<div dir="ltr">I think you guys always keep forgetting about the gremlins.<br><br><br><div class="gmail_quote">On Tue, Jan 12, 2010 at 1:25 PM, Eli Billauer <span dir="ltr"><<a href="mailto:eli@billauer.co.il">eli@billauer.co.il</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div class="im">Etzion Bar-Noy wrote:<br>
<br>
> Sequential IO is very simple, relatively, so that you will hardly feel<br>
> the performance impact testing it.<br>
> Test random IO loads with small packets (0.5K-4K) and you will<br>
> probably feel the performance impact there.<br>
><br>
</div>It crossed my mind that a sequential test was too simplistic. My ideal<br>
test would be a database server running like crazy on the data, but this<br>
is not my application anyhow. I agree that I should test the system<br>
"under load" but I can't really think about a realistic "under load"<br>
scenario for 4 hyperthreaded CPUs. At least not something that I could<br>
test without turning it into a project. So I went for the session below.<br>
<br>
/usr is unencypted. /secret is encrypted. My home directory sits on a<br>
740 GB ext4 partition, which has seen some activity. /secret and /fresh<br>
are 20 GB, and are pretty untouched.<br>
<br>
The whole system will be reinstalled pretty soon, so I don't care about<br>
being root and fooling around. ;)<br>
<br>
My finding is that encryption still speeds up writes. This is really<br>
weird. I swear that /fresh was created a minute before I used it on this<br>
test, so I suppose it can't be in a better shape. Still, writing to<br>
/secret is faster. The only thing that was slower was removing a lot of<br>
files in one go.<br>
<br>
It's also interesting to note that writing to my home directory was<br>
considerably slower. So even if encryption turns out to have some<br>
impact, other factors (partition size? Fragmentation?) seem to be more<br>
dominant.<br>
<br>
So while I still agree with the concept that encryption must surely slow<br>
down disk writes, my tests show me the contrary.<br>
<br>
Insights, anyone?<br>
Eli<br>
<br>
---------- Session follows ---------------<br>
<br>
[root@short eli]# tar -c /usr | { cd /secret/delete-this/ && time tar -x ; }<br>
tar: Removing leading `/' from member names<br>
tar: Removing leading `/' from hard link targets<br>
<br>
real 5m12.709s<br>
user 0m2.351s<br>
sys 0m31.486s<br>
[root@short eli]# time rm -rf /secret/delete-this<br>
<br>
real 0m13.319s<br>
user 0m0.180s<br>
sys 0m5.154s<br>
[root@short eli]# tar -c /usr | { cd /home/eli/delete-this/ && time tar<br>
-x ; }<br>
tar: Removing leading `/' from member names<br>
tar: Removing leading `/' from hard link targets<br>
<br>
real 7m0.826s<br>
user 0m2.344s<br>
sys 0m32.340s<br>
[root@short eli]# time rm -rf /home/eli/delete-this<br>
<br>
real 0m8.195s<br>
user 0m0.196s<br>
sys 0m5.620s<br>
[root@short eli]# tar -c /usr | { cd /fresh/delete-this/ && time tar -x ; }<br>
tar: Removing leading `/' from member names<br>
tar: Removing leading `/' from hard link targets<br>
<br>
real 5m42.639s<br>
user 0m2.332s<br>
sys 0m36.879s<br>
[root@short eli]# time rm -rf /fresh/delete-this<br>
<br>
real 0m9.612s<br>
user 0m0.203s<br>
sys 0m6.275s<br>
<font color="#888888"><br>
--<br>
</font><div><div></div><div class="h5">Web: <a href="http://www.billauer.co.il" target="_blank">http://www.billauer.co.il</a><br>
<br>
_______________________________________________<br>
Haifux mailing list<br>
<a href="mailto:Haifux@haifux.org">Haifux@haifux.org</a><br>
<a href="http://hamakor.org.il/cgi-bin/mailman/listinfo/haifux" target="_blank">http://hamakor.org.il/cgi-bin/mailman/listinfo/haifux</a><br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br>Maxim Kovgan<br>
</div>