Older blog entries for rcastro (starting at number 11)

27 Jun 2003 (updated 27 Jun 2003 at 02:28 UTC) »
SBAC paper is done. Now waiting for a review, but I don't think I will spend much more time on it. Thanks God!

Only the minor changes must be implemented for the final version of my dissertation. This time I have to fix the errors found and pointed out by other member of the committee. Well, I actually just tried to fix these errors and in 15 minutes it was done. Very quick. Only some suggestions left from the first committee member and we are done with this dissertation. Next week I will be finally free from my latex programming job, and will be able to code again.

Tomorrow I guess I will work on XSL stuff at my job. In particular, Xalan and Saxon implementations of XSLT specifications. I hope it will be fun.

Now working on the final version of my dissertation, fixing typos, graphics and a few corrections. I am eager to get finally rid of it in order to keep implementing the compressed cache. I guess I will end up spending most of my time this week on the modifications suggested by the committee. Besides finishing the dissertation, I intend to publish the latest performance results on the compressed caching website this week. (Note: next Thursday is a catholic holiday in Brazil).

This weekend I started reading The Fountainhead, by Ayn Rand, a great book. It is the second time I read this novel, and I think it will be worth reading it one more time, although I will have to leave my LOTR reading for a while.

First draft of my 8 page paper for SBAC (Simpósio Brasileiro de Arquitetura de Computadores, or, the English name, Symposium on Computer Architecture and High Performance Computing) is ready. Now I have to review it before sending it to my advisor for comments.

Unfortunatelly couldn't manage to find time to finish the new statistics for compressed caching website.

Debian installation with PPPoE connection
Just out of curiosity, a friend of mine wanted to install Debian in a laptop connected to a PPPoE ADSL internet connection, which isn't possible with the default boot floppies. He asked for help tonight, and I ended up finding out that this website has special floppies for this case. I have to give it a try in the future (hopefully Debian will be released soon, and this option will be available in the next stable release).

The defense of my master's dissertation on compressed caching took place on Jun 06th at Jacy Monteiro Amphitheater of University of Sao Paulo, at 2PM BRT. The audience wasn't very large, but it was composed of high quality people :-). I guess the presentation was pretty fine, but it took a little longer than I expected. The questions about the project were clever, but no surprise. All the committee members liked the dissertation and my master was approved. After the defense, some close friends and I celebrated it going to a pizzeria, where we ate a lot. Now I have to fix some mistakes and check some suggestions in order to deliver the final version of my text and make it available on my web site (the text is in Portuguese).

Besides this task, I am spending some time on two paper submissions (in English). I intend to do (a brazilian symposium and an american workshop), that is why my code development isn't going very well. Oh, gosh, I guess I have to stop working at my job (would there be anyone out there willing to sponsor my work? ;-).

Time to update compressed caching website with new performance statistics.

Finished all the OSDL DBT1 tests, and noticed that compressed caching does not help this scenario. It does not access disk so much, and disk is not its bottleneck. That is a pity, because such an application would be nice to have a performance gain.

I also could run OSDB, the Open Source Database Benchmark. I mean, I had already run it, but now I could run in the multithreaded disabling the creation of hash indexes in Postgres. Tiny performance gains. Not a good day for my new performance tests.

In the night I left the computer running the complete test suite with a compressed cache which used WK4x4 compression algorithm (download here), the algorithm from Wilson-Kaplan compression algorithm family for process data segment.

I still have to run some tests for a meeting with my advisor today. I guess we are pretty close to finishing all the tests and we should put a great effort into making the dissertation good enough for submission. It is a pain in the ass to be stuck with this, unable to spend time developing interesting stuff.

For people who does not love hot weather like me, check the weather forecast in Sao Paulo. It is 27 degrees Celsius (81F) at 9h AM. In the room I am now, probably it is over 30C (86F).

More than two months after the last entry, I am back. At the moment I write, I am running OSDL database benchmark 1, DBT1 to verify my project's performance. That will be used for my dissertation and also to a paper that I intend to submit (if I have enough time) to SOSP 2003.

Since my last entry, I finished and submitted a paper to Usenix Annual Technical conference (on Nov 18th 2002), but it was rejected (on Jan 22nd 2003). One out of three reviewers did not read the paper carefully, and had some bad impressions about the project and how useful it might be. We think that this reviewer was the one to blame for the rejection, since the other ones wrote positive reviews of the paper. Anyway, there are interesting comments that can be applied to the text in order to improve it.

I am working really hard on my dissertation. Now, since I am working on the experimental results chapter, I am trying to run some new tests to add to my dissertation. Among them, DBT1, which took me quite a while to make it run succesfully. Other tests are piGIMP, httperf and a scientific application that runs in Matlab to calcule the fractal dimension of a image.

Something I've been doing lately is to search for a job. It is really hard to get a nice job. I would like a challenging job, where I could keep working with Linux and open-source software, but it is hard to get a job like this, in particular here in Brazil. I tried some jobs to work in USA, but they do not even reply (I guess they do not want foreigners without work visa). If you have a job for me, please get in contact :-)

Oh, I probably did not comment, but I went to USA to attend OSDI symposium. I got a stipend from Usenix and spend a week there, including 3 days in New York. Nice trip, except for a flu I got before getting to NY, which lasted for over a week.

In this meantime I also tried the IBM Linux Scholar Challenge, but once again I was not among the winners. I do not think my short paper was not commercial enough to convince people it was a nice piece of work.

On Friday, in the morning I worked on the paper and in the afternoon I bought new computer parts for my father and mounted the new computer. On Saturday, I spent the whole morning installing the new computer. Unfortunately the video card didn't work so well (the fan was noisy) and we will have to return it. Something very cool is the POST reporter present on the Asus A7V333 motherboard.

In the afternoon, I printed and reviewed the paper, sending it to the co-authors (my advisors). I was so tired that couldn't work that night.

On Sunday, I configured my home computer to run a set of tests on a kernel with compressed cache. I had to install postgres, osdb, mummer (these programs were copied from my computer at office using a CDRW). Unfortunately compiling the kernel to be tested with all profiling hung my computer at boot time. I wasn't lucky because my vmlinuz.old link wasn't set, so I didn't have a sane kernel to boot after that crash. I had to come to the office to take a bootable CD in order to fix it.

I also created a picture to the paper describing the lists within compressed cache and fixed some sections that there were things to be written yet.

About my monitor display, I sent an email to LG and also to Folha de São Paulo (a famous newspaper in the country) complaining about the geometry problems that the staff at LG claimed to be normal. I hope I can get it fixed or a new one without this undesirable "feature".

Reading the linux kernel mailing list on Sunday, I also found that that kernel 2.0 needs swap partitions with version 0 of swap header (2.2 and 2.4 use version 1). I guess this swap thing may be the reason why we couldn't make the compressed swap project work on a Debian potato system (it oopsed). I will try to rerun a test this week.

There are two weeks to go. It is already time to start running the final set of tests to plot the graphics. I want to send this paper to be reviewed by some people before submission.

On Wednesday and yesterday, I worked a lot on the paper I am writing. I intended to send a version for review of my advisors yesterday, but couldn't make it. I hope I can do that today.

Livio gave me a lot of suggestions on the paper that were very helpful yesterday, thanks.

My advisor had suggested me to test two new adaptability policies some days ago. We focused on the first one, that displayed the best results and in the Sunday night I tried to run a complete test suite run to evaluate its performance. However, it hung. On Tuesday, running that at home, I found out where the problem is located: it is hanging trying to locking a swap buffer. That means there is still some race related to these buffers that I have to check.

Today I got an email from an Usenix manager telling that the notifications about the scholarship to attend OSDI 2002 will be out on Nov 11th, 12th and 13th. I thought it was going to be quicker, but they say the delay is due to the number of applications.

Oh, about the display monitor I bought, it looks like the first one didn't have problems. After calling and almost discussing with LG Electronics support, I was told that the problem I reported was normal. It probably is, given that the second monitor had the very same symptom.

I finally found out what was hanging my mozilla browser: the sound wrapper. I wasn't using any sound server, and my sound card driver isn't that good, then when something else was using the sound device, the mozilla browser hang trying to lock /dev/dsp. I enabled arts sound server (I use kde) and since then I can open how many mozilla windows I want without problems. I am relieved.

I hope I can manage to have time to read "The Hobbit" this weekend. I am near the end and the book is getting each time better.

Yesterday night I watched "Revolution OS" in the Sao Paulo International Movie Festival. Very nice movie, but the end gave me the impression that "the dream was over". After the movie, there was a discussion with Marcelo Tosatti and Rodrigo Stulzer, both from Conectiva, about open source software. It had been such a while since the last time I met Marcelo. I noticed he is a little shy in a public presentation. By the way, there I met Breno, a close friend of mine.

During the day, I worked on the comments I got for my paper on compressed caching. I changed the simple ones and only started the ones that demand more intellectual work. I also bought a new display monitor for my home pc. However, I am not so lucky buying things and got one with problems. Today I am going to call the store where I bought in order to change it.

On Monday, I attended a lecture about a paper that a friend of mine is going to present at a workshop at OOPSLA 2002 and had a metting with my advisor to talk about some results from the tests I have been running.

I was unsubscribed from the linux kernel mailing list yesterday afternoon. It's not the first time that happens and I don't know why. I ended up subscribing again today. Along with this fact, I had my internet connection down until 2PM.

Yesterday I recalled why I had a benchmark suite with good performance results (that actually had puzzled me). There were some changes in the hash table size that could decrease the completion time in 1 second. I am going to try to improve that hash table size at home (I think it must be dependant on the maximum compressed cache size).

From the results from the suite I ran in the night from Thursday to Friday, I forgot about my policy to shrink the compressed cache if the last portion didn't have many accesses.

I had also some results from two new suggestions made by my advisor. The first one performed well, but not better in all cases. The second one wasn't good. I scheduled a benchmark suite to run the first suggestion, but we are not sure it's actually worth it.

And since yesterday, there is a benchmark suite running for static compressed caches. It takes quite a while to run (over 12 hours), but it's very complete. Those data will be needed in order to implement a third adaptability policy my advisor suggested. Anyway, on Monday I continue writing my paper on compressed caching (oh, I'd rather be programming :-).

2 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!