allocation of multiple vCPUs - unipolar or multicore?

Similar Questions

• Multiple vCPUs are very slow

  I found whenever I place using several nuclei, the guest is super slow.  Does anyone know the reason?  The system specification is 2 Core i7, which has 4 cores, each with hyperthreading, i.e. 8 logical processors.

  This phenomenon occurs on OpenSolaris and Windows Server 2008 R2.

  Specifically, once before that I wrote a program, which is purely CPU bounded, which has no IO and memory very little access, to test the performance of the processor.  When I compiled and run it in OpenSolaris guest, running with a thread as quickly as if it was running on the host computer.  However, when I use 2 wires, each is significantly slower than the single thread.  The guest of OpenSolaris is configured to use 2 CPUs.

  The other observation is when I install server on Windows 2008 R2 comments RAID5, it's fast (16-20 MB/s), if I put it on 1 CPU.  But when I try to configure 2 CPUs, performance decreases signifantly (3 Mbps), and the guest is so slow that there can be no answer to my mouse click.

  My conclusion is therefore, on my machine, it seems a CPU is fast (near-native), but multiple vCPUs a performance problem.  Isn't anyone know why or know another thread has what?  Thank you very much.

  BTW: my host CPU's Core 4 w / HT, but I set up my guest for 2 processors and 1 for each processor core.  Change to 1 CPU with 2 hearts on this subject will be better?

  icando9 wrote:

  It is not quite explain why.

  Not necessarily. It is a rule of thumb, not a law. Depending on how many clients and their number of vCPU and on the load on the host computer, this may apply. Even if all clients except for one have only one vCPU this particular comment with still two of vCPU must wait until the two nuclei is free. Who takes always longer than if she is not to wait two hearts.

  In general, you should always start with the least possible vCPU. Virtualization environments attributing more vCPU can degrade the overall performance.

  In addition, according to the guest OS a huge load of CPU interrupts is generated by the virtual timers. According to the kernel Linux, for example, can generate 1000 + 1000 x n (n is the number of vCPU) interrupts per second per guest.

  Two guests with such a core, and each of them with two vCPU to generate a load of 6000 interupts per second, even when they do nothing else.

  BTW: assigning 2 processors with 1 core each and 1 CPU with 2 hearts makes a difference?

  Yes it can. Depending on the application. If she makes intensive use of the caches of the processor that best you do not want to have two vCPU running on a CPU cores because they can use the same hidden then.

  AWo

  \[:o]===\[o:]

  = You want to have this ad as a ringtone on your mobile phone? =

  = Send 'Assignment' to 911 for only $999999,99! =

• slow performance with CentOS 5.5 reviews and multiple vCPUs

  We installed a server in CentOS 5.5 postgresql database comments in vSphere 4.

  This prompt 'verik' (2 vCpu, 16 GB RAM, core EAP) is the only host running on the vSphere host (2 x 6 core AMD Opteron 2439SE, 128 GB of RAM).

  We are witnessing very low yield (for example, when executing a pg_dumpall). The CSTP % seems rather high, especially compared to % RUN - and surprising since there is no claim to the resources of the host.

  09:38:02 to 17:54, 131 worlds; CPU load average: 0.01, 0.02, 0.01

  PCPU USED (%): 11.2 0.6 0.4 0.3 0.2 0.3 0.4 2.1 0.3 0.3 0.1 2.9 AVG: 1.6

  PCPU UTIL (%): 11.4 1.1 0.9 0.9 0.8 0.6 0.9 2.5 0.6 0.7 0.2 3.1 AVG: 2.0

  PSC (%): 7 US, 2, 91 sy id wa 0.       CS/sec: 509

  ID NAME NWLD % USED GID % PERFORMANCE SYS % WAITING RDY % IDLE % OVRLP % CSTP % MLMTD %

  1 1 12 1181.67 slow 1184,71 0.00 0.00 15,87 2.47 0.00 0.00 0.00

  2 2 System 7 0.01 0.01 0.00 700.00 0.00 0.00 0.00 0.00 0.00

  6 6 75 helper 0.16 0.16 0.00 7500,00 0.05 0.00 0.00 0.00 0.00

  7 7 drivers 9 0.01 0.01 0.00 900.00 0.00 0.00 0.00 0.00 0.00

  8 8 vmotion 4 0.00 0.00 0.00 400.00 0.00 0.00 0.00 0.00 0.00

  10 10 console 2 10,59 10.83 0.02 0.00 189,27 89,19 0.35 0.00 0.00

  15 vmkapimod 15 7 0.03 0.03 0.00 700.00 0.00 0.00 0.00 0.00 0.00

  17 17 FT 1 0.00 0.00 0.00 100.00 0.00 0.00 0.00 0.00 0.00

  18 18 8 vobd.4261 0.00 0.00 0.00 800,00 0.00 0.00 0.00 0.00 0.00

  19 19 net - cdp.4269 1 0.00 0.00 0.00 100.00 0.00 0.05 0.00 0.00 0.00

  20 20 vmware-vmkauthd 1 0.00 0.00 0.00 100.00 0.00 0.00 0.00 0.00 0.00

  26 26 verik 4 5.12 4.75 0.35 348,10 0.05 18.24 0.51 47.30 0.00

  We tried a variety of 2.6.18 kernels EAP (including attempts with the CONFIG_HZ 100 and with a divider = 10 options). More or less consistently poor performance.

  The BIOS of the host settings are:

  HyperTransport technology: HT 3

  Assist HT: enabled

  Virtualization technology: enabled

  DRAM prefetcher: enabled

  Training Software Prefetch Prefetch material: enabled

  Hardware prefetcher: enabled

  Demand-based power management: disabled

  We have also tried to run this comments with 1 vCPU (2.6.18 - 194.8.1.el5PAE):

  The first run after a reboot:

  time pg_dumpall >/dev/null

  Real 8m1.764s

  user 0m12.649s

  sys 0m3.133s

  While it works, the system fails (for example, a 'top' which should be updated every second can be updated every 7 seconds). The guest spends a lot of time in the 'system', do what we do not know.

  CPU: 17.3%us, 81.9%sy, 0.0%ni, 0.0%id, 0.0%wa, 0.0%hi, 0.8%si, 0.0%st

  It takes 7 minutes on a server physics multi-core (4x2.8Ghz Xeon).

  It takes 17 minutes at best (more time) on "verik" with 2 vCPUs.

  We would really like to migrate our physical postgres server in VMware, but we cannot justify it doing so unless he can perform at least as well as on physical hardware. (We tried to store the database on the local disk and mounting NFS host datastore; similar performance in all cases. We also checked adequate bandwidth on our links for storage and front networks.)

  We reviewed several postings on execution of the guests with more than 1 vCpu and think we understand how vSphere CPU coscheduling works. But in our scenario, without conflict of use of CPU resources, we do not know why we're having performance issues.

  Any suggestions?

  Thank you!

  Craig

  First, why OS 32-bit VMS with 16 GB of RAM?

  My first and very strong suggestion is to go for the 64-bit version of CentOS, EAP is known to have poor performance. Run us multiple virtual CentOS 64-bit machines, and I can't say that we have unexplained performance issues.

  You might try to twist a little bit the 32 bit existing OS, but I don't expect a big improvement. Keep this divider = 10 in your kernel parameters, but if you like to try again with simple CPU add 'nosmp noapic nolapic' kernel parameters, those can drop the percentage of time processor rdy one quite a bit, at least I saw that happen on busy ESX host. Also assign some pages of huge memory in the Linux kernel and configure PostgreSQL to use.

  Your BIOS settings also appear to be inefficient, the General recommendation is to disable the CPU features on the ESX host preloading since when the CPU executes several processes (VMs) preload results in a high number of failures which is just waste of cycles CPU.

  Equipment training on Software Prefetch Prefetch: disabled

  Hardware Prefetcher: disabled

  But still, go for 64-bit operating system.

• Correct allocation of the vCPUs for virtual machines

  Hi all

  We have a physical server with 4 x Quad Core XEON E7310 1.6 GHz CPU

  VMware ESX 3.5

  We have 3 VMs that are assigned to the server

  VM1 (Production Server) as a result of config

  4 vCPUs, 16 GB of RAM and works according to:

  OS - Windows Svr 2008 R2 (64-bit)

  MS SQL Server 2008 (64-bit) (a database with the data of a line and two newspapers file)

  IIS7 with only a single application of ASP .NET (the App-based eXpress Foundation (XAF) from DevExpress) running in Classic .NET AppPool (a work process)

  Connection string Web Application points to the database on the same server (local)

  Web app has a userbase of around 250 users

  VM2 (Dev server) is the same as VM1 except it has only 8 GB of RAM

  Now 4 x 1.6GHz Quad Core CPU = 4 x 4 x 1.6 GHz = 25.6 GHz

  VM1 received 16GHz (reserved) and can use available CPU if necessary.

  Our problem is that when we don't have 10 users on VM1 it grinds to 80-100% CPU usage.

  Usually just open the app on VM1 is slower than if I use a laptop as the server.

  What can we do to improve the speed of VM1?

  Thank you!

  2 things.  DISC is your problem, not CPU.  It can run at 100% CPU, but it's probably because that the e/s DISK is queued up, and thus the CPU is waiting for data.

  try to reduce the CPU at only 2 vCPU, not 4.  More CPU for physical machines working, but the virtual computer without CPU more due to the work of how virtual machines (the same problem applies to any product of virtualization).

• Problem of migration of a virtual machine with multiple vCPUs

  When I try to migrate some of our virtual machines with vMotion, I get the following error.

  «Impossible to migrate from host a to host b: virtual machine has 4 virtual CPUs, but the host only supports 1.» The number of virtual processors can be limited by the selected guest OS for the virtual machine or the granting of licenses for the host".

  The two ESX Server are identical HP DL380 with two quad core Intel Xeon. They are both connected to the same server vCenter (2.5). They both use the same license server. And, the VM OS is set to Windows Server 2003 Enterprise (32-bit), which is responsible for 4 vCPUs (it does now after all).

  I tried moved under voltage power VM off and got the same error, so I changed the vCPU to 1 and it migrated correctly. Then, I changed it to 4 vCPUs with no problems on the new host. Only problem is, I can't stop some of the virtual machines to migrate to the new host because of the availability requirements.

  ~ Brian

  Have you tried to change the settings of your license and apply them, just in case where?  I had the weird services fails when there is a kind of latency hit the license of the ESX host server.

  -KjB

  VMware vExpert

• Allocation of multiple DNS in UCS.

  Hi, can somehow let me know which DNS server uses the UCS when you have set up several of them? Especially in the case of the cisco routers, this is the one you set up first, is that router queries. BUT I'm not sure that in the case of the UCS, as one that is configured, 1 goes down the list in the GUI.

  Hello

  Since it is based on linux, it would use a list of DNS servers in the order (from first entry to the bottom of the file), as specified in the /etc/resolv.conf file.

  To verify the behavior of GUI and theory, I did a quick test in the laboratory.

  The last entry added via GUI was the first entry in the /etc/resolv.conf file.

  He asks so in order as see you in the GUI.  It is therefore first level the last DNS server registered even though it appears at the bottom of the list in the GUI.

  tcpdump - neither eth0 dst port 53

  IP 22:12:47.381799 FI-MGMT - IP.33128 > 5.5.5.5.53: 35904 + A? www.Cisco.com. (31)

  IP 22:12:52.382920 FI-MGMT - IP.33129 > 3.3.3.3.53: 35904 + A? www.Cisco.com. (31)

  IP 22:12:55.392924 FI-MGMT - IP.33145 > 2.2.2.2.53: 35904 + A? www.Cisco.com. (31)

  Even once questioning followed 5.5.5.5 3.3.3.3 and 2.2.2.2 until maximum tires.

  [EDIT]   If refresh GUI (move to another section and then click again on the DNS management), it now shows the DNS server in the order configured by the user.

  Padma

• Questions about the allocation of multiple NETWORK cards in a machine virtual Windows XP Machine HOST/NAT or bridged?

  Hi all

  That is the situation. We have a Machine physics WindowsXP with 5 CARDS, all with static IP. Only be connected to the internal network without physical Internet connection is used.

  Map NETWORK 1 has a 10.34.78.241/255.255.255.0 IP Gateway 10.34.78.2

  Map NETWORK 2 has an IP 192.168.22.242/255.255.255.0

  Map NETWORK 3 has a 192.168.100.241/255.255.255.0 IP

  Map NETWORK 4 has an IP 172.27.254.240/255.255.0.0

  5 NIC has an IP 192.168.242.10/255.255.255.252

  I used VMWare Converter Standalone to do a P2V PC, which works well with the latest version of VMWare Workstation. All were brought to the course successfully. The virtual machine has 5 Networksas by the XP Machine of origin.

  I built a Win7 Machine with 5 physical NETWORK interface cards, I suppose to replicate the Windows original XP Machine, I need configure the static IP address of the new NICs as shown above and configure cards NETWORK the VM 5 HOST mode and use the virtual network Editor

  to adapt the VMS NICS? Or is there a better way to do it?

  Cheers for any help and advice that can be provided...

  If I'm wrong, you should be able to create a Bridged (vmnet) network for each physical card. Make sure that one (which is created by default) is mapped to a physical NIC dedicated rather than automatic.

  André

• Determine the use of vCPU using PowerCLI

  Is there a way to determine how many vCPUs uses makes it a virtual machine using PowerCLI?  For example, suppose you try to size just your virtual machines with powerCLI.  If the virtual machine uses several threads, it can use the multiple vCPUs.  If it is mono-thread, it can only use one vCPU even if she has a high workload.  You can determine what virtual machines have multiple vCPUs but use less than they are allocated?

  As much as I know there is no meter that indicates if a core is used or not.

  What you could do, is the use of the CPU, and then check if the assigned processors are required.

  For example, suppose that the average CPU usage is 25%, and the maximum is 50%.

  If this VM has 4 processors assigned, you may deduct 2 CPUs would be sufficient.

  But this isn't an accurate calculation.

  Once you lower the processors assigned to the virtual machine, you will need to see if there's an obvious increase of the ready time.

• Server 2008 extremely slow on ESXi 4 (208167)

  A few months ago I started having serious performance on our Terminal server VM server problems, and I have yet to track down the problem. I know that I inevitably left out some sort of vital information, please bare with me and slap my wrist when I forget.

  Our Vmware ESXi system:

  2 quad-core (8 logical cores) processors

  12 GB memory

  RAID 1 Bay for discs of OS and RAID 5 for the data.

  VM on the system:

  Nuclei of SBS 2008 = 4, 9 GB RAM

  Server 2008 Standard (32-bit) = 4 cores, 2.5 GB of RAM

  XP Pro = 1 core, 512 MB of RAM

  This system used to work very well without any problems, but now it is sometimes unusable. The CPU spikes and takes forever, and it's not just with a single process either. It is used to serve a small site with a SQL database (used internally and not very intense) and Terminal Server services, both accounting can remote in and the use of Quickbooks.

  I started, I was wondering if possibly, it started happening when I created the XP Pro VM and if he had to do with assigning a processor when technically all 8 have already been used. Which could cause problems? In addition, the RAM is only 2.5 (made at 3, up to what I gave to XP). I know it's low, even though 32-bit can go to 4, but if I look at the memory usage in the system, it is only at 60% or more. I will probably increase the RAM of anyway, but could not nail down the problem.

  It could be something with general reserve and system? It seems it's only wrong with this machine a virtual that don't is not really implemented any differently from others. Most of the settings for the virtual machine are by default, other than the allocation of resources.

  Even when something as simple as opening session, or open the reliability and Performance Monitor, watching the CPU in the vSphere client usage jump path to the top, but its use disk and memory remain relatively the same.

  How can I find and check to solve this performance problem?

  I found a few other Community assignments that go in-depth on the question of vCPU. Take a good look at both of the threads below. The general consensus is to not use multiple vCPUs unless required. Those positions explain why, and they do it well. I would say that your other admins coming in and changing things may affect the performance of your system. Can you take their rights to change things?

  http://communities.VMware.com/thread/157849

  http://communities.VMware.com/thread/131269

• Help me to understand the distribution of the vCPU, job scheduling and context switch.

  It was my understanding that each virtual computer must have all their vPCUs available at the time of execution of this machine. So if I have 24 available on the host vCPUs and allocate all the vm 24: 1, context switch will produce any time, VMware has to perform tasks. This would basically be 'pause' the virtual machine itself and adversely affect performance. The same thing happens if I have 2 virtual machines in place, 1 24vCPUs, 1 with 2vCPUs. Since they both are not all available vCPUs, they will never run at the same time. Reading, I found it is the so-called 'Gang Scheduling.

  Does this make sense? VMware will recognize that the 1st vm uses a number of vCPUs and plan the 2nd at the same time?

  Welcome to the community - don't forget the multiple vCPUS can if run on the same kernel that allows the high density of virtual machines on a host ESXi - in previous versions, a VM vCPUs wouldn't last unless all vCPUs could be scheduled at the same time - with vSphere 4 this problem has been addressed and now the vCPUs will be get laid at the same time as the other machines virtual - there is always an impact on the the vCPUs planning so that is why it is recommended always to start with one vCPU wingle and add extra would be not necessary -

• DB_nk_CACHE_SIZE in oracle 10g r2

  Hello everyone

  
  

  Official document... Memory architecture

  
  

  (1) how the logic applied here? I want to say 32 M how it came?

  
  

  Memory consumed by components manually sizes reduced the amount of memory available for the automatic adjustment.

  SGA_TARGET = 256M
  DB_8K_CACHE_SIZE = 32M  
  

  The instance has only 224 MB of memory (256-32) and

  remaining to be allocated between the components of size automatically.

  Thanks in advance. !

  SQL> select component,current_size, granule_size from V$SGA_DYNAMIC_COMPONENTS;
  
  COMPONENT                 CURRENT_SIZE GRANULE_SIZE
  ------------------------- ------------ ------------
  shared pool                  411041792      4194304
  large pool                     4194304      4194304
  java pool                      4194304      4194304
  streams pool                   4194304      4194304
  DEFAULT buffer cache         230686720      4194304
  KEEP buffer cache                    0      4194304
  RECYCLE buffer cache                 0      4194304
  DEFAULT 2K buffer cache              0      4194304
  DEFAULT 4K buffer cache              0      4194304
  DEFAULT 8K buffer cache              0      4194304
  DEFAULT 16K buffer cache      33554432      4194304
  
  COMPONENT                 CURRENT_SIZE GRANULE_SIZE
  ------------------------- ------------ ------------
  DEFAULT 32K buffer cache             0      4194304
  Shared IO Pool                       0      4194304
  ASM Buffer Cache                     0      4194304
  
  14 rows selected.
  
  SQL> alter system set db_16k_cache_size=28M;
  System altered.
  
  SQL> select component,current_size, granule_size from V$SGA_DYNAMIC_COMPONENTS where component='DEFAULT 16K buffer cache';
  
  COMPONENT                 CURRENT_SIZE GRANULE_SIZE
  ------------------------- ------------ ------------
  DEFAULT 16K buffer cache      29360128      4194304   <-- Allocated 28M (Multiple of 4)
  
  SQL> select component,current_size, granule_size from V$SGA_DYNAMIC_COMPONENTS where component='DEFAULT 16K buffer cache';
  
  COMPONENT                 CURRENT_SIZE GRANULE_SIZE
  ------------------------- ------------ ------------
  DEFAULT 16K buffer cache      29360128      4194304   <-- Allocated 32M (Multiple of 4)
  
  SQL> alter system set db_16k_cache_size=30M;
  System altered.
  
  SQL> select component,current_size, granule_size from V$SGA_DYNAMIC_COMPONENTS where component='DEFAULT 16K buffer cache';
  
  COMPONENT                 CURRENT_SIZE GRANULE_SIZE
  ------------------------- ------------ ------------
  DEFAULT 16K buffer cache      33554432      4194304   <-- Allocated 32M (Multiple of 4)
  
  SQL> alter system set db_16k_cache_size=31M;
  System altered.
  
  SQL> select component,current_size, granule_size from V$SGA_DYNAMIC_COMPONENTS where component='DEFAULT 16K buffer cache';
  
  COMPONENT                 CURRENT_SIZE GRANULE_SIZE
  ------------------------- ------------ ------------
  DEFAULT 16K buffer cache      33554432      4194304  <-- Allocated 32M (Multiple of 4)
  

  As you can see the db_16k_cache_size is always allocated in pellet-sized pieces (4M). Despite setting the size to 30 M, 31 M, 32M, it will always allocate 32 M.

  Hope this is clear.

  Kind regards

  Suntrupth

• How to optimize a new iMac for After Effects

  I bought the most tricked out available, with 32 GB of ram, a graphics card updated, all new iMac everything.

  But export my project out of After Effects takes so much time on my new iMac, as it did on my macbook 2011 - about 6 hours.

  I know how to handle my export settings, so maybe I just need to change my computer's system settings to work better with After Effects.

  Mac had no advice and showed me Adobe, so I'm here.

  Any advice?

  Learn more about optimizing your system for AE and follow these guidelines, allocate multiples of 4 GB of ram for your process. At the very least keep the ram allocated in multiples of 2 GB. If you use the rendering of the MP, then make sure that the total number of processors equals the total number of processors. It is easy to end up with 0 actually used processors if you allocate too much ram to a core. If you use traced rendered section make sure that you have configured correctly and that you have a NVIDIA GPU correctly configured and installed the latest drivers, CUDA. You must check the box "use maps not taken in charge" in your preferences for rendering raytraced using CUDA is is no longer developed by Adobe. If you do not have or have not properly configured your NVIDIA card, then DO NOT install Cuda drivers.

  Last point, note that all projects will work with MP because not all plugins and codecs are aware of MP well. Also, do not have your FAT disk cache. He has no sense cluttering up your available disk space and you strip total system bandwidth with a large number of redundant or unnecessary reading write cycles.

• Addition of SMP processors is not always advantageous

  Hello

  I was going to the best practices for vmware virtual SMP technology.

  However, I stumbled on the fact that the addition of processors to the smp system still does not flow. I have read the reasons explained why this is so.

  But I could not well what it is all new to me,

  Please could someone explain to me in simple why words isn't always advantageous to continue to add more CPU to an SMP system.

  Thank you.

  I try to keep things simple, while planning for the CPU isn't really simple

  All processors/cores are shared between the host itself and all virtual machines and all of them need sufficient CPU cycles to work with the performance you expect. If you assign multiple vCPUs to a virtual machine, the UC Planner must ensure that the number of virtual processors is available at the same time to execute commands. For example, assuming that you assign 4 vCPUs, the Planner must wait until the 4 cores are available. To achieve this, you need to pause the first carrots until the 4th is available. Meanwhile suspended hearts are not available for other processes.

  In addition, vCPUs more generally does not add much performance to a virtual machine unless the applications running on the virtual machine are optimized to use multiple processors.

  For more information, see http://www.vmware.com/resources/techresources/10131

  André

• Get information from virtual hardware in Web mode

  Hello

  I have a problem in my workflow Web view.

  CPU and RAM to a virtual computer workflow changes.

  In an instant, the workflow looks like this:

  

  What I want to do is insert the CPU and the RAM of the current elected VM here:

  

  But I have no idea how do. I hope someone here can help me.

  MFG

  Patrick

  SERVUS Patrick!

  You can set "default values" in the presentation of entry of the workflow. (there is no way to include information in the tree-search-browser (from the less any documented way).)

  Here you can perform Actions, which returns the current vCPUs and RAM to the virtual machine and fills the value in the input field after selecting the virtual machine.

  to do this:

  1. create an action that returns the memory (and a second for the number of processors) of a given virtual machine.

  

  2 change the workflow, go to the presentation tab, select the numberCPU input parameter, click on the "Add property" icon in the lower section, select 'default value '.

  

  3 select the fall to the low fields to the symbolized by an arrow (ognl expression) and click on the puzzle on the right icon to select the action. filter for the name of your action, then select the 'VM' input parameter as a parameter for the action

  

  

  

  

  4. the wizard should "make" the appropriate expression of ognl

  

  5. save / finishing

  Tips: Think of the Guest OS HAL when switching between 1 and multiple vCPUs and prove the value of memory the user inserts is divisible by 4... (but not orchestrator, vsphere-questions)

  Grüße aus dem Allgäu!

  Jörg

  clicked 'send message' too early

• in vm CPU cores

  Hello

  Running windows 2003 standard 64-bit vm and try to get allows them to use multiple cores CPU instead of multiple vcpus.

  As a test, I gave the vcpu image 4 x and it displays correctly 4 graphics processor in the Manager of tasks and 4 processors in Device Manager.

  I also tried setting the cpuid.coresPerSocket using different values (2, 4 etc.) but that does not increase the amount of graphics and the specified processors.

  I it is dropped back to 1 and 2 vcpu and tried carrots by socket yet still no extra processor graphics.

  Is there anything else I should do as the plan is to get 4 cores for a sql database instead of having to use several vcpu for licensing reasons.

  dhedges wrote:

  Running windows 2003 standard 64-bit vm and try to get allows them to use multiple cores CPU instead of multiple vcpus.

  The physical representation of a vCPU is a kernel.

  As a test, I gave the vcpu image 4 x and it displays correctly 4 graphics processor in the Manager of tasks and 4 processors in Device Manager.

  I also tried setting the cpuid.coresPerSocket using different values (2, 4 etc.) but that does not increase the amount of graphics and the specified processors.

  That's right, as you do not increase the number of vCPU'. Example: If you have a host with 4 dual-core CPUS (8 cores) and run Windows 2003 SE on it, get physical Windows is reported that there are 4 CPU with 2 hearts of each. Even that Windows 2003 SE is limited to 4 CPUS, it would still run on all 8 cores.

  In a virtual machine, it gets reported that you assigned. Suppose that 4 vCPU. If Windows thinks that it works on 4 CPUS with the nucleus of one (!). So now the physical box that he must deal with a maximum of 4 cores (out of 8).

  Change that and let the guest to run on all 2 4 CPU cores, you must assign 8 vCPU for prompt and set parameter 2. Then Windows will run on 8 cores because he sees 4 CPU with 2 hearts of each. And Windows 2003 himself is limited by the CPU is not carrots.

  Read this: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1010184

  AWo

  VCP 3 & 4

  \[:o]===\[o:]

  = You want to have this ad as a ringtone on your mobile phone? =

  = Send 'Assignment' to 911 for only $999999,99! =