Re: Re: Re: open-source virtualization software...and moving off-topic

["paul.cour1" <user=1263@nabble.blu.org>]

 Kent Borg said: 
I think the assumed requirements of what an OS does is in need of change. 

what other things would you change, given the resources to do it ??? 

thanks 

paulc 


>From:  <[hidden email]> 
>Date: 2007/12/14 Fri PM 05:18:38 CST 
>To: Kent Borg <[hidden email]>, Eugene Gorelik <[hidden email]> 
>Cc: Blu <[hidden email]> 
>Subject: Re: Re: open-source virtualization software...and moving off-topic 

> 
>>From: Kent Borg <[hidden email]> 
>>Date: 2007/12/14 Fri PM 05:18:38 CST 
>>To: Eugene Gorelik <[hidden email]> 
>>Cc: Blu <[hidden email]> 
>>Subject: Re: open-source virtualization software...and moving off-topic 
> 
>>Eugene Gorelik wrote: 
>>> I am considering different virtualization software to create isolated 
>>> environment for Development and QA and let developers shell access to the 
>>> virtual environments. 
>>> I am comparing Xen and OpenVZ. They provide different level of 
>>> virtualization and OpenVZ is like an advanced jail, but I really like it 
>>> because it has minimal performance overhead comparing to Xen. 
>> 
>>Xen is brittle.  Choose something where the client is more independent 
>>of the host.  A host upgrade in Xen can easily break your existing 
>>clients because both the host and client have to run matched kernels.   
>>The fact that Openvz is Linux-only makes me worry it might also have a 
>>too-tight coupling between host and guest. 
>> 
>>Being completely virtual is a real nice feature.  It lets you run old or 
>>obscure stuff without any pesky version-skew issues. 
>> 
>>If I had KVM-compatible hardware I would run it, as it is I don't, so I 
>>chose what I think is the next closest thing: straight Qemu. 
>> 
>> 
>>Somewhat off topic... 
>> 
>>The popularity of virtualization brings up an interesting point.  To me 
>>it indicates a failure of Linux, et al. 
>> 
>>Linux offers some very cool features and APIs, yet the really *hot* API 
>>in the year 2007 is...<drum roll please>...a slightly evolved, bastard 
>>version of the original BIOS-architecture IBM hastily put in their 
>>original PC.  This is a design from over a quarter century ago, it was 
>>made for an 8-bit CPU, that ran at under 5 MHz, and could not have an 
>>entire megabyte of RAM.  I do note, however, that, as old as the PC is, 
>>the fundamental design of Unix (which Linux largely copies), is even older. 
>> 
>>Unix/Linux is a multiuser design, the users are protected from each 
>>other.  So why isn't it good enough?  What is lacking from the services 
>>offered by the OS that so many of us are drooling over a simulated raw 
>>IBM PC?  What is so wrong with the OS that, as bloated as the OS can be, 
>>we are wanting to run multiple copies on a single computer? 
>> 
>>I think the assumed requirements of what an OS does is in need of change. 
>> 
>>Note that using virtualization as a way to host old software *is* pretty 
>>compelling, but people doing new engineering are using virtualization in 
>>their *NEW* designs!  Something is rotten in the state of our OS 
>>assumptions. 
>> 
>> 
>>-kb, the Kent who frequently laments the lack of innovation in the 
>>tradition-bound computer world. 
>> 
>> 
>>P.S.  It is almost as if shared libraries were the big mistake, that the 
>>savings came at such a high price in version hell, that we are willing 
>>to duplicate the whole OS to dig ourselves out.  This is clearly not the 
>>entire story, but it is part of it. 
>> 
>>-- 
>>This message has been scanned for viruses and 
>>dangerous content by MailScanner, and is 
>>believed to be clean. 
>> 
>>_______________________________________________ 
>>Discuss mailing list 
>>[hidden email] 
>>http://lists.blu.org/mailman/listinfo/discuss