer some insight into these numbers and maybe make some more suggestions for me? I apologize for my ignorance on these numbers - as I said, I'm new to the whole Flex profiling thing.
Many thanks in advance!
Edit: I just did another check, this time profiling only from the start of my drag to the end of my drop, and I still see [mouseEvent] taking almost 1000ms of Cumulative Time. However, when I double click that item to see the Method Statistics, no Callers or Callees are listed. What's causing this [mouseEvent] and how come it's taking so long?

Poor performance with WebI and BW hierarchy drill-down...

Hi
We are currently implementing a large HR solution with BW as backend
and WebI and Xcelcius as frontend. As part of this we are experiencing
very poor performance when doing drill-down in WebI on a BW hierarchy.
In general we are experiencing ok performance during selection of data
and traditional WebI filtering - however when using the BW hierarchy
for navigation within WebI, response times are significantly increasing.
The general solution setup are as follows:
1) Business Content version of the personnel administration
infoprovider - 0PA_C01. The Infoprovider contains 30.000 records
2) Multiprovider to act as semantic Data Mart layer in BW.
3) Bex Query to act as Data Mart Query and metadata exchange for BOE.
All key figure restrictions and calculations are done in this Data Mart
Query.
4) Traditionel BO OLAP universe 1:1 mapped to Bex Data Mart query. No
calculations etc. are done in the universe.
5) WebI report with limited objects included in the WebI query.
As we are aware that performance is an very subjective issues we have
created several case scenarios with different dataset sizes, various
filter criteria's and modeling techniques in BW.
Furthermore we have tried to apply various traditional BW performance
tuning techniques including aggregates, physical partitioning and pre-
calculation - all without any luck (pre-calculation doesn't seem to
work at all as WebI apparently isn't using the BW OLAP cache).
In general the best result we can get is with a completely stripped WebI report without any variables etc.
and a total dataset of 1000 records transferred to WebI. Even in this scenario we can't get
each navigational step (when using drill-down on Organizational Unit
hierarchy - 0ORGUNIT) to perform faster than minimum 15-20 seconds per.
navigational step.
That is each navigational step takes 15-20 seconds
with only 1000 records in the WebI cache when using drill-down on org.
unit hierachy !!.
Running the same Bex query from Bex Analyzer with a full dataset of
30.000 records on lowest level of detail returns a threshold of 1-2
seconds pr. navigational step thus eliminating that this should be a BW
modeling issue.
As our productive scenario obviously involves a far larger dataset as
well as separate data from CATS and PT infoproviders we are very
worried if we will ever be able to utilize hierarchy drill-down from
WebI ?.
The question is as such if there are any known performance issues
related to the use of BW hierarchy drill-down from WebI and if so are
there any ways to get around them ?.
As an alternative we are currently considering changing our reporting
strategy by creating several higher aggregated reports to avoid
hierarchy navigation at all. However we still need to support specific
division and their need to navigate the WebI dataset without
limitations which makes this issue critical.
Hope that you are able to help.
Thanks in advance
/Frank
Edited by: Mads Frank on Feb 1, 2010 9:41 PM

Hi Henry, thank you for your suggestions although i´m not agree with you that 20 seconds is pretty good for that navigation step. The same query executed with BEx Analyzer takes only 1-2 seconds to do the drill down.
Actions
suppress unassigned nodes in RSH1: Magic!! This was the main problem!!
tick use structure elements in RSRT: Done it.
enable query stripping in WebI: Done it.
upgrade your BW to SP09: Has the SP09 some inprovements in relation to this point ?
use more runtime query filters. : Not possible. Very simple query.
Others:
RSRT combination H-1-3-3-1 (Expand nodes/Permanent Cache BLOB)
Uncheck prelimirary Hierarchy presentation in Query. only selected.
Check "Use query drill" in webi properties.
Sorry for this mixed message but when i was answering i tryied what you suggest in relation with supress unassigned nodes and it works perfectly. This is what is cusing the bottleneck!! incredible...
Thanks a lot
J.Casas

Poor performance with Oracle Spatial when spatial query invoked remotely

Is anyone aware of any problems with Oracle Spatial (10.2.0.4 with patches 6989483 and 7003151 on Red Hat Linux 4) which might explain why a spatial query (SDO_WITHIN_DISTANCE) would perform 20 times worse when it was invoked remotely from another computer (using SQLplus) vs. invoking the very same query from the database server itself (also using SQLplus)?
Does Oracle Spatial have any known problems with servers which use SAN disk storage? That is the primary difference between a server in which I see this poor performance and another server where the performance is fine.
Thank you in advance for any thoughts you might share.

OK, that's clearer.
Are you sure it is the SQL inside the procedure that is causing the problem? To check, try extracting the SQL from inside the procedure and run it in SQLPLUS with
set autotrace on
set timing on
SELECT ....If the plans and performance are the same then it may be something inside the procedure itself.
Have you profiled the procedure? Here is an example of how to do it:
Prompt Firstly, create PL/SQL profiler table
@$ORACLE_HOME/rdbms/admin/proftab.sql
Prompt Secondly, use the profiler to gather stats on execution characteristics
DECLARE
l_run_num PLS_INTEGER := 1;
l_max_num PLS_INTEGER := 1;
v_geom    mdsys.sdo_geometry := mdsys.sdo_geometry(2002,null,null,sdo_elem_info_array(1,2,1),sdo_ordinate_array(0,0,45,45,90,0,135,45,180,0,180,-45,45,-45,0,0));
BEGIN
dbms_output.put_line('Start Profiler Result = ' || DBMS_PROFILER.START_PROFILER(run_comment => 'PARALLEL PROFILE')); -- The comment name can be anything: here it is related to the Parallel procedure I am testing.
v_geom := Parallel(v_geom,10,0.05,1); -- Put your procedure call here
dbms_output.put_line('Stop Profiler Result = ' || DBMS_PROFILER.STOP_PROFILER );
END;
SHOW ERRORS
Prompt Finally, report activity
COLUMN runid FORMAT 99999
COLUMN run_comment FORMAT A40
SELECT runid || ',' || run_date || ',' || run_comment || ',' || run_total_time
FROM plsql_profiler_runs
ORDER BY runid;
COLUMN runid       FORMAT 99999
COLUMN unit_number FORMAT 99999
COLUMN unit_type   FORMAT A20
COLUMN unit_owner FORMAT A20
COLUMN text        FORMAT A100
compute sum label 'Total_Time' of total_time on runid
break on runid skip 1
set linesize 200
SELECT u.runid || ',' ||
       u.unit_name,
       d.line#,
       d.total_occur,
       d.total_time,
       text
FROM   plsql_profiler_units u
       JOIN plsql_profiler_data d ON u.runid = d.runid
                                     AND
                                     u.unit_number = d.unit_number
       JOIN all_source als ON ( als.owner = 'CODESYS'
                             AND als.type = u.unit_type
                               AND als.name = u.unit_name
                            AND als.line = d.line# )
WHERE u.runid = (SELECT max(runid) FROM plsql_profiler_runs)
ORDER BY d.total_time desc;Run the profiler in both environments and see if you can see where the slowdown exists.
regards
Simon

Poor Performance of Safari with YouTube

Recently I have noticed really poor performance of Safari when I try to watch videoes on YouTube.
I have experienced this both with my 2009 iMac desktop and my brand new MacBook Pro w/ Retina Display.
This can't be a network problem because I pay Verizon way too much money every month for their FiOS Quantum service and various speedtests that I run look OK.
Is the problem with Safari or what?

You guys are not following me. And, I'm not sure if I used that term right. I was talking about answers you see from google searches, not from people here. I am a 40 year old business owner not a kid. What I meant by "fanboy" is that a lot of times when you search online for something like "Itunes on my PC is not seeing my ipod", you see the most popular answer is "Buy a mac". Yes it may be funny, but it is not really helpful. That is the type of person I was referring to, and I was not talking about people here. People on this discussion board seem to be very helpful. I own macs and PC's, so don't like the answer "get a mac". I have one (three in fact), but I also have PC's. I love my mac computers, my iPhone, my iPad, my AppleTV and all of that just like you guys probably do.
I still think it is an interesting and strange issue. If you contacted the IE folks, they would say "Are you having this problem on all websites?". As soon as you said "No, just this one". They would say "It must be the way the website is written". If you talk to someone here, the answer might lean towards "If it runs fine on Safari, but not on IE, it must be an IE issue".
I was just thinking it was an interesting enough scenario that someone had looked into it. Sorry if I was misunderstood and riled someone up.
Message was edited by: Austin Rob

Photoshop laggy, poor performance with GPU enabled

I'm using Photoshop CS6 for some time now. I had enabled GPU usage earlier and I remember that Photoshop ran absolutely smoothly without and lag what-so-ever. This morning, all of a sudden, it has become extermely laggy. I don't understand the reason behind this. This had happened to me before as well but it became alright after a few days. This also happened to me once while using Photoshop CS5. I didn't install any plugin and didn't change any software. I also want to point out that I had 13.4 driver version when Photoshop was working fine, and now too I have the same driver version.
CPU: AMD Phenom II X4 955
Memory: 6 GB
Free storage: 8 GB
GPU: ATI Radeon HD 5850

I've also noticed poor performance since installing the 10.1 plugin. Not just in HD video etc, but in general use. So much for "improved performance"...
No idea on solutions at this stage - In the process of downgrading to previous version to see if that fixes my problems

Non jdriver poor performance with oracle cluster

Hi,
we decided to implement batch input and went from Weblogic Jdriver to Oracle Thin 9.2.0.6.
Our system are a Weblogic 6.1 cluster and an Oracle 8.1.7 cluster.
Problem is .. with the new Oracle drivers our actions on the webapp takes twice as long as with Jdriver. We also tried OCI .. same problem. We switched to a single Oracle 8.1.7 database .. and it worked again with all thick or thin drivers.
So .. new Oracle drivers with oracle cluster result in bad performance, but with Jdriver it works perfectly. Does sb. see some connection?
I mean .. it works with Jdriver .. so it cant be the database, huh? But we really tried with every JDBC possibility! In fact .. we need batch input. Advise is very appreciated =].
Thanx for help!!
Message was edited by mindchild at Jan 27, 2005 10:50 AM
Message was edited by mindchild at Jan 27, 2005 10:51 AM

Thx for quick replys. I forget to mention .. we also tried 10g v10.1.0.3 from instantclient yesterday.
I have to agree with Joe. It was really fast on the single machine database .. but we had same poor performance with cluster-db. It is frustrating. Specially if u consider that the Jdriver (which works perfectly in every combination) is 4 years old!
Ok .. we got this scenario, with our appPage CustomerOverview (intensiv db-loading) (sorry.. no real profiling, time is taken with pc watch) (Oracle is 8.1.7 OPS patch level1) ...
WL6.1_Cluster + Jdriver6.1 + DB_cluster => 4sec
WL6.1_Cluster + Jdriver6.1 + DB_single => 4sec
WL6.1_Cluster + Ora8.1.7 OCI + DB_single => 4sec
WL6.1_Cluster + Ora8.1.7 OCI + DB_cluster => 8-10sec
WL6.1_Cluster + Ora9.2.0.5/6 thin + DB_single => 4sec
WL6.1_Cluster + Ora9.2.0.5/6 thin + DB_cluster => 8sec
WL6.1_Cluster + Ora10.1.0.3 thin + DB_single => 2-4sec (awesome fast!!)
WL6.1_Cluster + Ora10.1.0.3 thin + DB_cluster => 6-8sec
Customers rough us up, because they cannot mass order via batch input. Any suggestions how to solve this issue is very appreciated.
TIA
>
>
Markus Schaeffer wrote:
Hi,
we decided to implement batch input and went fromWeblogic Jdriver to Oracle Thin 9.2.0.6.
Our system are an Weblogic 6.1 cluster and a Oracle8.1.7 cluster.
Problem is .. with the new Oracle drivers ouractions on the webapp takes twice as long
as with Jdriver. We also tried OCI .. same problem.We switched to a single Oracle 8.1.7
database .. and it worked again with all thick orthin drivers.
So .. new Oracle drivers with oracle cluster
result in bad performance, but with
Jdriver it works perfectly. Does sb. see someconnection?Odd. The jDriver is OCI-based, so it's something
else. I would try the latest
10g driver if it will work with your DBMS version.
It's much faster than any 9.X
thin driver.
Joe
I mean .. it works with Jdriver .. so it cant bethe database, huh? But we really
tried with every JDBC possibility!
Thanx for help!!

Poor Performance with Converged Fabrics

Hi Guys,
I'm having some serious performance issues with Converged Fabrics in my Windows Server 2012 R2 lab. I'm planning on creating a Hyper-V cluster with 3 nodes. I've built the first node, building and installing/configuring OS and Hyper-V pretty straight forward.
My issue is with Converged Fabrics, I'm absolutely getting very slow performance in the sense of managing the OS, Remote Desktop connections taking very long and eventually times out. Server unable to find a writable domain controller due to slow performance.
If I remove the converged fabric everything is awesome, works as expected. Please note that the cluster hasn't even been built yet and experiencing this poor performance.
Here is my server configuration:
OS: Windows Server 2012 R2
RAM: 64GB
Processor: Intel I7 Gen 3
NICS: 2 X Intel I350-T2 Adapters, supporting SRIOV/VMQ
Updates: All the latest updates applied
Storage:
Windows Server 2012 R2 Storage Spaces
Synology DS1813+
Updates: All the latest updates applied
Below is the script I've written to automate the entire process.
# Script: Configure Hyper-V
# Version: 1.0.2
# Description: Configures the Hyper-V Virtual Switch and
#              Creates a Converged Fabric
# Version 1.0.0: Initial Script
# Version 1.0.1: Added the creation of SrIOV based VM Switches
# Version 1.0.2: Added parameters to give the NLB a name, as well as the Hyper-V Switch
param
    [Parameter(Mandatory=$true)]
    [string]$TeamAdapterName="",
    [Parameter(Mandatory=$true)]
    [string]$SwitchName="",
    [Parameter(Mandatory=$true)]
    [bool]$SrIOV=$false
#Variables
$CurrentDate = Get-Date -Format d
$LogPath = "C:\CreateConvergedNetworkLog.txt"
$ManagmentOSIPv4="10.150.250.5"
$ManagmentOS2IPv4="10.250.251.5"
#$CommanGatewayIPv4="10.10.11.254"
$ManagmentDNS1="10.150.250.1"
$ManagmentDNS2="10.150.250.3"
$ManagmentDNS3="10.250.251.1"
$ManagmentDNS4="10.250.251.3"
$ClusterIPv4="10.253.251.1"
$LiveMigrationIPv4="10.253.250.1"
$CSVIPv4="10.100.250.1"
$CSV2IPv4="10.250.100.1"
#Set Excution Policy
Write-Host "Setting policy settings..."
Set-ExecutionPolicy UnRestricted
try
    # Get existing network adapters that are online
    if($SrIOV)
        #$sriov_adapters = Get-NetAdapterSriov | ? Status -eq Up | % Name # Get SRIOV Adapters
        $adapters = Get-NetAdapterSriov | ? Status -eq Up | % Name # Get SRIOV Adapters
        Enable-NetAdapterSriov $adapters # Enable SRIOV on the adapters
    else
        $adapters = Get-NetAdapterSriov | % Name
        #$adapters = Get-NetAdapter | ? Status -eq Up | % Name
    # Create NIC team
    if ($adapters.length -gt 1)
        Write-Host "$CurrentDate --> Creating NIC team $TeamAdapterName..."
        Write-Output "$CurrentDate --> Creating NIC team $TeamAdapterName..." | Add-Content $LogPath
        #New-NetLbfoTeam -Name "ConvergedNetTeam" -TeamMembers $adapters -Confirm:$false | Add-Content $LogPath
        New-NetLbfoTeam -Name $TeamAdapterName -TeamMembers $adapters -TeamingMode SwitchIndependent -LoadBalancingAlgorithm Dynamic -Confirm:$false | Add-Content $LogPath
    else
        Write-Host "$CurrentDate --> Check to ensure that at least 2 NICs are available for teaming"
        throw "$CurrentDate --> Check to ensure that at least 2 NICs are available for teaming" | Add-Content $LogPath
    # Wait for team to come online for 60 seconds
    Start-Sleep -s 60
    if ((Get-NetLbfoTeam).Status -ne "Up")
        Write-Host "$CurrentDate --> The ConvergedNetTeam NIC team is not online. Troubleshooting required"
        throw "$CurrentDate --> The ConvergedNetTeam NIC team is not online. Troubleshooting required" | Add-Content $LogPath
    # Create a new Virtual Switch
    if($SrIOV) #SRIOV based VM Switch
        Write-Host "$CurrentDate --> Configuring converged fabric $SwitchName with SRIOV..."
        Write-Output "$CurrentDate --> Configuring converged fabric $SwitchName with SRIOV..." | Add-Content $LogPath
        #New-VMSwitch "ConvergedNetSwitch" -MinimumBandwidthMode Weight -NetAdapterName "ConvergedNetTeam" -EnableIov $true -AllowManagementOS 0
        New-VMSwitch $SwitchName -MinimumBandwidthMode Weight -NetAdapterName $TeamAdapterName -EnableIov $true -AllowManagementOS 0
        $CreatedSwitch = $true
    else #Standard VM Switch
        Write-Host "$CurrentDate --> Configuring converged fabric $SwitchName..."
        Write-Output "$CurrentDate --> Configuring converged fabric $SwitchName..." | Add-Content $LogPath
        #New-VMSwitch "ConvergedNetSwitch"-MinimumBandwidthMode Weight -NetAdapterName "ConvergedNetTeam" -AllowManagementOS 0
        New-VMSwitch $SwitchName -MinimumBandwidthMode Weight -NetAdapterName $TeamAdapterName -AllowManagementOS $false
        $CreatedSwitch = $true
    if($CreatedSwitch)
        #Set Default QoS
        Write-Host "$CurrentDate --> Setting default QoS policy on $SwitchName..."
        Write-Output "$CurrentDate --> Setting default QoS policy $SwitchName..." | Add-Content $LogPath
        #Set-VMSwitch "ConvergedNetSwitch"-DefaultFlowMinimumBandwidthWeight 30
        Set-VMSwitch $SwitchName -DefaultFlowMinimumBandwidthWeight 20
        #Creating Management OS Adapters (SYD-MGMT)
        Write-Host "$CurrentDate --> Creating and configuring virtual NIC for Management OS"
        Write-Output "$CurrentDate --> Creating and configuring virtual NIC for Management OS" | Add-Content $LogPath
        Add-VMNetworkAdapter -ManagementOS -Name "SYD-MGMT" -SwitchName $SwitchName
        Set-VMNetworkAdapter -ManagementOS -Name "SYD-MGMT" -MinimumBandwidthWeight 30 -VmqWeight 80
        Set-VMNetworkAdapterVlan -ManagementOS -VMNetworkAdapterName "SYD-MGMT" -Access -VlanId 0
        #Creating Management OS Adapters (MEL-MGMT)
        Write-Host "$CurrentDate --> Creating and configuring virtual NIC for Management OS"
        Write-Output "$CurrentDate --> Creating and configuring virtual NIC for Management OS" | Add-Content $LogPath
        Add-VMNetworkAdapter -ManagementOS -Name "MEL-MGMT" -SwitchName $SwitchName
        Set-VMNetworkAdapter -ManagementOS -Name "MEL-MGMT" -MinimumBandwidthWeight 30 -VmqWeight 80
        Set-VMNetworkAdapterVlan -ManagementOS -VMNetworkAdapterName "MEL-MGMT" -Access -VlanId 0
        #Creating Cluster Adapters
        Write-Host "$CurrentDate --> Creating and configuring virtual NIC for Cluster"
        Write-Output "$CurrentDate --> Creating and configuring virtual NIC for Cluster" | Add-Content $LogPath
        #Add-VMNetworkAdapter -ManagementOS -Name "Cluster" -SwitchName "ConvergedNetSwitch"
        Add-VMNetworkAdapter -ManagementOS -Name "HV-Cluster" -SwitchName $SwitchName
        Set-VMNetworkAdapter -ManagementOS -Name "HV-Cluster" -MinimumBandwidthWeight 20 -VmqWeight 80
        Set-VMNetworkAdapterVlan -ManagementOS -VMNetworkAdapterName "HV-Cluster" -Access -VlanId 0
        #Creating LiveMigration Adapters
        Write-Host "$CurrentDate --> Creating and configuring virtual NIC for LiveMigration"
        Write-Output "$CurrentDate --> Creating and configuring virtual NIC for LiveMigration" | Add-Content $LogPath
        Add-VMNetworkAdapter -ManagementOS -Name "HV-MIG" -SwitchName $SwitchName
        Set-VMNetworkAdapter -ManagementOS -Name "HV-MIG" -MinimumBandwidthWeight 40 -VmqWeight 90
        Set-VMNetworkAdapterVlan -ManagementOS -VMNetworkAdapterName "HV-MIG" -Access -VlanId 0
        #Creating iSCSI-A Adapters
        Write-Host "$CurrentDate --> Creating and configuring virtual NIC for iSCSI-A"
        Write-Output "$CurrentDate --> Creating and configuring virtual NIC for iSCSI-A" | Add-Content $LogPath
        Add-VMNetworkAdapter -ManagementOS -Name "iSCSI-A" -SwitchName $SwitchName
        Set-VMNetworkAdapter -ManagementOS -Name "iSCSI-A" -MinimumBandwidthWeight 40 -VmqWeight 100
        Set-VMNetworkAdapterVlan -ManagementOS -VMNetworkAdapterName "iSCSI-A" -Access -VlanId 0
        #Creating iSCSI-B Adapters
        Write-Host "$CurrentDate --> Creating and configuring virtual NIC for iSCSI-B"
        Write-Output "$CurrentDate --> Creating and configuring virtual NIC for iSCSI-B" | Add-Content $LogPath
        Add-VMNetworkAdapter -ManagementOS -Name "iSCSI-B" -SwitchName $SwitchName
        Set-VMNetworkAdapter -ManagementOS -Name "iSCSI-B" -MinimumBandwidthWeight 40 -VmqWeight 100
        Set-VMNetworkAdapterVlan -ManagementOS -VMNetworkAdapterName "iSCSI-B" -Access -VlanId 0
        Write-Host "Waiting 40 seconds for virtual devices to initialise"
        Start-Sleep -Seconds 40
        #Configure the IP's for the Virtual Adapters
        Write-Host "$CurrentDate --> Configuring IPv4 address for the Management OS virtual NIC"
        Write-Output "$CurrentDate --> Configuring IPv4 address for the Management OS virtual NIC" | Add-Content $LogPath
        #New-NetIPAddress -InterfaceAlias "vEthernet (SYD-MGMT)" -IPAddress $ManagmentOSIPv4 -PrefixLength 24 -DefaultGateway $CommanGatewayIPv4
        New-NetIPAddress -InterfaceAlias "vEthernet (SYD-MGMT)" -IPAddress $ManagmentOSIPv4 -PrefixLength 24
        Set-DnsClientServerAddress -InterfaceAlias "vEthernet (SYD-MGMT)" -ServerAddresses ($ManagmentDNS1, $ManagmentDNS2)
        Write-Host "$CurrentDate --> Configuring IPv4 address for the Management OS virtual NIC"
        Write-Output "$CurrentDate --> Configuring IPv4 address for the Management OS virtual NIC" | Add-Content $LogPath
        New-NetIPAddress -InterfaceAlias "vEthernet (MEL-MGMT)" -IPAddress $ManagmentOS2IPv4 -PrefixLength 24
        Set-DnsClientServerAddress -InterfaceAlias "vEthernet (MEL-MGMT)" -ServerAddresses ($ManagmentDNS3, $ManagmentDNS4)
        Write-Host "$CurrentDate --> Configuring IPv4 address for the Cluster virtual NIC"
        Write-Output "$CurrentDate --> Configuring IPv4 address for the Cluster virtual NIC" | Add-Content $LogPath
        New-NetIPAddress -InterfaceAlias "vEthernet (HV-Cluster)" -IPAddress $ClusterIPv4 -PrefixLength 24
        #Set-DnsClientServerAddress -InterfaceAlias "vEthernet (HV-Cluster)" -ServerAddresses $ManagmentDNS1
        Write-Host "$CurrentDate --> Configuring IPv4 address for the LiveMigration virtual NIC"
        Write-Output "$CurrentDate --> Configuring IPv4 address for the LiveMigration virtual NIC" | Add-Content $LogPath
        New-NetIPAddress -InterfaceAlias "vEthernet (HV-MIG)" -IPAddress $LiveMigrationIPv4 -PrefixLength 24
        #Set-DnsClientServerAddress -InterfaceAlias "vEthernet (LiveMigration)" -ServerAddresses $ManagmentDNS1
        Write-Host "$CurrentDate --> Configuring IPv4 address for the iSCSI-A virtual NIC"
        Write-Output "$CurrentDate --> Configuring IPv4 address for the iSCSI-A virtual NIC" | Add-Content $LogPath
        New-NetIPAddress -InterfaceAlias "vEthernet (iSCSI-A)" -IPAddress $CSVIPv4 -PrefixLength 24
        #Set-DnsClientServerAddress -InterfaceAlias "vEthernet (iSCSI-A)" -ServerAddresses $ManagmentDNS1
        Write-Host "$CurrentDate --> Configuring IPv4 address for the iSCSI-B virtual NIC"
        Write-Output "$CurrentDate --> Configuring IPv4 address for the iSCSI-B virtual NIC" | Add-Content $LogPath
        New-NetIPAddress -InterfaceAlias "vEthernet (iSCSI-B)" -IPAddress $CSV2IPv4 -PrefixLength 24
        #Set-DnsClientServerAddress -InterfaceAlias "vEthernet (CSV2)" -ServerAddresses $ManagmentDNS1
        #Write-Host "$CurrentDate --> Configuring IPv4 address for the VMNet virtual NIC"
        #Write-Output "$CurrentDate --> Configuring IPv4 address for the VMNet virtual NIC" | Add-Content $LogPath
        #New-NetIPAddress -InterfaceAlias "vEthernet (VMNet)" -IPAddress $VMNetIPv4 -PrefixLength 24
        #Set-DnsClientServerAddress -InterfaceAlias "vEthernet (VMNet)" -ServerAddresses $ManagmentDNS1
        Write-Host "$CurrentDate --> Hyper-V Configuration is Complete"
        Write-Output "$CurrentDate --> Hyper-V Configuration is Complete" | Add-Content $LogPath
catch [Exception]
    throw "$_" | Add-Content $LogPath
I would really like to know why I'm getting absolutely poor performance. Any help on this would be most appreciated.

I didn't parse the entire script, but a few things stand out.
SR-IOV and teaming don't mix. The purpose of SR-IOV is to go straight from the virtual machine into the physical adapter and back, completely bypassing the entire Hyper-V virtual switch and everything that goes with it. Team or SR-IOV.
You're adding DNS servers to adapters that don't need them. Inbound traffic is going to be confused, to say the least. The only adapter that should have DNS addresses is the management adapter. For all others, you should run Set-DnsClient -RegisterThisConnectionsAddress
$false.
I don't know that I'm reading your script correctly, but it appears you have multiple adapters set up for management. That won't end well.
It also looks like you have QoS weights that total over 100. That also won't end well.
I don't know that these explain poor performance like you're describing, though. It could just be that you're a victim of network adapters/drivers that have poor support for VMQ. Bad VMQ is worse than no VMQ. But, VMQ+teaming+SR-IOV sounds like recipe for
heartache to me, so I'd start with that.
Eric Siron Altaro Hyper-V Blog
I am an independent blog contributor, not an Altaro employee. I am solely responsible for the content of my posts.
"Every relationship you have is in worse shape than you think."

Poor Performance with Fairpoint DSL

I started using Verizon DSL for my internet connection and had no problems. When Fairpoint Communications purchased Verizon (this is in Vermont), they took over the DSL (about May 2009). Since then, I have had very poor performance with all applications as soon as I start a browser. The performance problems occur regardless of the browser - I've tried Firefox (3.5.4), Safari (4.0.3) and Opera (10.0). I've been around and around with Fairpoint for 6 months with no resolution. I have not changed any software or hardware on my Mac during that time, except for updating the browsers and Apple updates to the OS, iTunes, etc. The performance problems continued right through these updates. I've run tests to check my internet speed and get times of 2.76Mbps (download) and 0.58Mbps (upload) which are within the specified limits for the DSL service. My Mac is a 2GHz PowerPC G5 runnning OSX 10.4.11. It has 512MB DDR SDRAM. I use a Westell Model 6100 modem for the DSL provided by Verizon.
Some of the specific problems I see are:
1. very long waits of more than a minute after a click on an item in the menu bar
2. very long waits of more than two minutes after a click on an item on a browser page
3. frequent pinwheels in response to a click on a menu item/browser page item
4. frequent pinwheels if I just move the mouse without a click
5. frequent messages for stopped/unresponsive scripts
6. videos (like YouTube) stop frequently for no reason; after several minutes, I'll get a little audio but no new video; eventually after several more minutes it will get going again (both video and audio)
7. response in non-browser applications is also very slow
8. sometimes will get no response at all to a mouse click
9. trying to run more than one browser at a time will bring the Mac to its knees
10. browser pages frequently take several minutes to load
These are just some of the problems I have.
These problems all go away and everything runs fine as soon as I stop the browser. If I start the browser, they immediately surface again. I've trying clearing the cache, etc with no improvements.
What I would like to do is find a way to determine if the problem is in my Mac or with the Fairpoint service. Since I had no problems with Verizon and have made no changes to my Mac, I really suspect the problem lies with Fairpoint. Can anyone help me out? Thanks.

1) Another thing that you could try it is deleting the preference files for networking. Mac OS will regenerate these files. You would then need to reconfigure your network settings.
The list of files comes from Mac OS X 10.4.
http://discussions.apple.com/message.jspa?messageID=8185915#8185915
http://discussions.apple.com/message.jspa?messageID=10718694#10718694
2) I think it is time to do a clean install of your system.
3) It's either the software or an intermittent hardware problem.
If money isn't an issue, I suggest an external harddrive for re-installing Mac OS.
You need an external Firewire drive to boot a PowerPC Mac computer.
I recommend you do a google search on any external harddrive you are looking at.
I bought a low cost external drive enclosure. When I started having trouble with it, I did a google search and found a lot of complaints about the drive enclosure. I ended up buying a new drive enclosure. On my second go around, I decided to buy a drive enclosure with a good history of working with Macs. The chip set seems to be the key ingredient. The Oxford line of chips seems to be good. I got the Oxford 911.
The latest the hard drive enclosures support the newer serial ata drives. The drive and closure that I list supports only older parallel ata.
Has everything interface:
FireWire 800/400 + USB2, + eSATA 'Quad Interface'
save a little money interface:
FireWire 400 + USB 2.0
This web page lists both external harddrive types. You may need to scroll to the right to see both.
http://eshop.macsales.com/shop/firewire/1394/USB/EliteAL/eSATAFW800_FW400USB
Here is an external hd enclosure.
http://eshop.macsales.com/item/Other%20World%20Computing/MEFW91UAL1K/
Here is what one contributor recommended:
http://discussions.apple.com/message.jspa?messageID=10452917#10452917
Folks in these Mac forums recommend LaCie, OWC or G-Tech.
Here is a list of recommended drives:
http://discussions.apple.com/thread.jspa?messageID=5564509#5564509
FireWire compared to USB. You will find that FireWire 400 is faster than USB 2.0 when used for a external harddrive connection.
http://en.wikipedia.org/wiki/UniversalSerial_Bus#USB_compared_toFireWire
http://www23.tomshardware.com/storageexternal.html

ZFS on HW RAID with poor performance. Ideas?

Similar Messages

Maybe you are looking for