Hi all,
A few weeks ago, we did some additional tests on the Intel Atom processor to see if it was suitable to use as an Rsync Server. It’s possible to get inexpensive PCs with low power consumption and small footprint cases for under AUS $500 (US $400) – and we felt they could make good Rsync servers.
Previously we noted that the ARM processors in many entry-level NAS devices were insufficient in their computing power. But how does the Atom perform?
Stress test results – single 42GB file
Intel Atom – Vista / cwRsyncServer:
Initial time taken: 1h8m
Update: 1h19m (see note)
QNAP – ARM / Embedded Linux:
Initial time taken: 7h55m
Update time taken: 4h57m
Intel Dual Core – Ubuntu Linux:
Initial time taken: 1h22m
Update time taken: 0h35m
So the Atom performs far better than the embedded NAS devices. We traced the cause of the slower update time to slow Disk I/O, and having to copy the old file to apply the in-file changes. We’re seeing if we can speed that up further.
Conclusion: The Atom based PC is suitable as an Rsync server, and is far cheaper than the equivalent NAS.
Note: Our test here is Rsyncing a single 42GB file with approximately 2MB of changes between the initial upload and the update. This is most definitely a stress test and the average user will not place such demands on the backup. Rsync performs very well on a typical file system that has lots of small files. It starts struggling for massive files – and that is why we ran this particular test. On typical filesystems that don’t have massive files, the difference in performance between an Atom and Core Duo processor may even be too small to notice.
4 thoughts on “Rsync performance on different processors – Atom performs quite well”
hmm, when I try to comment this post, your stupid webserver claims:
“You tried to run some Request/Code, which seems to aim to damage this Website, therefore the Request was blocked!
If you didn’t run any Attack please notice the Webmaster about your Blocking.”
Could you look into it?
I tried to post the following message, which I included in base64 since I guess your blog-tool triggers on some character I have in the post. You may want to base64 decode the message and try posting it to your blog yourself to see what triggers the problem:
== BEGIN ==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== END ===
Oh, and I realized that your blog tool doesn’t even seem to handle the english alphabet plus digits correctly, since it *transformed* one of the characters in the base64 post above. The character preceding the substring “5IH” should be the letter “x” (0x78), but the blog-program translated it to 0xD7. Might want to submit a bug report to whomever created the web app…
And the same exact invalid substitution again preceding the substring “5Ii”..
That should be the last of it. Over and out..