2007年09月19日

プロトコル破綻

「パーティ出席確認の再送です。すでにお返事いただいている方は 無視してください」

それはわかったけど、返事した時その確認がされてないので 「こっちは送ったつもりになってるけど、そちらは受け取ってない (つもり)」というケースじゃない、とどうやって判断したら よかでしょうか??プロトコル的に破綻してないでしょうかね??

タグ:不可解
posted by oi at 10:14 | 福島 🌁 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年08月09日

USJ2

先日の こいつは何とかなった (ような気になっている) が、 今やっていることはもう少し複雑で自分自身の中で こなれていない事もあり時間がかかる。 (集中力の限界が低い、低くなっている、 という要因も多分にあろう)。

現実的な目標設定をしなおした方が賢明かと思い始めている。

posted by oi at 10:48 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年07月30日

オープンキャンパス

オープンキャンパスとやらに参加。 客がくるといいなぁ。
posted by oi at 22:56 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年07月25日

VMware、仮想環境でのベンチマークプログラムを提供開始

以前 VMWareのホームページで "VMmark: A Scalable Benchmark forVirtualized Systems (VMware white paper)"なんていうのをみつけて 眺めてはいたのだが、それがリリースされた、というニュースが。

ITmediaニュース
VMware、仮想環境でのベンチマークプログラムを提供開始

VMWareのNews Release
VMware Unveils Industry’s First Benchmark Designed to Measure Performance of Enterprise Workloads on Virtualized Environments

おもしろそうなので VMWareのサイトにいきながめていると VMmark System Requirements

One server with 2 CPUs, 6GB RAM, 80G disk space and a 1Gb NIC
むむむ、6GB
Purchased Software
You must purchase licenses for the following software packages if you do not already own them:
  • Microsoft Windows Server 2003 Release 2 Enterprise Edition (32-bit)—Four copies per tile (three for virtual machines and one for that tile’s client system)
  • Microsoft Exchange Server 2003 Enterprise Edition (This is available as part of the Microsoft Developer’s Network (MSDN) Universal subscription.)
だめだこりゃ。 とりあえず耐えがたきを耐え、忍びが忍ぶことにする。

Open Sourceの代替ソフトで似たようなものを作ってみるのは 面白い、という気がする。

posted by oi at 18:08 | 福島 ☀ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年07月24日

Challenge

I'll do a little challenge which is a kind of task meant by the title of this album:

タグ:Char U.S.J.
posted by oi at 01:02 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年06月15日

Some tips on distributed processing

For distributed processing work properly, we may need:
  • having 'trust-worthy' (could be a small number of) agents, rather than (possibly a large number of) 'randomly collected' guys,
  • give them enough privilledge to perform given tasks. otherwise, they all ways have to ask permission to the centralized authority,
  • simple and effient communication channel and protocol. unnecessary broadcasting makes each agent pay less attention to messages (even if some of them area really important).
  • multiple layers of safe-nets, rather than keep shouting 'be careful'
posted by oi at 20:43 | 福島 ☁ | Comment(1) | TrackBack(0) | しごと(おおまかに言えば)

2007年06月12日

paper review

外に出ていてもやはり「もってきた仕事」がある (言い換えると外でもできるから出かけられる、となるわけだが)。 とりあえず自分の分の査読終了。 締め切りは近いので「お知らせ」を送ったものの、 「柳に風」で結局締め切りまでに提出してくれなさそうな人も いそうな雰囲気が濃厚。
posted by oi at 04:40 | 福島 ☀ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年06月10日

Meeting with Prof. Cretu and Visit to Siemens

Today I had a meeting with Prof. Cretu, the Head of Departamentul "Stiinta si Ingineria Calculatoarelor", heard his presentation about the department, and discussed on possibile collaboration and exchange programs between us.

After the meeting, Christi, who is a PhD student at UPT and also a Siemens employee, took us for a visit to his office. Siemens Romania has quite new and sophisticated buildings. I saw the presentation of their development project.

w_prof_cretu.jpg w_christi.jpg siemens.jpg

posted by oi at 00:02 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年06月09日

お仕事その1終了

とりあえずメインのお仕事終了。 昨日も今日も大変暑い。 日本は8月が暑さのピークだが、ヨーロッパは日照時間と 暑さのピークが同期しているような気がしないでもない。
Timisoara の天気 (Yahoo Weather).
写真はポスターと ACSAの人々。

poster_upt.jpg acsa1.jpg acsa2.jpg

posted by oi at 01:04 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年06月08日

Finally at UPT

tsr_church.jpg

長い長い移動の果て、ついに到着。 宿に落ち着き一休みして Advanced Computing Systems and Architectures訪問中。

posted by oi at 02:03 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年05月30日

Today's harvest

harvest.jpg

These are what I got today for free (or in change for my business cards), but what was more important than these ad-stuff was to have a chance to talk with people in the industry, which is a hard to get oppotunity for the people like us living in isolated areas. The talk of the guy from The Linux Foundation was quite understandable and agreaable: just writing new codes cannot make linux as another choice for people using computers; rather it needs to make linux look more reliable and dependable. Also, there's a compapatibily issue with de facto standard. However, it is definitely better to have more choices than monopoly.

posted by oi at 21:28 | 福島 ☁ | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年04月20日

Lunch with COOL Chips Participants

Today, it was last day for COOL Chips conference. During the lunch break, we've been to a small Chinese restaurant called "李さんの店" which is near Yamashita Koen (Park). It is outside China Town and a place for people working in this area to have lunch and drink after work. I hope people coming from outside Japan (two from the US and one from Spain) enjoyed food and atomospher of such a non-touristic place.

lunch_apr20_2007_2.jpg lunch_apr20_2007.jpg

タグ:中華
posted by oi at 23:35 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年04月10日

I've lost ...

IveLost.jpg

I've lost 19337Yen for some unable to specify reason.

posted by oi at 18:27 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年03月19日

TPC-E

A new benchmark for OLTP workload from TPC has just been released. It is called TPC-E and it should replace the long time existing TPC-C. (it seems the link to "TPC-E PowerPoint Presentation" is incorrect).

In an article on Cnet ( Database servers get a new yardstick), it is said that

One problem with TPC-C is that high scores require an inordinate amount of storage--not just raw capacity, but numerous hard drives.
We have found that this is quite true. Probably, we have to invest around 10 times more money on the disk system than the CPUs to obtain interesting results. I hope TPC-E is a little more tractable for non-vendor researchers and engineers.

業界団体TPC、新サーバ性能テスト「TPC-E」を承認 (Japanese translation of the above article).

タグ:TPC-E Benchmark
posted by oi at 21:48 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年03月14日

昔とった...

問題大有りであったが、とりあえず『仕事』のヤマは越え、 自発的な『業務』の方に時間が費やせるようになる。 比較的新しい種類の workloadとして data miningに 興味を持ち NU-MineBenchで遊び始め、 論文 (pdf)を読んでみる。 Fuzzyだの Baysian Methodだの、昔 川崎のはずれでやってたことを思い出す。 ちょっとここら辺にまた手を出したくもなる。
タグ:Fuzzy ネタ 回顧
posted by oi at 13:55 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2007年03月06日

出張中

takino.jpg

昼間、ひょんなきっかけで連絡が取れるようになった 昔の知り合いの会社にお邪魔する。

夜、こちらもまた古い知り合いと飲みに (左端)。 ただしこちらはあまり長いブランクはなし。 研修で東京滞在とのこと。

タグ:再会 現実 T氏
posted by oi at 02:18 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2006年11月09日

DVDs from the Sunshine State

dvd1.jpg dvd2.jpg dvd_cap.jpg

I received small packet from Abdel, which were video recording of my seminar at USF Lakeland.

タグ:DVD Florida
posted by oi at 00:00 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2006年10月28日

Oprofile didn't work in the performance counter mode on a Pentium-D platform.

We have recently acquired a Pentium-D based server. We installed CentOS4.4 and Oprofile which is a workload measurement tool using performance counters in the processor. Oprofile works in two measurement modes: The timer Interrupt mode which is for processors without performance counters (such as processors for laptop systems) and the performance counter mode. Obviously, we need to use the latter mode.

From the past experience, we knew that LAPIC must be turned on to use oprofile in the performance counter mode. To do so, first we have to re-compile the kernel because the kernel that comes with RedHat based Linux implementations has LAPIC disabled (I saw some articles saying the using the kernel for SMP works, but in our system it didn't). Also, the use of LAPIC must be included in the boot parameters. For other machines (including a Pentium-D based server), these options made oprofile working in the performance counter mode.

Oddly enough, these options didn't work for the new Pentium-D based server. The kerlnel we used was 2.6.15.4. We then tried a newer kernerl (2.6.18), and it did work. However, we had another problem with 2.6.18: the iptables didn't work (at least with the filter rule file for 2.6.15.4). This is a serious threat for the security of the system. Just running the system with 2.6.18 for 30 minutes, we found many attacks to the system from outside. So, at least until we find a way to use iptables on 2.6.18, we have to stick with an older version.

I looked into the internal of oprofile. Oprofile creates a special file system under /dev/oprofile/ directory. /dev/oprofile/cpu_type indicates (as the name says) the CPU type and in the problematic system, it was "timer". We searched the source files of oprofile to find where this information is created, but could not find a clear answer. Later, I found that this file system was created and initialized not in the oprofile, but in the kernel: (linux-2.6.15.4/arch/i386/oprofile). So I started looking at the files in this directory.

Finally, I found where this problem was caused.
In linux-2.6.15.4/arch/i386/oprofile/nmi_init.c, there is a function "p4_init" which checks the model information for p4 family processors. There is a conditional statement:

if (cpu_model > 4) 
         return 0;
and "cpu_mode" is one of fields in a structure defined in
linux-2.6.15.4/include/asm-i386/processor.h
struct cpuinfo_x86 {
        __u8    x86;            /* CPU family */
        __u8    x86_vendor;     /* CPU vendor */
        __u8    x86_model;
...
}
After some literature seaerch, I found this field is specfied in Intel's document "Intel(R) Processor Idientification and the CPUID Instruction"

pentium.jpg

and also found that in the kernel 2.6.15.4 handled recent Pentium-D as 'illegal' cases. I realized 2.6.18 has this function corrected but 2.6.18. Assuming 'future models' do not have counters must be a safe choice but does not seem very smart. Also, Pentium-D 830 does not fall into "Pentium-D" categoty in this document. Once we have found a solution for iptables on 2.6.18, we will move to 2.6.18 kernel.

posted by oi at 23:50 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2006年09月15日

Visit to USF Lakeland

I am visiting University of South Florida Lakeland campus , where my friend Abdel Ejnioui is teaching IT courses as an assistant professor. I gave a seminar about my research as well as a brief introduction to our school. It was broadcast to another regional campus of USF in Sarasota.
(The announcement of seminar in Sarasota)

lakeland1.jpg lakeland2.jpg DrPonticell.jpg

After the seminar, I talked with Dr. Judith Ponticell, who is Associate Vice President in Academic Affairs, on the possible development of exchange programs between USF Lakeland and our school. Abdel made all arrangements for these activities of my visit.

タグ:USF
posted by oi at 07:32 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

2006年08月25日

UCD CSI Seminar

ucd_seminar0.jpg ucd_seminar1.jpg ucd_seminar2.jpg

I have just finised my talk at the UCD CSI Seminar. My time at UCD is running out...

Read Seminar Announcement

posted by oi at 23:56 | Comment(0) | TrackBack(0) | しごと(おおまかに言えば)

 

広告


この広告は60日以上更新がないブログに表示がされております。

以下のいずれかの方法で非表示にすることが可能です。

・記事の投稿、編集をおこなう
・マイブログの【設定】 > 【広告設定】 より、「60日間更新が無い場合」 の 「広告を表示しない」にチェックを入れて保存する。


×

この広告は180日以上新しい記事の投稿がないブログに表示されております。