[rescue] emulation fun

Joshua Boyd jdboyd at jdboyd.net
Mon Feb 2 09:08:22 CST 2009


On Mon, Feb 02, 2009 at 09:22:50AM -0500, Lionel Peterson wrote:

> Execution of code doesn't equate to capable of handling the same  
> workload. Mainframes have I/O throughput that PCs can't touch.
> 
> It would be interesting to identify/calculate the lag where a PC can  
> equal the throughput of a particular mainframe model/series. (for  
> example, a current model Dell 2950 8-core Xeon, SAS drives equals the  
> performance and throughput of an IBM 3090/200)

What was the bandwidth of each of the channels on the 3090?  From a
least one paper, I see the figure of 402 IOPS across all the DASD, but
it doesn't specify the size of each IO.  

I don't see why IBM can't be more straightforward about their machine
performance though...

I'm sure that PCs trail mainframe bandwidth by quite a few years, but I
find it a bit hard to believe that a few SAS drives can't deliver more
IOPS and bandwidth than an over 20 year old mainframe.



More information about the rescue mailing list