[SPAM] About the differences between Version 3.3 and Version 3.0

[SPAM] About the differences between Version 3.3 and Version 3.0

[Fang Yuming]

[-- Attachment #1: Type: text/plain, Size: 1508 bytes --]


I want to do some research about the I/O performance of VMM and want to use 
Xen to do some experiments. I read many materials about the Xen. But there are 
some questions about the code of Xen. I know the virtual disk is implemented 
through the frontend driver and backend driver. But there are some questions I could 
not be sure about the virtual disk. I would appreciate it greatly if anyone could help me.

1. If there are many disk requests coming from different domains simultaneously, how 
would the backend driver of the domain0 process these requests? Does it processes them
through FIFO algorithm?

2. In version 3.0, where is the code about disk backend driver of domain0? Is it in 
linux-2.6-xen-sparse\drivers\xen\blkback? Do these codes include the algorithm processing 
disk requests from different domains?

3. I found the newest Version is Version 3.3, can anyone tell me the majors differences between 
Version 3.3 and Version 3.0? And is there any change in the part of virtual disk, especially 
the scheduling alogithm of different disk requests from domains?

I would really appreciate it if anyone could help me to answer tha above questions.
PS: I read some papers about the Xen and wanted to understand some parts of the code so that 
I could do some experiments using Xen. However, I found it was difficult, for there is few 
materials about the code. Could anyone tell me how I could be quickly familar with the code 
related to the virtual disk?





[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

[SPAM] About the differences between Version 3.3 and Version 3.0

[Fang Yuming]

[-- Attachment #1: Type: text/plain, Size: 1508 bytes --]


I want to do some research about the I/O performance of VMM and want to use 
Xen to do some experiments. I read many materials about the Xen. But there are 
some questions about the code of Xen. I know the virtual disk is implemented 
through the frontend driver and backend driver. But there are some questions I could 
not be sure about the virtual disk. I would appreciate it greatly if anyone could help me.

1. If there are many disk requests coming from different domains simultaneously, how 
would the backend driver of the domain0 process these requests? Does it processes them
through FIFO algorithm?

2. In version 3.0, where is the code about disk backend driver of domain0? Is it in 
linux-2.6-xen-sparse\drivers\xen\blkback? Do these codes include the algorithm processing 
disk requests from different domains?

3. I found the newest Version is Version 3.3, can anyone tell me the majors differences between 
Version 3.3 and Version 3.0? And is there any change in the part of virtual disk, especially 
the scheduling alogithm of different disk requests from domains?

I would really appreciate it if anyone could help me to answer tha above questions.
PS: I read some papers about the Xen and wanted to understand some parts of the code so that 
I could do some experiments using Xen. However, I found it was difficult, for there is few 
materials about the code. Could anyone tell me how I could be quickly familar with the code 
related to the virtual disk?





[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: [SPAM] About the differences between Version 3.3 and Version 3.0

[Samuel Thibault]

Fang Yuming, le Thu 28 Aug 2008 09:48:47 +0800, a écrit :
> 1. If there are many disk requests coming from different domains
> simultaneously, how would the backend driver of the domain0 process
> these requests? Does it processes them through FIFO algorithm?

They are just queued in the dom0 kernel, which schedules them just like
process requests.

> 2. In version 3.0, where is the code about disk backend driver of
> domain0? Is it in linux-2.6-xen-sparse\drivers\xen\blkback?

Yes.

> Do these codes include the algorithm processing disk requests from
> different domains?

No, Xen just lets linux do that, in block/*iosched.c

> 3. I found the newest Version is Version 3.3, can anyone tell me the majors differences between 
> Version 3.3 and Version 3.0?

See the announces of 3.1, 3.2 and 3.3.

> And is there any change in the part of virtual disk, especially the
> scheduling alogithm of different disk requests from domains?

Again, that's just dom0's Linux matter.

Samuel

[SPAM] Re: [SPAM] About the differences between Version 3.3 and Version 3.0

[Fang Yuming]

[-- Attachment #1: Type: text/plain, Size: 1837 bytes --]


----- Original Message ----- 
From: "Samuel Thibault" <samuel.thibault@eu.citrix.com>
To: "Fang Yuming" <fangyuming@emails.bjut.edu.cn>
Cc: <xen-devel@lists.xensource.com>
Sent: Thursday, August 28, 2008 4:51 PM
Subject: Re: [Xen-devel] [SPAM] About the differences between Version 3.3 and Version 3.0


> Fang Yuming, le Thu 28 Aug 2008 09:48:47 +0800, a écrit :
>> 1. If there are many disk requests coming from different domains
>> simultaneously, how would the backend driver of the domain0 process
>> these requests? Does it processes them through FIFO algorithm?
> 
> They are just queued in the dom0 kernel, which schedules them just like
> process requests.
> 
>> 2. In version 3.0, where is the code about disk backend driver of
>> domain0? Is it in linux-2.6-xen-sparse\drivers\xen\blkback?
> 
> Yes.
> 
>> Do these codes include the algorithm processing disk requests from
>> different domains?
> 
> No, Xen just lets linux do that, in block/*iosched.c

Does this mean that I should modify the dom0(linux) kernel( in block/*iosched.c) if I want to do some experiments 
about the disk scheduling algorithm of virtual disk scheduling algorithm?


> 
>> 3. I found the newest Version is Version 3.3, can anyone tell me the majors differences between 
>> Version 3.3 and Version 3.0?
> 
> See the announces of 3.1, 3.2 and 3.3.
> 
>> And is there any change in the part of virtual disk, especially the
>> scheduling alogithm of different disk requests from domains?
> 
> Again, that's just dom0's Linux matter.

From your answers, I know that if I want to do some research on virtual disk schedling algorithm, I should 
modify the disk scheduling part of linux kernel and it is no need to change other parts of Xen?

> 
> Samuel


Thank you very much for your help.

Fang Yuming

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

[SPAM] Re: [SPAM] About the differences between Version 3.3 and Version 3.0

[Fang Yuming]

[-- Attachment #1: Type: text/plain, Size: 1837 bytes --]


----- Original Message ----- 
From: "Samuel Thibault" <samuel.thibault@eu.citrix.com>
To: "Fang Yuming" <fangyuming@emails.bjut.edu.cn>
Cc: <xen-devel@lists.xensource.com>
Sent: Thursday, August 28, 2008 4:51 PM
Subject: Re: [Xen-devel] [SPAM] About the differences between Version 3.3 and Version 3.0


> Fang Yuming, le Thu 28 Aug 2008 09:48:47 +0800, a écrit :
>> 1. If there are many disk requests coming from different domains
>> simultaneously, how would the backend driver of the domain0 process
>> these requests? Does it processes them through FIFO algorithm?
> 
> They are just queued in the dom0 kernel, which schedules them just like
> process requests.
> 
>> 2. In version 3.0, where is the code about disk backend driver of
>> domain0? Is it in linux-2.6-xen-sparse\drivers\xen\blkback?
> 
> Yes.
> 
>> Do these codes include the algorithm processing disk requests from
>> different domains?
> 
> No, Xen just lets linux do that, in block/*iosched.c

Does this mean that I should modify the dom0(linux) kernel( in block/*iosched.c) if I want to do some experiments 
about the disk scheduling algorithm of virtual disk scheduling algorithm?


> 
>> 3. I found the newest Version is Version 3.3, can anyone tell me the majors differences between 
>> Version 3.3 and Version 3.0?
> 
> See the announces of 3.1, 3.2 and 3.3.
> 
>> And is there any change in the part of virtual disk, especially the
>> scheduling alogithm of different disk requests from domains?
> 
> Again, that's just dom0's Linux matter.

From your answers, I know that if I want to do some research on virtual disk schedling algorithm, I should 
modify the disk scheduling part of linux kernel and it is no need to change other parts of Xen?

> 
> Samuel


Thank you very much for your help.

Fang Yuming

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: [SPAM] Re: [SPAM] About the differences between Version 3.3 and Version 3.0

[Samuel Thibault]

Fang Yuming, le Thu 28 Aug 2008 21:49:29 +0800, a écrit :
> >> Do these codes include the algorithm processing disk requests from
> >> different domains?
> > 
> > No, Xen just lets linux do that, in block/*iosched.c
> 
> Does this mean that I should modify the dom0(linux) kernel( in block/*iosched.c) if I want to do some experiments 
> about the disk scheduling algorithm of virtual disk scheduling algorithm?

Since that's the component that does it, yes.

> >> 3. I found the newest Version is Version 3.3, can anyone tell me the majors differences between 
> >> Version 3.3 and Version 3.0?
> > 
> > See the announces of 3.1, 3.2 and 3.3.
> > 
> >> And is there any change in the part of virtual disk, especially the
> >> scheduling alogithm of different disk requests from domains?
> > 
> > Again, that's just dom0's Linux matter.
> 
> From your answers, I know that if I want to do some research on virtual disk schedling algorithm, I should 
> modify the disk scheduling part of linux kernel and it is no need to change other parts of Xen?

Indeed. You may want to make the block backend provide the domid to the
io scheduler for e.q. better fairness, however.

Samuel