Comments for An enjoyed kernel apprentice http://blog.coly.li Just another WordPress weblog Sat, 21 Aug 2010 15:19:37 +0000 hourly 1 http://wordpress.org/?v=3.0.1 Comment on BeijingLSF 2009 by jebtang http://blog.coly.li/?p=24&cpage=1#comment-248 jebtang Sat, 21 Aug 2010 15:19:37 +0000 http://blog.coly.li/?p=24#comment-248 Hi, just wait for the blog for LSF 2010... Hi, just wait for the blog for LSF 2010…

]]> Comment on Random I/O — Is raw device always faster than file system ? by jebtang http://blog.coly.li/?p=87&cpage=1#comment-247 jebtang Sat, 21 Aug 2010 15:17:58 +0000 http://blog.coly.li/?p=87#comment-247 This question indeed been asked million time. For testing result, the bufferI/O(page cache) .vs. Driect I/O, the buffer I/O should show better result than direct I/O on the file system level testing. The ext3 in your testing did show this, but ext4 don't have this behavior. Why? Also, all of testing is read only operation, the page cache almost no use if you didn’t enable the readahead on the block device layer. I may find a choice to do the compare with read/write mix operation to see the result. Also, may try to use the iozone to control the I/O block size to see any effect on the result. Anyways, this is a good approaching to do no-philosophical problem  This question indeed been asked million time. For testing result, the bufferI/O(page cache) .vs. Driect I/O, the buffer I/O should show better result than direct I/O on the file system level testing. The ext3 in your testing did show this, but ext4 don’t have this behavior. Why? Also, all of testing is read only operation, the page cache almost no use if you didn’t enable the readahead on the block device layer.

I may find a choice to do the compare with read/write mix operation to see the result.
Also, may try to use the iozone to control the I/O block size to see any effect on the result.
Anyways, this is a good approaching to do no-philosophical problem 

]]> Comment on Don’t waste your SSD blocks by Tamás Kiss http://blog.coly.li/?p=124&cpage=1#comment-218 Tamás Kiss Tue, 03 Aug 2010 21:58:44 +0000 http://blog.coly.li/?p=124#comment-218 Hi! Thank you for the interesting post. I learned much from it. I noticed though that there is a small error in the post: you mention the '-N' option that specifies the number of inodes for the filesystem, but then in the example at the end you write '-I 16' instead of '-N 16'. Maybe while writing the example you were thinking of another space wasting filesystem feature: the possibility to create large inodes for storing extended attributes. If someone doesn't use extended attributes, then it is pointless to create the filesystem with large inodes. According to the mkfs.ext3 man page the smallest possible inode size is 128 bytes if you don't use extended attributes or any other feature that requires the extra inode space. Hi!

Thank you for the interesting post. I learned much from it.

I noticed though that there is a small error in the post: you mention the ‘-N’ option that specifies the number of inodes for the filesystem, but then in the example at the end you write ‘-I 16′ instead of ‘-N 16′. Maybe while writing the example you were thinking of another space wasting filesystem feature: the possibility to create large inodes for storing extended attributes. If someone doesn’t use extended attributes, then it is pointless to create the filesystem with large inodes. According to the mkfs.ext3 man page the smallest possible inode size is 128 bytes if you don’t use extended attributes or any other feature that requires the extra inode space.

]]> Comment on Taobao joins open source by Joris Willems - Amazing Things http://blog.coly.li/?p=112&cpage=1#comment-191 Joris Willems - Amazing Things Sat, 24 Jul 2010 13:33:28 +0000 http://blog.coly.li/?p=112#comment-191 Congratulations from Belgium! Congratulations from Belgium!

]]> Comment on Taobao joins open source by richard http://blog.coly.li/?p=112&cpage=1#comment-190 richard Sat, 24 Jul 2010 10:56:09 +0000 http://blog.coly.li/?p=112#comment-190 This is fantastic news - really great news. For those of us who believe strongly in open software and in China, this is a great day. This sort of move will do more than any campaign I can imagine to promote openness in Chinese development and, at the same time give devs in China a home grown foss product to prove that not only can foss software find a place in China, it can come from here. Congratulations Taobao... you show 'em all! This is fantastic news – really great news.

For those of us who believe strongly in open software and in China, this is a great day. This sort of move will do more than any campaign I can imagine to promote openness in Chinese development and, at the same time give devs in China a home grown foss product to prove that not only can foss software find a place in China, it can come from here.

Congratulations Taobao… you show ‘em all!

]]> Comment on Taobao joins open source by Pockey http://blog.coly.li/?p=112&cpage=1#comment-177 Pockey Tue, 13 Jul 2010 11:42:25 +0000 http://blog.coly.li/?p=112#comment-177 Congratulations! It's definitely a good news for all of us :) Congratulations! It’s definitely a good news for all of us :)

]]> Comment on Taobao joins open source by Suresh Jayaraman http://blog.coly.li/?p=112&cpage=1#comment-167 Suresh Jayaraman Thu, 01 Jul 2010 05:17:49 +0000 http://blog.coly.li/?p=112#comment-167 It seems your efforts have impacted positively and started paying off (though a little first step, it matters a lot sometimes). Way to Go!! It seems your efforts have impacted positively and started paying off (though a little first step, it matters a lot sometimes). Way to Go!!

]]> Comment on Taobao joins open source by colyli http://blog.coly.li/?p=112&cpage=1#comment-166 colyli Thu, 01 Jul 2010 04:26:36 +0000 http://blog.coly.li/?p=112#comment-166 Nikanth, thanks for the comments :-) This is a great team work, many people contribute valuable efforts to make it come to true. Let's see whether there are more positive news/contributions coming. Nikanth, thanks for the comments :-)
This is a great team work, many people contribute valuable efforts to make it come to true. Let’s see whether there are more positive news/contributions coming.

]]> Comment on Taobao joins open source by Nikanth http://blog.coly.li/?p=112&cpage=1#comment-165 Nikanth Thu, 01 Jul 2010 04:06:58 +0000 http://blog.coly.li/?p=112#comment-165 Wow! You rock. Wow! You rock.

]]> Comment on a conversation on DLM lock levels used in OCFS2 by TaoMa http://blog.coly.li/?p=81&cpage=1#comment-159 TaoMa Thu, 17 Jun 2010 04:48:46 +0000 http://blog.coly.li/?p=81#comment-159 hi coly, I think it is really not easy for a starter to understand the whole stuff of dlmglue. So thanks for the sharing. ;) hi coly, I think it is really not easy for a starter to understand the whole stuff of dlmglue. So thanks for the sharing. ;)

]]>