Years ago, I ran an IBM datacenter. We ran VM/370 with OS/VS1 as our production OS and CMS for our programmers to use. While not a realistic benchmark, we were converting our payroll from Burroughs to IBM. I ran payroll one night with OS/VS1 native, and the next night, nearly the same data with some corrections, except that OS/VS1 was running under VM/370 with a few logged in CMS users. If I recall correctly the payroll actually finished earlier under VM/370 earlier than native. Later I went to a SHARE conference. One presentor showed some benchmarking results where they were running DOS (IBM mainframe DOS) under VM/370 and they were able to achieve better throughput with DOS under VM/370 than native. In the case of VM/370, the big win was paging and spooling. I found that double spooling (using the OS native spooling which then spooled to VM/370) was actually faster than the spooling with the OS booted native. In the case of DOS and OS/VS1 there were many hacks both in VM/370 and the firmware on the 370 which provided some efficiencies. VM/370 paging was more efficient than the DOS or OS/VS1 paging. In any case, under Win4Lin, you are using the native Linux file system which is faster than the Windows VFAT file system. But, as above, the answer is throughput. Windows9x has some limitations due to its dependence on MS-DOS. A VM can mitigate some of these limitations. I'm just generalizing here because I don't have the more detailed knowledge I used to have back when I ran an IBM shop. On 3 Jun 2002 at 22:17, Anders Johansson wrote:
Explain please. I genuinely want to know. How could any program running in a virtual environment be as fast or faster than running on the bare hardware? If I had an application where speed was the #1 priority I wouldn't choose linux, I'd run it without any OS at all.
-- Jerry Feldman Enterprise Systems Group Hewlett-Packard Company 200 Forest Street MRO1-3/F1 Marlboro, Ma. 01752 508-467-4315 http://www.testdrive.compaq.com/linux/