From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Venefax" Subject: RE: Windows SMP Date: Sun, 28 Dec 2008 21:48:44 -0500 Message-ID: <03e401c9695f$f7ebb8c0$e7c32a40$@com> References: <036501c96898$959aaef0$c0d00cd0$@com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============2033863167==" Return-path: In-Reply-To: Content-Language: en-us List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: 'Dirk Utterback' Cc: xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org This is a multipart message in MIME format. --===============2033863167== Content-Type: multipart/alternative; boundary="----=_NextPart_000_03E5_01C96936.0F15B0C0" Content-Language: en-us This is a multipart message in MIME format. ------=_NextPart_000_03E5_01C96936.0F15B0C0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit The problem is not SMP, is ACPI. I installed a non-ACPI Hal, but SMP capable, and the performance went right up with 4 virtual processors. I hope the developers can look into this mess. Federico From: Dirk Utterback [mailto:dirk.utterback@gmail.com] Sent: Sunday, December 28, 2008 9:40 PM To: Venefax Cc: xen-devel@lists.xensource.com Subject: Re: [Xen-devel] Windows SMP It seems something wrong. It is hard to achieve linear performance with more processors, but it should be at least something "normal". Another choice might be halsign turbogate drivers. I used it half a year ago, and it is stable and performance is really nice. 2008/12/27 Venefax I think that we need to tackle the absurd loss of performance that any Windows HVM incurs with more than one processor. I have measured it and it is amazing. I have an application that opens 15 independent network clients, and with one single processor and the Standard PC Hal, it takes a little over a minute to load the 15 clients and also to register with the SIP provider. If I leave that unchanged, I mean "ceteris paribus" , and only change the HAL to ACPI and add 8 processors, then the same operation takes over 10 minutes. Hello!!! I am using the GPLVL drivers from James Harper. Right now I am about to test the MPS -Non-ACPI HAL, but the driver still create a BSOD. Maybe the problem is only ACPI, and not MPS, who knows. In any case, what can we do so any Windows virtual machine can achieve a "normal" performance, I mean, a linear performance when using more processors and not this declining curve that can kill any host. Federico _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel ------=_NextPart_000_03E5_01C96936.0F15B0C0 Content-Type: text/html; charset="US-ASCII" Content-Transfer-Encoding: quoted-printable

The problem is not SMP, is ACPI. I installed a non-ACPI = Hal, but SMP capable, and the performance went right up with 4 virtual = processors.

I hope the developers can look into this = mess.

Federico

From:= Dirk = Utterback [mailto:dirk.utterback@gmail.com]
Sent: Sunday, December 28, 2008 9:40 PM
To: Venefax
Cc: xen-devel@lists.xensource.com
Subject: Re: [Xen-devel] Windows SMP

It = seems something wrong. It is hard to achieve linear performance with more processors, = but it should be at least something "normal". Another choice = might be halsign turbogate drivers. I used it half a year ago, and it is stable = and performance is really nice.

2008/12/27 Venefax <venefax@gmail.com>

I think that we need to tackle the absurd loss of performance that = any Windows HVM incurs with more than one processor. I have measured it and = it is amazing. I have an application that opens 15 independent network = clients, and with one single processor and the Standard PC Hal, it takes a little = over a minute to load the 15 clients and also to register with the SIP = provider. If I leave that unchanged, I mean "ceteris paribus" , and only = change the HAL to ACPI and add 8 processors, then the same operation takes over 10 minutes. Hello!!! I am using the GPLVL drivers from James Harper. Right = now I am about to test the MPS –Non-ACPI HAL, but the driver still = create a BSOD. Maybe the problem is only ACPI, and not MPS, who knows. In any case, = what can we do so any Windows virtual machine can achieve a "normal" performance, I mean, a linear performance when using more processors and = not this declining curve that can kill any host.

Federico

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.c= om
http://lists.xensource.com/xen-devel

=

------=_NextPart_000_03E5_01C96936.0F15B0C0-- --===============2033863167== Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Disposition: inline _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel --===============2033863167==--