Latency of storage

Similar Questions

• Foglight for virtualization - Standard Edition - storage latency

  Hello

  I would like to know how to tell the difference between the latency of storage on different data warehouses - rather than having the same threshold at all.

  The default settings suits us well on our warehouses of data production, but we have some data warehouses where 200ms is no problem and would like to have that reflected in foglight to avoid SPAM-alerts.

  Best regards

  Bjarke Emborg Kragelund

  Hello

  There are probably 2 ways to handle this, if it's that the first option will be faster to implement a large number of data warehouses:

  1 change the rule and put a condition on the name of the store of data and comparison with a different value for data warehouses where this latency is different.

  2. When you look at the rule condition there is usually some kind of variable, something like datastoreLatency.warning/critical/fatal or something similar. When you go to administration > manage registry variables, you can find your variable and edit and brought it. To have a different value for data warehouses where you want other Scope values

  

  Press next and choose the right database and set a value for it

  

  Hope this helps

  Golan

• vSphere vCenter - vCenter 4.0 Update 1 resolution VM issues slow because the disk i/o latency?

  vSphere vCenter - vCenter 4.0 Update 1 resolution VM issues slow because the disk i/o latency?

  My client is slow VM problems that appear to be disk I/O latency on Hitachi storage 1000. I heard in VMware meetings vCenter 4.0 Update 1 solves these problems... without any explanation as to why.

  Rumor is that "after working all the angles of the slowness of the virtual machine including vmtools hardware version, latency of storage, etc. called VMware support and advised to get vCenter to U1 to solve the 'known problem' VM performance issues." "U1 fact and problem solved.

  It is VMware User meeting and I have no other details.

  Customer did not of U1 and having problems of slowness VM believes may be due to the assets/Hitachi 1000 s and favorite paths questions (known to cause problems). Preferred fixed pathing but questions continue.

  My limited participation in the issue of troubleshooting in the date limit information available now. It is just an issue if 'known problem' exists and is known to be resolved via U1 vCenter or ESX 4.0 U1?

  Not vCenter 4.0 Update 1 solve disk i/o latency issues considered the slow pace of the vm? Why?

  All points will be awarded.

  that strike!  You can use round robin or MRU for active/passive.

• VMware View 4.6 and vSphere Client performance

  Hello world

  I prepare a demo for VMware View 4.6, a small, with internal disks (3 x 600 GB, 6 GB/s SAS, 10 k RPM RAID 5), network: 100 MB (it was never a problem demos), within IBM x 3650 M3, 32 GB of RAM, 2 x xeon 4 c E5620 2.4 GHz.

  There are 3 virtual machines of Windows 2008 R2 (4 GB RAM and 20 GB disks) and 6 virtual desktops (2 GB RAM) and 30 GB drive XP

  In 10 k RPM there is 150 IOPS / s, so 3 drives x 150 = 450 IOPS / s-> 450 / 9 mV = 50 IOPS / s p/vm. I know that if IOPSp/vm is over 30, it's good, so this seems like a good script.

  I don't know what should I look to find the cause of poor performance.

  VI Client is really slow, and worse even desktop computers.

  Can you guide me to find the problem please?

  Thank you!

  Your latency of storage time seems a bit high, how the store looks like it? Possible alignment issues?

  You have a lot of CPU resources, I'd give workstations an another vCPU.

  Linjo

• Windows 7 or 8.1?

  my system

  I7-3960

  64 gb

  SSD 512 boot

  6 WD on areca raid

  NVIDIA gtx 770

  
  

  mostly use cinema 4 d, photoshop, pp, ae

  
  

  Noticing these last months, just a general slugishness to develop (worsened when cc 2014 adobe apps came out)

  
  

  This installation of win 7 ult 64 bit is prolly 2 years, maybe more.  pretty much when intel has published my chip.  so im guessing a good facility will help alleviate some problems.

  
  

  y at - it a diff in performance between 7 and 8.1?  don't like not 8 but if things will run a little smoother that 7 patients switch to him since im reinstalling everything anyway.  or if they have the same, ill just stick with 7.  TIA
  

  The performance is a bit better on Win 8.1 but not significant. The main performance difference is around network latency and storage with what a difference there is. In this case have a clean image of Win 7 improve probably what you see now.

  Eric

  ADK

• Adapter of storage latency Metrics

  Hello

  I notice some read latency in most of my guests showing a maximum of 19 ms, as high as 31 ms, the runtime currently around 3-7 ms.  Is this acceptable?  For some reason, I feel like they are normally less than that.

  You can set only acceptable.  For some, 19 is excellent, for some applications, the application might as well be down.

  Very generally,.<15 ms="" averages="" with="" occasional="" spikes="" to="" 20-30ms="" would="" be="" pretty="">

• Acceptable storage NFS latency

  We have 2 hosts vsphere 4.1, connected to a NAS Aberdeen using NFS shares to store the virtual machine.

  There are currently 6 virtual machines per host.  Very low CPU/memory use.  I am watching the latency on the performance tab and was curious to know what's acceptable limits?  I did a quick capture of the results in real time in the middle of the afternoon and it shows a maximum of 136ms and an average of 36.

  

  I would say no more than 10 to 15 ms. But what is acceptable depends on your users. If they agree with it... Also, if this is not acceptable and what will be the cost of the idea to improve this last? This could be very important. If these figures in itself will not give you much.

• Satellite L750 - Lag latency of registration issue

  MY PROBLEM.

  I used my desktop pc to record guitar on a pre backing track recorded using programs: fruity-loop, Audacity etc. without problem.

  When I try the same on my laptop the problem of latency L750 does not listen to what I play it IE 0.5 second delay of the note to be played to the sound being heard.

  I adjusted the settings of latency in the programs that I use, but this makes no difference.

  Can anyone come up with a fix for this problem because my new laptop becomes a zone of decanting for dust on my desk. ???

  Hello

  I can recommend you to try tools like LatencyMon and DPC latency checker for information on the driver causing the latency issues.
  Here's a thread on the same topic:
  http://forums.computers.Toshiba-Europe.com/forums/thread.jspa?threadID=67087

  I remember that and old TSRI (Intel Rapid Storage Technology Driver) caused this show too.
  This could be resolved by the updated version of the TSRI at least V10.1.0.1008

• Since the last update of Windows, I have problems with iastor.sys dpc latencies...

  ... Technology Intel Rapid Storage Technology on my Ideapad Y560

  Has been published since the last Windows Update (happened to autoamtically in the background ) I now have a few latencies.

  It seems to come from the LatencyMon also says iastor.sys driver.

  Tried to install the version the more recent but no luck. STIL have these DPC latencies every few minutes.

  I also see the spike in ProcessExplorer and DPC latency.

  How can I fix it (without breaking my Windows 7)?

  Seems, I solved it, updated Atheroes AR9285 to 10.0.0.75 downloaded from http://www.atheros.cz/atheros-wireless-download.php?chipset=37&system=6

  Lenovo should add these drivers to their download pages ;-)

• Custom report for several Datastore latencies

  Hello

  I would like to create a report from the latancy VMware Storage model, but have the ability to account for the multiple datatstores. I need to create a new report template to do this, or do I have to create a Service with the data store and to report on this service?

  The only option I seem to get a store of data one selection.

  Thanks in advance.

  Giles program

  Hi Giles

  You use WVE 7.1 or earlier? I remember that the VMware storage latency report could run for the storage of personal data.

  From WVE 7.2, you should be able to run against all data warehouses. Keep in mind the report displays only with latency data warehouses.

  WVE 7.2 should be available in a few weeks.

  I hope this helps.

  Gaston.

• 6100XV vs 6210 X: difference in latency/write speed

  Hello

  In November 2011, we bought a 6100XV and recently a 6210 X was built up to be placed in the same group as the old unit. For testing purposes, both units are run in their own group for now until I am happy with what I see.

  The installation program:
  As storage, we have a PS6100XV (4 x 1 GB + MGT) and a 6210 X (2 x 10 Gbit (copper or fiber) + MGT) the two FW 7.07 and connected to our iSCSI power running.

  Our storage network are made up of two M6348 placed in our M1000E chassis. The switches are stacked and running version 5.1.3.7.

  The processing power for our vSphere cluster comes from our M1000E chassis with 4 blades (M710/M820). Each blade is running vSphere 5.5 (1892794) with MEM 1.2 installed and is connected with 4 x 1Gbps ports internal M6348 using the 4 paths.

  The "problem":
  When JJ - ing 2 GB of data from/dev/zero to drive in the virtual machine, I see different results. Please note that I use the same VMware host, network of equallogic only different etc.:
  6100XV: 101 Mbps
  6210 X: 239 MB/s

  You search in esxtop that I see the following about my ports, iSCSI for the 2 GB write:
  6100XV:
  Latency triggers to 7 ms (also well seen by SanHQ)
  We see and spread also charge all the 4 links of ~ 256 Mbit

  ESXTOP - network
  PORT-ID USED BY TEAM-TEDDY DNAME MbTX/s PKTTX/s PKTRX/s MbRX/s % DRPTX % DRPRX
  50331658 vmk2 vmnic4 vSwitch1 2182,96 256.54 3934,29 4.20 0.00 0.00
  50331659 vmk3 vmnic5 2249,15 vSwitch1 265.84 4087,83 4.37 0.00 0.00
  50331660 vmk4 vmnic8 2182,39 vSwitch1 269.51 4053,12 4.44 0.00 0.00
  50331661 vmk5 vmnic9 vSwitch1 2155,30 267,10 4032,52 4.43 0.00 0.00

  ESXTOP - virtual disk
  GID VMNAME VDEVNAME NVDISK CMDS/sec READING/s WRITTEN/s MBREAD/s MBWRTN/s LAT/rd LAT/wr
  7115 mail38 - 2 576.02 69.81 506.21 1.06 118.68 22,27 7.01

  6210 X:
  Latency in 74 ms (also well seen by SanHQ) triggers
  We see and close enough maxed out the load between the two links of 900 + Mbit

  ESXTOP - network
  PORT-ID USED BY TEAM-TEDDY DNAME MbTX/s PKTTX/s PKTRX/s MbRX/s % DRPTX % DRPRX
  50331658 vmk2 vmnic4 vSwitch1 10995,86 918.20 16286,85 14.44 0.00 0.00
  50331659 vmk3 vmnic5 11036,87 vSwitch1 945.63 16856,19 11.83 0.00 0.00
  50331660 vmk4 vmnic8 247,96 8.55 vSwitch1 251,77 3.59 0.00 0.00
  50331661 vmk5 vmnic9 258,45 8.35 vSwitch1 251,77 4.51 0.00 0.00

  ESXTOP - virtual disk
  GID VMNAME VDEVNAME NVDISK CMDS/sec READING/s WRITTEN/s MBREAD/s MBWRTN/s LAT/rd LAT/wr
  3477742 eql02bench01 - 1 436.24 2.15 434.09 0.01 216.48 11.68 74.16

  Because the 6210 X has only two active ports it uses only two lanes but as you can see he can combined the two links to a connection GB 2 where the 6100XV can not combined all four links to a 4 Gbit connection. In addition to the 'problem' of the bandwidth I see a huge diffrene of latency during the writing between the two units.

  Issues related to the:
  (1) what is the diffrence between the two units which lead to be limited to 1 GB and the other does not
  (2) why is the latency of writing so high? The two controllers are in the Green return entry should be active.

  Kind regards

  -Henk

  Looking at the average size of the latency block is appropriate.  Larger Blocksizes means more latency but greater speed.

  I'm guessing you're under Linux?

  Try "dt".

  www.scsifaq.org/.../DT.html

  Once you compile.  Try something like this:

  . /DT = bs = 8 k = iot disable = reason compare capacity =

  for example

  . of=/mnt/disk/test.dt /dt bs = 8 k = iot disable = ability to compare pattern = 20G

  Then try with bs = 64 k.

  This will do a sequential write then read.   The disable = compare stop checking the data read to find errors, which slows down the test.

  Then try on a blank partition.

  . /DT = / dev/sdx bs = 8 k = iot disable = reason compare capacity =

  And once again with 64K.

  When you run on a file system, it tends to break up great Scriptures.

  If you want to run multiple passes add passes = X at the end.

  Kind regards

• Storage vMotion with Syncrep on Equallogic San

  Hello

  We recently bought 2 San Equallogic (PS6510ES) and turned on synchronous replication between them.

  We use ESXi 5.5 and followed the best practices guide to disable TCP ack and 60 logintimeout value.

  Everything seems to work fine until we tried Storage vMotion/migration and cloning VMS on the San.

  When we did the migration of 2 volumes on the EQL SAN storage (the target volume has active syncrep), the target volume will be out-of-sync and then the storage vMotion will hang sometimes.

  The worst is when such a problem has occurred, all other volumes on the EQL constantly to respond also. All volumes will not respond until we rebooted the corresponding ESXi host (NOTE: simple shut down/power outside the host won't help.) The San EQL will only return to normal after a reboot of the ESXi host).

  If we pause / off the syncrep before Storage vMotion, the problem will not happen.

  Is this normal or not?

  In addition, whenever the problem occurred, the latency of writing the corresponding volume will be dry and we will receive alerts similar to the following:

  connection iSCSI target ' xx.xx.xx.xx:3260, iqn.2001 - 05.com.equallogic:xxxxx - test-volume ' initiator 'xx.xx.xx.xx:xxxxx, eqlinitiatorsyncrep' failed for the following reason:

  Initiator, disconnected from the target when connecting.

  Thanks in advance.

  Hi Joerg,

  What is a #SR? I opened a record of support to 24 June and still waiting for Dell to contact me.

  Ryan

• Storage of hybrid matrices 3 with Betclic

  I have 2 6500ES and 1 6100XS used by 8 ESX hosts approximately 130 virtual machines running.  Performance is important, but I migrated just my MD3000i data on the 6500's, so I already see an improvement... the SSD is impressive.  :)

  I have 3 questions

  1.) can I do 1 pool or 2?  I currently have them in 2 pools, however, I wonder if just do 1 pool and let the faders EQL to load all the work assumes that do it for me... I don't want to have to check the use of the e/s between the basins as I went and manually or using DRS storage so I don't have to. E/s of store DRS requires turning on storage i/o control, which is not recommended according to practice Documentation best EqualLogic VMWare - I'm guessing that because it could hinder the decisions of swing for the loading of the SAN.

  I think I read that best practices says 2 pools, but I don't know if it takes 3 hybrids account.

  With the exception of the virtual machine our data are pretty ephemeral. Receive us a small amount of data, create a large amount of data, duplicate, save it, deliver, then he sits for a period of time before being archived and purged.

  (2.) I think higher than the latency of reading expected when Ops ARE on the 6500 / s fall below 500 Kbps.  It of not crazy high, but revolves around the range 8ms.  I disabled delayed ACK, who brought 12-15 latency time, and want to disable TCP offloading the next maintenance window.  But, is this normal?  My 6100 remains generally less than 2ms read and write

  Updates to the firmware 3) -is it safe to do in activity? Or I really need to stop all my ESX hosts before you do?  The documentation says to do when there is low usage... Does this mean that status at the HQ of SAN said low? or what is considered to be low? Is the same applies to the hard drive firmware updates?

  Thanks in advance!
  David

  Re: 1. I suggest that keep you in the two pools.  Only because of the difference in size between the 6500's and 6100 s.   You could go to 3 swimming pools, but they run you the risk of having to move virtual machines to balance IO.

  Re: 2. I'll try also turn off RSO, great Recieve Offload.

  Re: 3 upgrades.  E/s low means when the table is not widely used.  So yes, a low e/s of SANHQ reading would show that.  If your company is not active 24 x 7, so it's normal after hours time span weekend.   Most important is giving the login_timeout 60 ESXi servers and by setting the disktimeout value to 60 in virtual machines.  With the recent firmware the time required to failover and early response is usually less than 30 seconds.  But not always, one may be closer to 60 seconds under load.  It is also essential to ensure that the switch ports happen immediately on the transition.  (AKA portfast)   This way when the CM goes down, the ports are online immediately.    In addition, to be able to handle a FW update also means you can manage a CM failover on a HW problem.

  Disc FW updates.   Yes.  The updates take place one disc at a time.   Then, the player is power rolled, to update the FW.   That time is very fast, and single e/s to this player is paused, then resumed once the player comes back online.   But again, the best time is low i/o.

  In a perfect scenario, yes make upgrades in a window of Maint would be ideal.  At least stop the virtual machines.   In this something unexpected occurred offshore lucky.

  Since it is ESXi you run at least 6.0.7 firmware?

  Kind regards

• The storage load balancing process

  We have a group of 4 members, total approximately 30 TB of space. In the past, we had a lot of volumes of 500 GB for ESX, after the installation of the ESXi5.1, we use higher volumes, about 1.5 TB. So there are a lot of changes in the storage pool. We find that the EQ is balancing of storage compared to 4 members as expected. We thought it was over two or three days agoo, but now he's balancing... then again for a couple of day all volumes are spread over 3 members, also, as expected, but he decides to charge storage it swings again... and again...  It is coming to an end sometimes?

  We have another group with 4 members and also the same amount of space, but it we do not use the 1.5 TB volumes again... here, we see that EQ is continuous storage load balancing pool.

  Im not worried yet :)), but whenever volumes are redestributing compared to members, it does not follow on some reconnects for ESX as the part of a volume can and will reside on another Member where on that time esx is not a connection to that session at that time.

  Hello René,.

  It will not be stuck in a loop.  Part of the balance to the maintenance of free space between the members.  So when several pages are written to a specific member, space balancing will create plans to move the blocks to the other members to free up space on this member.  It's an ongoing balancing act.  In addition, extended periods of heavy i/o load balancer Advanced Performance will swap pages in order to compensate for the latency between members.

  Kind regards

• Need help to create a storage report/monitoring

  vFoglight V 6.1

  So our storage has been really slow lately by some of our virtual machine crawl. Whenever I talk to our team of storage, that they speak for EMC and the answer is BUY new more fast/drives or take VM off the LUNS are getting slammed. Well that's fine, but now I have a few LUNS with more space free 300Gig because they can not will not try and work with me to find a solution.
  I need to create a report and monitor my LUN so that they show that last year the "slowness" was getting worse. Just storage model reports show stats 'space' unless I'm not totally see models of storage performance.
  Y at - there someone here who can lead me in the right direction to create reports of storage/performance problems? Is there some sort of pre canned reports, I do not see or a 'cartridge', I'm missing here?
  I'd love some help here... Thank you

  Hello
  The reason why you get n/a, it is that the data do not exist. Queue length data are only available once we deployed agents OS like VMware does not provide this in a meaningful way (I am happy check if you can show me otherwise) monitoring. Disk latency information should however be available but only for the host ESX 4 and vCenter 2.5 U4 and later as this was the point where she became exposed.
  .
  I also encourage you to take a look at vFoglight storage, if you want to enter more deeply into the fabric of storage.
  -Mike C