ESXi - N40L - Slow Network Speeds


Recommended Posts

Hi Guys,

I am running 2 VMS, WHS2011 and WinXP, both have vmxnet3 drivers installed. The XP network says its running at 1.4GB/s and the WHS network is running at 10GB/s.

When I transfer files from one VM to the other i am getting speeds of 2MB/s. Why is this so slow??

I have run iperf and get speeds of 102 Mbits/sec

whilst the file transfer is in progress CPU usage is around 10% on both VM machines, but when I run iperf, CPU usage maxs out at 100% on the XP VM and about 55-65% on the WHS2011 VM.

Is the CPU being at 100% throttling the network?

I am running one physical network card,

Can anyone shed some light on this? I am a newbie to VMs but not to computers :)

thanks!

Link to comment
https://www.neowin.net/forum/topic/1128838-esxi-n40l-slow-network-speeds/
Share on other sites

Those speeds (That are reported in Windows) are for the Virtual Adapters. Although its true that transferring speeds between VM's on the same does write directly between Memory and not hit the wire so to speak. The speed will only be 1Gbps on the Physical Adapter by the way. Have you applied all the latest patches to ESXi, I would highly recommend it. My CPU usage is always really low, Infact I am surprised at how efficient it is. I just need some more memory in my baby!

http://www.vmware.com/patchmgr/findPatch.portal

SK[' timestamp=1357165637' post='595429082]

If your using the vmxnet3 driver shouldn't it be showing 10GB?

*Edit*

Just checked my XP VM out and indeed it shows 1.4GB. Never noticed that before.

102Mbit is about right for a 1GB NIC no?

It might be for XP that it shows 1.4Gbps, Don't have one to test it. I have only seen it at 10GBps. Budman could chime in though and tell us if this is the case.

Yes your right just over 100MB for 1Gbps.

So I don't have an XP vm currently - but I could fire one up for sure.

But between my nas vm and win7 vm I show this, btw 102mbps is SLOW if you showing connected at gig speeds.

C:\Windows\System32>iperf -c storage -w 256k
------------------------------------------------------------
Client connecting to storage, TCP port 5001
TCP window size:  256 KByte
------------------------------------------------------------
[276] local 192.168.1.210 port 49233 connected with 192.168.1.8 port 5001
[ ID] Interval	   Transfer	 Bandwidth
[276]  0.0-10.0 sec   599 MBytes   502 Mbits/sec

I should see if I can get that up a bit - my physical machines see much higher. 800Mbps should be seen on a gig network fairly easy.

So file copy from my nas vm to my win7 vm

post-14624-0-27431800-1357221554.png

So I am seeing 50MBytes per second in this test.

But curious to what disks your writing/reading from - are you just trying to copy a file from datastore drive of vm to another datastore drive on the other vm. My nas has raw access to the 3 drives it uses for storage. So the above tests were from one of those disks to the datastore drive of the win7 vm

Also curious --- you gave 4GB to each of your VMs, do you have more than 8 in your N40L?

So I did a test that I think more reflects what your doing more likely. And that is from datastore drive to datastore drive of each VM. And in that test I got only 23MBps

post-14624-0-56310600-1357222192.png

What else is your datastore disk doing at the time?? That performance is not going to be the best, your 2 different VMs are accessing the same physical disk. So yeah there is going to be some contention for time.

thanks, i have 10GB of memory in my N40L, an 8Gb and a 2Gb stick.

I have a drivepool in WHS, consisting of 2TB and 1.5TB WD Green drives

I am transferring files from the pool to the XP machine which has a second 2TB drive attached.

The VMs themselves were stored on the 250Gb drive, but I have moved them to another 2TB drive for now.

The ESXi is installed on a 8GB USB pen drive.

Thanks

a bit of progress maybe?

under the WHS2011 VM

If i "pull" a file from the xp machine to the WHS drivepool i get speeds of 13.5MB/s

if i "push" a file from the WHS drivepool to the xp machine i get speeds of 2MB/s

could that be a write issue with the XP HDD ? (E: 2TB)

thats still pretty crappy for a gig network. If I do a iperf or file copy to a physical machine from the vm I get this

C:\test>robocopy \\storage\media\cleanup c:\test test.mp4

-------------------------------------------------------------------------------
   ROBOCOPY	 ::	 Robust File Copy for Windows
-------------------------------------------------------------------------------

  Started : Thu Jan 03 10:31:22 2013

   Source = \\storage\media\cleanup\
	 Dest : c:\test\

	Files : test.mp4

  Options : /COPY:DAT /R:1000000 /W:30

------------------------------------------------------------------------------

						   1	\\storage\media\cleanup\
100%		New File			 650.8 m		test.mp4

------------------------------------------------------------------------------

			   Total	Copied   Skipped  Mismatch	FAILED	Extras
	Dirs :		 1		 0		 1		 0		 0		 0
   Files :		 1		 1		 0		 0		 0		 0
   Bytes :  650.86 m  650.86 m		 0		 0		 0		 0
   Times :   0:00:09   0:00:09					   0:00:00   0:00:00


   Speed :			74150984 Bytes/sec.
   Speed :			4242.953 MegaBytes/min.

   Ended : Thu Jan 03 10:31:31 2013

C:\test>iperf -c storage -w 256k
------------------------------------------------------------
Client connecting to storage, TCP port 5001
TCP window size:  256 KByte
------------------------------------------------------------
[336] local 192.168.1.100 port 16077 connected with 192.168.1.8 port 5001
[ ID] Interval	   Transfer	 Bandwidth
[336]  0.0-10.0 sec  1012 MBytes   849 Mbits/sec

Notice the 800+ Mbits per second over gig.. And in the 70MBps for file pull from my VM.

Do an iperf test from your physical to your VM, using a higher window size, say -w 256k You can see example of this test in my above output. On the server set it up with iperf -s -w 256k

Do you have 2 physical boxes we can play with - 295Mbps is that your iperf test.. That is pretty crappy for a gig network. Did you use the larger window size for your test?

at 8.59 mbps -- is that over wireless or something? Are you reading writing from a USB disk or something.. That is just horrific performance.

Damn you beat me too it! I was going to say that. How are they formatted?

I added the drives under the vclient and formatted them via windows - is that right? am a novice with ESXi

Do you have 2 physical boxes we can play with - 295Mbps is that your iperf test.. That is pretty crappy for a gig network. Did you use the larger window size for your test?

at 8.59 mbps -- is that over wireless or something? Are you reading writing from a USB disk or something.. That is just horrific performance.

yes, those figures are are with iperf on a wired Gb network

I've just performed another iperf test

physcial XP machine set as server (1Gb network) - wired

physical win7 laptop set as client (100Mb network) - wired

and i get 87Mbits/sec

the laptop has only got a 100Mb network card, so thats the max on this network - this is for this test only.

Well 87mbps on 100mbit connection is about right.. Best you could see on 100mbit is around mid 90's -- think best I have ever seen is 97 something. Never ever going to see actual speed of the connection.

But on gig you should be seeing closer to 800s vs your under 300 number.

So question for you on your VM whs - did you give whs raw access to the disks, or did you create vmfs on the disks that the OS has access to.

So for example on my storage box

post-14624-0-36278900-1357239638.jpg

Notice that the 3 disks I have given it to use in a drive pool the disks are raw mapped, and then above that you see the virtual disk (on the datastore drive) that is the vmfs for the OS installed too.

Here is a guide on how to do that

http://vm-help.com/esx40i/SATA_RDMs.php

So for example here is my maps

/vmfs/volumes/4f677e4a-81443006-5233-2c768aadf656/RDMs # ls -la

drwxr-xr-x 1 root root 1120 Dec 26 13:12 .

drwxr-xr-t 1 root root 5180 Dec 29 14:59 ..

-rw------- 1 root root 2000398934016 Apr 7 2012 rdm1-rdmp.vmdk

-rw------- 1 root root 521 Dec 26 13:12 rdm1.vmdk

-rw------- 1 root root 750156374016 Apr 16 2012 rdm2-rdmp.vmdk

-rw------- 1 root root 520 Dec 26 13:12 rdm2.vmdk

-rw------- 1 root root 750156374016 Apr 14 2012 rdm3-rdmp.vmdk

-rw------- 1 root root 520 Dec 26 13:12 rdm3.vmdk

/vmfs/volumes/4f677e4a-81443006-5233-2c768aadf656/RDMs #

So couple of things to deal with on your slow performance - why your iperf between vms is so slow. And then make sure that your drive access is as fast as it can be. How are attaching the disks to your VMs could be a limiting factor - and have to look into the speed between vms from a network as well. I do recall seeing above 1Gbps before on iperf between machines

yeah just looked up a old thread about N40L performance - and in a test there I was seeing 1.5Gbps between vms - have to take a look why only seeing 500mbps now

https://www.neowin.net/forum/topic/1071115-esxi-hp-microserver-n40l-performance/page__view__findpost__p__594826393

post-14624-0-44916300-1335455395.jpg

So I got something not quite right as well - wonder what would of changed?? I know the client is different in this test. Have to look to getting my between vms speeds back to that over 1Gbps ;)

This is a copy from A Win7 VM to my physical Desktop over a 1GB switch. The discs are thin provisioned.


-------------------------------------------------------------------------------
ROBOCOPY :: Robust File Copy for Windows
-------------------------------------------------------------------------------
Started : 04 January 2013 09:16:13
Source : VM7COMPUTER
Dest : C:\Users\Andrew\Desktop\
Files : SOME.FILE
Options : /DCOPY:DA /COPY:DAT /R:1000000 /W:30
------------------------------------------------------------------------------
1 VM7COMPUTER

100% New File 701.8 m SOME.FILE
------------------------------------------------------------------------------
Total Copied Skipped Mismatch FAILED Extras
Dirs : 1 0 0 0 0 0
Files : 1 1 0 0 0 0
Bytes : 701.88 m 701.88 m 0 0 0 0
Times : 0:00:06 0:00:06 0:00:00 0:00:00

Speed : 108010209 Bytes/sec.
Speed : 6180.393 MegaBytes/min.
Ended : 04 January 2013 09:16:20
[/CODE]

This is copying from a Win XP machine to my physical desktop over the same switch with the VM in the same datastore. Basically, there is nothing different from the above other than its a VM with XP rather than 7.

[CODE]
-------------------------------------------------------------------------------
ROBOCOPY :: Robust File Copy for Windows
-------------------------------------------------------------------------------
Started : 04 January 2013 10:10:43
Source : VMXPCOMPUTER
Dest : C:\Users\Andrew\Desktop\
Files : SOME.FILE
Options : /DCOPY:DA /COPY:DAT /R:1000000 /W:30
------------------------------------------------------------------------------
1 VMXPCOMPUTER
100% New File 851.2 m SOME.FILE
------------------------------------------------------------------------------
Total Copied Skipped Mismatch FAILED Extras
Dirs : 1 0 0 0 0 0
Files : 1 1 0 0 0 0
Bytes : 851.20 m 851.20 m 0 0 0 0
Times : 0:00:13 0:00:13 0:00:00 0:00:00

Speed : 65230532 Bytes/sec.
Speed : 3732.521 MegaBytes/min.
Ended : 04 January 2013 10:10:57
[/CODE]

XP seems to be much slower than its predecessor. I know they are slightly different file sizes but the speed speaks for itself.

TBH as I run with such a small amount of resources I would sooner use XP over 7 yet XP feels much slower than 7 does as a VM on my MicroServer. Windows 2008 R2 feels much snappier than XP does. I need to ditch my XP machine really as these tests above prove 7 is not performing how it should.

The host is a HP MicroServer (N36L so the mk1 version).

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • Well, it's good to know that they have found a workaround to a problem that they helped create, I guess...
    • Meta is reusing old DDR4 RAM in its servers instead of buying new hardware by Ivan Jenic Image: Meta The global hardware shortage isn’t exactly news, as the entire world has been struggling with rising component prices for quite some time now. And while big companies certainly aren’t as affected as the average consumer, even they aren’t opposed to the idea of saving a few (million) bucks. Meta appears to have found a way to spend less on new hardware while also putting its outdated infrastructure to use, essentially killing two birds with one stone. The company has built a custom chip that lets it reuse memory from retired servers rather than buying new hardware. The chip is called Vistara and allows for connecting old DDR4 RAM from obsolete servers into new servers that rely on DDR5. The problem Vistara solves goes back to a basic mismatch in how long hardware lasts. Meta replaces its servers every three to five years, but the memory modules inside them are good for seven to ten. When a server gets decommissioned, perfectly usable DDR4 RAM goes with it. Meta is presenting the new method at today’s ISCA symposium, but The Register has got hold of a paper that explains how Vistara works. It's a custom ASIC that bridges DDR4 memory to newer processors via aCXL 2.0/1.1 interface over PCIe Gen5 x16. Meta pulls DDR4 sticks from old machines and installs them in dedicated units it calls MemServers, each of which pairs 768GB of DDR5 with 256GB of recovered DDR4. The operating system sees the DDR4 as an additional memory node and draws from it when the primary DDR5 is running low. Off-the-shelf CXL hardware couldn't do this, so Meta built its own. Existing interfaces bundle their own memory with the controller, which makes reusing old RAM sticks impossible. But Vistara separates the controller from the memory entirely, so Meta can plug in whatever DDR4 sticks it has on hand. Meta plans to deploy the new architecture in hyperscale infrastructure with millions of servers, which should mean that Meta’s AI datacenters will now be more efficient. The company is investing heavily in AI infrastructure, especially with its new AI model, Muse Spark, now widely available. All of this doesn't mean that Meta will exclusively rely on "recycled" RAM, but the company is still looking at considerable savings at scale.
    • Save up to 87% on ChatPlayground AI lifetime subscriptions by Steven Parker Today's highlighted deal comes via our Apps + Software section of the Neowin Deals store, where for only a limited time, you can save up to 87% on ChatPlayground AI: lifetime subscriptions. ChatPlayground AI puts the world’s top AI models in one powerful interface, letting you enter a single prompt and instantly compare outputs from multiple models to choose the perfect response for your needs. Boost productivity and creativity with access to the latest AI giants like GPT-4o, Claude Sonnet 4, Gemini 1.5 Flash, DeepSeek V3, and dozens more — all in one window. Whether you’re chatting, coding, generating images, or refining prompts, ChatPlayground AI equips you with advanced tools like prompt engineering, image/PDF chat, saved conversations, and AI image creation, plus priority support to keep your workflow seamless. Access the world’s best AI models Side-by-Side Comparisons: Enter one prompt & instantly view results from multiple AI models to find the best output for your needs 40+ AI Models: Includes GPT-4o, Claude Sonnet 4, Gemini 1.5 Flash, DeepSeek V3, Llama, Perplexity, and many more Multi-Function Platform: Access AI for chat, image generation & coding all within a single interface Web Browser Extension: Offers a Chrome extension to seamlessly integrate the platform into your browsing workflow Boost productivity with powerful features ChatPlayground Interface: Designed for seamless AI model comparison in one window Prompt Engineering: Refine & optimize your prompts for better, more accurate responses Chat with Images & PDFs: Upload visuals and documents to get context-aware answers Saved Chat History: Keep track of past conversations for reference & ongoing projects AI Image Generation: Create high-quality visuals powered by top AI image models Priority Customer Support: Get faster assistance whenever you need it What you'll get with the Unlimited Plan Includes unlimited messages/month Built for prompt engineers, startups, and teams who run experiments nonstop Includes priority access to new features and future models Good to know Length of access: lifetime Redemption deadline: redeem your code within 30 days of purchase Access options: Desktop Max number of device(s): Unlimited Available to both NEW & Existing users Updates included A lifetime subscription to ChatPlayground AI (Unlimited Plan) normally costs $619, but you can pick it up for just $79 for a limited time - that represents a saving of $530 (87% off). Click the link below for more details, always check terms and specifications before making a purchase. Get this ChatPlayground AI (Unlimited) for $79 (was $619) There are also two other discounted plans to choose from. Although priced in U.S. dollars, this deal is available for digital purchase worldwide. Support queries If you have queries or need support for any of the Neowin Deals, please use the contact form here. Neowin Deals are managed and sold by StackCommerce who represent Neowin on an affiliate basis. Why we post these deals We post these because we earn commission on each sale so as not to rely solely on advertising, which many of our readers block. It all helps toward paying staff reporters, servers and hosting costs. So for those that keep moaning and complaining, be thankful we're still online for you to even do that. Other ways to support Neowin Whitelist Neowin by not blocking our ads Create a free member account to see fewer ads Make a donation to support our day to day running costs Subscribe to Neowin - for $14 a year, or $28 a year for an ad-free experience Disclosure: Neowin benefits from revenue of each sale made through our branded deals site powered by StackCommerce.
    • I like Tidal, but it still does not control devices from the mobile/app and still no surround support. And yeah re: above comment I still get a lot of network errors and I am on a 4/4 Gbit Fiber connection.
  • Recent Achievements

    • Reacting Well
      NovaEdgeX earned a badge
      Reacting Well
    • Week One Done
      NovaEdgeX earned a badge
      Week One Done
    • One Year In
      BA the Curmudgeon earned a badge
      One Year In
    • Conversation Starter
      rosiecharles earned a badge
      Conversation Starter
    • First Post
      KMilenkoski1202 earned a badge
      First Post
  • Popular Contributors

    1. 1
      +primortal
      539
    2. 2
      +Edouard
      266
    3. 3
      PsYcHoKiLLa
      151
    4. 4
      Steven P.
      98
    5. 5
      macoman
      66
  • Tell a friend

    Love Neowin? Tell a friend!