1 |
Hi Jeroen, |
2 |
on Saturday, 2006-04-22 at 22:26:43, you wrote: |
3 |
> >Lacking any benchmarks from before the change, that's the best measure I |
4 |
> >have. It might have been the firmware upgrade, but I doubt it. If I have |
5 |
> >to reboot any time soon, maybe I'll do another test to verify it. |
6 |
> > |
7 |
> Rrright.. that's the big disadvantage of having to sort this stuff out |
8 |
> on a production machine :) |
9 |
> We're only now starting to convince (read: force) our developers *not |
10 |
> *to put live code on production systems /before /we have had a chance to |
11 |
> test it. |
12 |
> You know how it goes... |
13 |
|
14 |
Sure...actually I tested most of the installation on another machine |
15 |
before switching it over to the server, but it didn't have quite the |
16 |
same hardware. This is a fairly small university department so we just |
17 |
can't justify a second ProLiant box for the occasional upgrade :) |
18 |
|
19 |
> Still only 300mbits, though... a server like that should be able to |
20 |
> handle SSH at near-line speed. |
21 |
|
22 |
Yup...that one hasn't been faster before either though. |
23 |
|
24 |
> >OK, that was a bit misleading, I meant that even assuming things about |
25 |
> >the flat file the scheduler sees of the disk like that offsets in the |
26 |
> >file sort of linearly correspond to cylinders |
27 |
> That *may* be true for old IDE drives, but it isn't even remotely true |
28 |
> for SCSI, which is its own higher-level abstraction on top of the |
29 |
> physical drive interface already, |
30 |
|
31 |
Huh? OK, block remapping for error recovery and stuff, but usually you |
32 |
can observe in benchmarks how the transfer rate corresponds to the |
33 |
offset, just like you'd expect it if the offset corresponded to the |
34 |
cylinder. |
35 |
|
36 |
> not to mention the layer the SmartArray puts on top of /that/. |
37 |
|
38 |
Sure. That's why I was thinking of the NOOP scheduler. |
39 |
|
40 |
> >---which is what it does to |
41 |
> >implement things like the elevator algorithm---are virtually always |
42 |
> >right for simple drives but may not be for a RAID. |
43 |
> > |
44 |
> Erm.. that's what I said :) |
45 |
|
46 |
Me too :) Just haven't benchmarked NOOP yet but I'm gonna try. |
47 |
Thanks for your input, I'll ask if we can afford the BBWC module...not |
48 |
that anybody was crying for it, but an admin needs a new toy once in a |
49 |
while as well! |
50 |
|
51 |
cheers! |
52 |
Matthias |
53 |
-- |
54 |
I prefer encrypted and signed messages. KeyID: FAC37665 |
55 |
Fingerprint: 8C16 3F0A A6FC DF0D 19B0 8DEF 48D9 1700 FAC3 7665 |