





 
Blogs I Read Alan Bell Alan Lepofsky Andre Guirard Ben Langhinrichs Bob Congdon Bob Obringer Bruce Elgort Christopher Byrne Damien Katz Dec's Dom Blog DominoHelper Ed Brill Esther Strom Gregg Eldred InsideLotus Jeff Eisen John D. Head Jon Udell Julian Robichaux Larry Niskala Libby Ingrassia Marko Vesapuisto Mary Beth Raven Nathan Freeman Ned Batchelder Olaf Björklund Ray Davies Richard Schwartz Rob Novak Rocky Oliver Stan Rogers Steve Castledine Tom Duff Toni Heinonen Ulrich Krause Vladislavs Tatarincevs
About me Mika Heinonen
Counter 53413
Referers
| | | Notes and Domino 8.0.1 have too many fatal bugs Thursday, 27 March 2008 01:14:27 EET | | | | | Some things I have experienced:
Domino 8.0.1 fatal bugs:
- Domino 8.0.1 on an AIX5L 5.3 server keeps crashing several times per day
- Domino 8.0.1 on one Windows32 server has crashed 4 times total
- Domino 8.0.1 on another Windows32 server crashed today for the first time
- Domino 8.0.1 x32 on Windows x64 makes Notes C API Terminate() call to hang for several minutes.
Notes 8.0.1 fatal bugs:
- Notes 8.0.1 has caused the harddisk's Master File Table to get corrupted multiple times on 2 different PCs, obviously because I am replicating my mail file to Local every 1 minute, and it has also FTI (on the Local replica only).
The symptoms that Notes 8.0.1 has damaged your harddisk's MFT are like this:
- Many error dialog windows in Windows saying: "The file or directory C:\$Mft is corrupt and unreadable. Please run the Chkdsk utility."
- The local replica will be 0 bytes in size after you have restarted Windows XP32, and it has fixed the MFT and corrupted file links
- No other files on the harddisk are corrupted or set to 0 byte size
The next time a Domino 8.0.1 crashes, I will downgrade it to Domino 8.0.
The next time a Notes 8.0.1 damages my harddisk's Master File Table, I will downgrade it to Notes 8.0.
The good news is that Notes and Domino 8.0 also support document and design compression, although you can't set or see them in the database properties. But they can still read and write the databases using compression.
| | | | | | |
| | | Bruce Elgort keeps deleting my comments! Wednesday, 26 March 2008 02:09:30 EET | | | | | I think he must be mad at me because I didn't deliver the next generation Domino Search engine to him yet (for IdeaJam). Well, it's not 100% ready for production, and needs lots of manual setup to make it work for custom applications. I'm planning to make an easy frontend interface so that it can be used like the normal Domino Search engine.
The interface will including things like include/exclude lists of field names, update frequency, etc...
But deleting my comments... I think it's the biggest sin of mankind to delete any Notes documents - I've personally not deleted any mails since I got Notes in 1995.
I've seen also similar things happen in discussion boards, but there the reason was that the phpBB/SMF update was not able to convert old formats to the new format.
So, if you post to Bruce Elgort's blog, please make sure you have a backup of your post, else your valuable time and idea might be lost forever! Same goes for all non-Domino based discussion boards on the web of course.
| | | | | | |
| | | The cheapest NAS of the world Tuesday, 25 March 2008 22:35:00 EET | | | | | If you want the cheapest network area storage with the biggest harddisk space on the planet, then there is only one choice for you. Siipi Roc features up to 11 harddisk slots: 6 internal 3.5" slots, 4 external 5.25" slots, 1 external 3.5" slot, equipped with 750GB Western Digital drives in the Antec P182 black case, running at Raid 0,1,5,6,10 with the Asus P5K-E motherboard, an Intel Quad Core CPU, 4GB Kingston RAM, and even a GeForce 7 series 3D card on openSUSE 10.3 64-bit.
For an amazing low price as 1800€ you can get a fully Raid 5 backed up NAS storage with 6TB and more free harddisk space.
| | | | | | |
| | | Lotus Notes All Client 8.0.1 on Linux Friday, 07 March 2008 22:30:33 EET | | | | | It just works:

I tried Notes 8.0 earlier on Ubuntu, and it was a bit crashy. But with 8.0.1 on openSUSE it seems more stable, when you learn to avoid certain things :)
I'll try to work with this setup a few weeks and see if I need Windows XP anymore.
On my work Laptop, the ThinkPad T61p, I have also PCLinuxOS 2008 minime, it will be a good challenger against my home openSUSE. Also kubuntu 7.10 x32 is still in the run, when one of those fails.
My server has been already running for a long time Domino 8.0.1 on openSUSE, and it has been running very smooth, without any reboots or crashes, and it hasn't even used the swap disk yet.
| | | | | | |
| | | AIX - change size of file system Thursday, 06 March 2008 01:18:26 EET | | | | | AIX has a command to change the size of a partition on the fly.
Lets make the root directory 4GB bigger:
bash-3.00# chfs -a size=+4G /
Filesystem size changed to 60555264
Lets look how the filesystem looks:
bash-3.00# df -g
Filesystem GB blocks Free %Used Iused %Iused Mounted on
/dev/hd4 28.88 24.75 15% 5471 1% /
/dev/hd2 4.62 3.23 31% 33600 5% /usr
/dev/hd9var 0.12 0.11 11% 433 2% /var
/dev/hd3 0.12 0.11 11% 66 1% /tmp
/dev/fwdump 0.12 0.12 1% 4 1% /var/adm/ras/platform
/dev/hd1 0.12 0.12 6% 378 2% /home
/proc - - - - - /proc
/dev/hd10opt 32.62 31.39 4% 30128 1% /opt
Lets try to make it another 4GB bigger:
bash-3.00# chfs -a size=+4G /
0516-404 allocp: This system cannot fulfill the allocation request.
There are not enough free partitions or not enough physical volumes
to keep strictness and satisfy allocation requests. The command
should be retried with different allocation characteristics.
Oh, there was not enough free space on the disk partition.
No problem, lets get more free disk space from another partition (from the partition which is mounted as /opt):
bash-3.00# chfs -a size=-4G /opt
Filesystem size changed to 60030976
Now lets try to make the root directory 4GB bigger again:
bash-3.00# chfs -a size=+4G /
Filesystem size changed to 68943872
Lets look how the file system looks now:
bash-3.00# df -g
Filesystem GB blocks Free %Used Iused %Iused Mounted on
/dev/hd4 32.88 28.75 13% 5471 1% /
/dev/hd2 4.62 3.23 31% 33600 5% /usr
/dev/hd9var 0.12 0.11 11% 433 2% /var
/dev/hd3 0.12 0.11 11% 66 1% /tmp
/dev/fwdump 0.12 0.12 1% 4 1% /var/adm/ras/platform
/dev/hd1 0.12 0.12 6% 378 2% /home
/proc - - - - - /proc
/dev/hd10opt 28.62 27.39 5% 30128 1% /opt
bash-3.00#
You can't do that in Windows, Linux or Mac!
| | | | | | |
| | | No collaboration without integration Tuesday, 04 March 2008 19:25:33 EET | | | | | First we had e-mails. Our paper mails, artistic handwriting and other personalized communication was gone.
Then we had websites. Our e-mails were reliefed and we could share information with everyone.
Then we had blogs. Our websites were updated less frequently and all structured information was gone.
Then we had rss feeds. Our custom look and style of blogs was gone and only raw information was left.
Then we had technorati. Actually I forgot about it completely.
Then we had ideajam. This was the first time we had a common place to share our ideas.
Then we had planetlotus. Our rss feeds were gone and we got back our blogs, but this time all of them together.
Then we had bleedyellow. Our blogs were still there, but we had also duplicate blogs, but we got also new ways of collaboration.
Then we had twitter. Our collaboration tools screamed for integration into one single global place, as we have now to update information in 4 or more different places seperately.
Planetlotus strikes again, and tries to keep all the places together. But it can do only so much, and it can't fix the source of the problem.
I'd like to see one single place for global collaboration.
I'd like to see it to running on Domino 8.0.1 on SUSE or AIX.
I'd like it to have everything what all the above systems provide plus be easily expandable for future ideas.
I'd like that the collaboration starts on the system level, and not only on the end user level.
And actually I think it should be in IBM's own interest also to host a global server, as they have the iron. I'm sure they can hack some Blue Gene/P to run like an AIX, and I'm sure many companies and individuals would love to join for monthly fees to host their own virtual server, when they know their site is running on the best hardware in the world.
And hey, collaboration could help also on the virtual server level, since not everyone needs their own virtual server, but they could be joined into communities, running on a common Domino server. Also those who prefer an own virtual server, could collaborate and share their idle time, then others, and themselves would have huge benefits from eachothers' idle times using IBM System p hypervisor technology.
| | | | | | |
| | | Why I love the Police Monday, 03 March 2008 19:31:19 EET | | | | | Look at their URL:

I wonder if they are running on AIX also...
| | | | | | |
| | | ODS48 Speed Test Monday, 03 March 2008 00:58:20 EET | | | | | At least in the core functions the new ODS48 (Notes/Domino 8 database format) is not remarkably faster than ODS43 (Notes/Domino 6/7 database format).
I made a few tests with new blank databases (one for ODS43 and one for ODS48):
First I created an agent which creates 100000 documents (no fields).
The first test was to run the agent from the Notes 8 Basic Client:
notes:
ods43: 90.914s
ods48: 90.547s
The remaining tests were done using Mozilla Firefox using the Preview function in the Notes 8 Basic Client (ods48nc means ODS48 format with no design and document compression enabled):
http:
create 100k docs:
ods43: Timer: 00108.78906s.
ods48: Timer: 00089.81250s.
ods48nc: Timer: 00086.27002s.
change 100k docs (using documentcollection):
ods43: Timer: 00024.68001s, 00024.50000s, 00023.56000s.
ods48: Timer: 00025.16000s, 00025.02002s, 00024.13000s.
ods48nc: Timer: 00024.96997s, 00024.80005s, 00024.02002s.
change 100k docs (using view and getnextdocument):
ods43: Timer: 00037.73999s, 00037.33008s, 00037.97998s.
ods48: Timer: 00038.25000s, 00037.83997s, 00038.50000s.
change 100k docs (using view and getnthdocument):
ods43: Timer: 00038.94995s.
ods48: Timer: 00039.75000s.
read 100k docs:
ods43: Timer: 00005.71997s, 00005.69995s, 00005.72003s.
ods48: Timer: 00005.71002s, 00005.71997s, 00005.70001s.
ods48nc: Timer: 00005.72998s, 00005.72998s, 00005.75000s.
The speed advantage of ODS48 was better visible on a slower system. There I got times like:
create 100k docs:
ods43: 207s.
ods48: 194s.
read 100k docs:
ods43: 14s.
ods48: 11s.
As a further observation, I noticed that ODS48 with design and document compression enabled can be much slower when opening documents with the Notes Client.
This might be especially noticable on servers with lower CPU power and documents with big RichText fields.
The fastest database option in general might thus be: Domino 8.0.1 with ODS48 and no compression.
| | | | | | |
| | | Welcome a new SPCLP Lotus blogger! Thursday, 28 February 2008 20:30:05 EET | | | | | Today I met him again, and he told me that he has just started his own blog a few days ago.
Please welcome Vladislavs Tatarincevs (or short: Vlad) from Latvia!
He is often referred by my colleagues as my "Master" (because I often learn new things from him), but I tend to say that it's a fruitful cooperation between me and him (since I also can teach him some things). Call it 50/60 if you want, but not 0/100 :)
My last LotusSphere was in 2005, so he is a tad ahead of me, since he has probably now PCLP 8.0 Designer (=CAAD) (and probably also Admin), but as I also never go home without the newest PCLP certification, I'd like to call the almost 100% PCLP scorers SPCLP (Superior Principal Certified Lotus Professional).
His blog can be found at: Latvia Lotus User Group, and I told him already to ask Yancy Lent to put him up in the planetlotus.org member list.
| | | | | | |
| | | LotusScript: over 780% faster checking for empty strings Wednesday, 27 February 2008 19:44:34 EET | | | | | I read on planetlotus.org about Bob Balaban's test of the speed difference between s="" vs len(s)=0 checking ( http://www-10.lotus.com/ldd/bpmpblog.nsf/dx/performance-testing-technique ).
My test agent looks like this:
Sub Initialize
Dim n As Long, ni As Long
n=10 ' number of test repeats
Dim t1 As Double, t2 As Double, tn As Double, tt As Double
Dim i As Integer, j As Integer, k As Long
Const x=10000
Const s="xxxxxxxxxx"
For ni=1 To n
t1=Timer
For i=1 To x
For j=1 To x
k=Len(s)=0
Next
Next
t2=Timer
tn=t2-t1
tt=tt+tn
Print ni & ": " & Format(tn,"00000.00000") & "s."
Next
Print "Average: " & tt / n & "s."
End Sub |
The line with k=Len(s)=0 varied and gave the following average timings over 10 tests:| Const s="" | k=Len(s)=0 | 7.532s |
| Const s="" | k=s<>"" | 58.938s |
| Const s="xxxxxxxxxx" | k=Len(s)=0 | 7.688s |
| Const s="xxxxxxxxxx" | k=s<>"" | 63.089s |
| Const s="xxxxxxxxxx" | k=s="" | 65.770s |
That means, even with small loops like 1 billion iterations, you get almost a minute faster results! Anyway, the number of iterations is not so important and you can't really judge on that, as someone might need even bigger iterations, but the fact that it's over 780% faster to use len(s) is important.
I don't think it's concidence that several people started to blog about LotusScript performance practices, just after I started to think about revising the LotusSphere 2007 LotusScript speed presentation, even before I posted anything on my blog. I think there is somekind of shared mind, which we can't explain scientifically yet.
| | | | | | |
| | | Deployment Status: 8x Domino 8.0.1, 5x Notes 8.0.1, more to come... Saturday, 23 February 2008 19:20:59 EET | | | | | I have deployed Domino 8.0.1 for x32, x64, AIX, Linux so far (Experimental, Research, Production (Linux) servers), and Notes 8.0.1 for x32 (Production Clients).
No problems have occured except the problems which were caused by "officially not supported by IBM" configurations:
- Domino 64-bit + DB2
- Domino 8.0.1 + Quickr 8.0.0.3
- DWA/Lite + Nokia (Symbian)
The next deployments will continue on Development, then Test, followed by Stage, and finally Production servers after a meeting with our consultant next week.
| | | | | | |
| | | Things what IBM Lotus should really focus on Thursday, 21 February 2008 21:29:33 EET | | | | | After evaluating the Notes/Domino 8.0.1 release, I found a few things which should be really fixed and done before doing anything else:
- Domino 8.0.1 64-bit support for NSFDB2. Currently it destroys DB/2 databases and permanently destroys the DB/2 interface too, so it has to be reinstalled for Domino 8.0.1 32-bit.
- Symbian support for Notes Traveller. Nobody really uses Windows Mobile devices (vanishing 12%), the vast majority on earth uses Symbian devices (65% market share).
- Symbian support for Domino Web Access Lite (or even Domino Web Access Full). Currenty neither works with any business phone: Nokia Communicator 9210, 9300, 9500, E90.
- Native Basic All Client for Linux: Lotus Notes, Domino Designer, Domino Administrator. Currently only Notes Standard Client works on Linux, and even there the implementation of Eclipse is horribly slow.
IBM has a clear business case here: Basic All Client for Linux reduces the Total Cost of Ownership even 3 times:
1) Basic Clients vs Standard Clients hardware needs
2) Linux and it's applications hardware needs
3) License fees for Windows (and also the cost for solving problems caused by Windows).
| | | | | | |
| | | 8.0.1 incoming at 980KB/s Wednesday, 20 February 2008 20:56:10 EET | | | | | 
And another 5 minutes later:

Domino 8.0.1 64-bit with DB/2 9.1.3 64-bit running on Windows 2003 R2 Enterprise 64-bit on IBM System x3650 with Dual Quad Core (=8 cores).
Oh, that reminds me I need to upgrade to DB/2 9.5 Viper next.
| | | | | | |
| | | Notes/Domino 8.0.1 went to Step 5: Gold Build Tuesday, 19 February 2008 22:52:00 EET | | | | | Finally some progress again, after it was stuck for long time in Step 4, it went today to Step 5: Gold Build:
http://www-10.lotus.com/ldd/r5fixlist.nsf/(Progress)/801
Maybe 8.0.1 is released even this week, or latest next week.
| | | | | | |
| | | kubuntu 8.04 + KDE 4.0.1 + Firefox 3.0b3 Monday, 18 February 2008 22:21:32 EET | | | | | I could say it's the most beautiful and effective desktop so far, technically and visually:
See Picture
What's missing from the picture? You guessed it right: Lotus Notes 8.0.1 for Linux!
But that will come very soon...
| | | | | | |
| | | openSUSE 10.3 + Domino 8.0 Saturday, 16 February 2008 01:10:01 EET | | | | | Finally I made the switch from Windows to Linux. I decided to go for openSUSE 10.3, as it has the fastest I/O of all operating systems in the world. This site, http://www.siipi.com is now running on openSUSE 10.3.
The installation was just as easy as installing Ubuntu (which is actually easier than installing Windows), but openSUSE gives you more freedom to make choices. It was literally just like inserting the CD and turning the power on, I didn't need to install any 3rd party drivers or anything else, everything was done by the openSUSE CD.
Of course I needed to convert the harddrives from NTFS to ext3 (I used a new harddisk for the "C:" drive), but that's just normal partitioning, formatting and moving files from one harddisk to another. openSUSE can read and write NTFS drives, but for production use only ext3 drives should be used (much better performance).
I think I will use this combo also on our IBM System p570 servers, then we have a unified environment between Intel and PowerPC servers. Oh wait, I can't! There's no Domino for Linux on PowerPC, ok AIX is actually better since it has JFS2.

| | | | | | |
| | | The best SAN of the world Wednesday, 13 February 2008 22:21:17 EET | | | | | If you want a HUGE disk storage, for example for daily backups of multiple servers, then this baby - the IBM DS4200 Express - is the only choice for you: for an amazing price of 4600€ and 512€ per 750GB additional disk, one 3U rack supports 16x 750GB = 12TB, and with optional expansion units (3U, 16x disks each) this single SAN supports even up to 112x 750GB = 84TB!
Similar systems with just 20x disks cost up to 20000€ from other manufacturers, and consume even more rack U space due to horizontal placement of the disks. And with up to 4x 4GB/s Fibre Channels the IBM SAN performs an effective transfer rate of 1550MB/s for applications (with Domino 8.0.1's database compression even more MB/s), which makes it even faster than some internal hard disks of servers.
As we are getting more non-production Domino servers (stage, testing, development, research), there is a need to backup their data also on a daily interval. Once again IBM has proven their leadership and competence in server hardware.
| | | | | | |
| | | Collaboration Riddle Friday, 01 February 2008 21:12:56 EET | | | | | The guard riddle in the movie Labyrinth (with David Bowie) reminds me of the everlasting battle between IBM and Microsoft, which was brought up once again during the last LotusSphere 2008.
Solve the riddle, and you will know the absolute truth between the claims of IBM and Microsoft.
SPOILER WARNING! Don't watch the video before you have solved the riddle yourself.
Here goes the riddle dialog:
There are 2 doors. Sarah needs to find out which door leads to the castle at the center of the labyrinth, as the other one leads to certain death.
Guard: "You can't ask us. You can only ask one of us."
Guard: "It's in the rules, and I should warn you that one of us always tells the truth, and one of us always lies. That's a rule too." Gesturing to the Guard, "He always lies!"
Guard: "I do not! I tell the truth!"
Guard: "Oh, what a lie!"
Sarah: "Alright," to the Guard, "answer yes or no: would he (Guard) tell me that this door leads to the castle?"
Guard: "Uhhh...yyyes?"
Sarah: "Then, the other door leads to the castle, and this door leads to certain death."
Guards: "oooOOoooh."
Guard: "How do you know? He could be telling the truth!"
Sarah: "But then he wouldn't be. So if you told me that he said 'yes', I know the answer is 'no'."
Guard: "But I could be telling the truth!"
Sarah: "But then he would be lying. So if you told me that he said 'yes', then I know the answer would still be 'no'"
Guard: "Wait a minute," to the Guard, " is that right?"
Guard: "I don't know – I've never understood it!"
Is Sarah right?
Spoiler Video: Labyrinth Riddle
Please note that there are also other hidden truths in the video...
| | | | | | |
| | | New Testament source code released Saturday, 26 January 2008 01:45:05 EET | | | | | I already knew for many years how the Old Testament was compiled (using the Babylonian Enuma Elish and it's original source from the Sumerians), but Jesus and the New Testament remained a mystery to me, until today.
Here is the full source code of Jesus and the New Testament, and it's all Open Source and Free:
http://www.zeitgeistmovie.com
It seems religion has been now completely disassembled and there is no mystery anymore. Way to go science!
| | | | | | |
| | | Domino 8.0 on AIX 5L 5.3 Friday, 11 January 2008 18:56:21 EET | | | | | Today I installed Domino 8.0 on our POWER5 AIX 5L 5.3 test server, and it doesn't want to accept our existing server id files. I saw on a IBM SPR that a similar issue has been reported, but I don't know yet - until it's fixed in 8.0.1 - if that fixes also this issue.
When I force it to eat the id file (by ftp:ing the id file to the server and using vi to edit the server's notes.ini file to use that id file), it runs fine for a few minutes, although still complaining that the public certificate does not match, but after a few minutes it brings up another error message related to the id file ("The subject's public key found in the certificate is not the one stored in the ID file for that entity."), and you can't connect to the server anymore.
Now I have 3 options:
1) Wait until Domino 8.0.1 comes out.
2) Try Domino 7.0.3 on AIX 5.3.
3) Try Domino 8.0 on SUSE 10.3.
Let's see:
1) I want to continue the migration testing, waiting must wait, so this is no option for me.
2) This will probably work, and it's what I will do on the AIX machine next.
3) This might work, but I don't want to trash the AIX installation just yet, since so far AIX has worked well. I will try this on a Windows 2003 server though.
VNC server works great on AIX, it's very smooth and fast to use, much faster than Windows Remote Desktop on a Windows 2003 server. And it supports even multiple simultaneous users (multiple mouse cursors), while the VNC for Windows supports only 1 mouse cursor for simultaneous users. Windows Remote Desktop doesn't even support simultaneous users at all, which is a waste of valuable work time.
The funny thing is that the AIX server (IBM System p) doesn't even have a graphics card, but well what is a graphics card anyway, just an additional CPU to focus on graphics operations, a POWER5 CPU can do that with almost idling:

| | | | | | |
| | | The Final Blow Monday, 07 January 2008 20:53:55 EET | | | | | I'm quite objective and tolerant when it comes to operating systems, applications and hardware. I just want the best for money, and temporarily can even accept functionality over the value for money factor. Indeed, I'm still looking for the best hardware/software combo.
One thing I've learned is that Linux has been predicted to take the throne of Windows too many times, it never did yet, but now something has happened which finally might give it the final blow. Even I made the mistake, and predicted it way too early: http://siipi.com/public/mika.nsf/blogs/D3F28A19C60356C5C2257361006FDA6A
Microsoft had many good products, like Visual Studio 2005 C++, but now that has been beaten by Linux GNU C++ too (see Speedtest article: http://siipi.com/public/mika.nsf/blogs/E2A4779723C2F147C22573C600673696). Was THAT the final blow?
For me it's a very important issue, as I want CPUs to do their job: compute a lot and very fast, at the lowest cost. If a CPU cannot do that, then I look for a better one.
Of course speed is not everything, but it is more important than anything else. You might compare it to a truck which drives at 50km/h and carries 20 tons of load with a moped which drives also at 50km/h but can carry only 100kg of load. However, in most cases it's not enough to drive at 50km/h with whatever load, but the need is for speed, like 320km/h, and at those speeds everyone carries about the same load, except for airplanes... But then the price goes up to the sky accordingly.
| | | | | | |
| | | if(thisblog.isshown()){planetlotus=rocks}; Monday, 07 January 2008 20:43:09 EET | | | | | lotus notes designer domino linux ibm aix windows
Just testing if it really works :)
If it works, then I agree with Andy Donaldson: http://blog.macian.net/2008/01/youve-got-to-love-planetlotusorg.html
I had kinda the same idea also for the Lucid project, which is still under development (the 3D desktop replacement for all OS),
but well, good ideas are always invented by someone at some point of time, and not only once.
It seems planetlotus.org is not able to auto-pick lotus related blogs, so I sent a mail to the admin.
| | | | | | |
| | | Surprising CPU Benchmark Results Friday, 04 January 2008 20:49:19 EET | | | | | I would never have guessed that my laptop would beat them all.
Does this mean that Linux x32 is the fastest OS in the world, and that 32-bit is faster than 64-bit, and that clockspeed means everything, and dual cores are better than quad and single cores?
I want to find out what is the fastest linear and parallel processing system on the market, and I really hope M5 will be it.
| System ID | M1 | M2 | M3 | M4 | M5 | M6 | M7 | M8 |
| System Name | IBM System p5 505 | IBM System x 3650 | Lenovo ThinkPad T61p | Siipi Blackhawk | IBM System p 570 | IBM System x 346 | Siipi Falcon | Siipi Osprey |
| System Specifications | POWER5
1.9GHz | 2x Intel Xeon Quad Core
2.66GHz | Intel Dual Core
2.66GHz | Intel Quad Core
2.4GHz | 4x POWER6
4.7GHz | 2x Intel Xeon 2.4GHz | Intel Pentium 4 3.2GHz | Intel Dual Core 1.8GHz |
| Operating System | AIX 5L 5.3 | Windows 2003 Enterprise R2 x64 | Windows XP SP2 x32 | Windows XP SP2 x32 / Linux Ubuntu 7.10 x32 | AIX 6.1 | Windows 2003 Enterprise x32 | Windows 2003 Enterprise x32 | Windows XP SP2 x32 |
| Speedtest 1.0 32-bit | - | 25.125s | 26.640s (14.312s+12.328s) | 27.781s (Windows)
48.010s (Linux, default)
21.590s (Linux, optimized 16.660s+4.930s) | - | 58.406s | 45.312s | 38.296s |
| Speedtest 1.0 64-bit | 199.391s (default)
51.450s (optimized) | 122.937s | - | - | coming in 2008-03 | - | - | - |
Maybe I should explain a bit further why these systems were used in the test:
M1: the cheapest System p you can get, we wanted to have a test server for evaluating the production server (possibly JS22 or 570)
M2: our currently fastest System x server
M3: my office laptop
M4: my home gaming PC
M5: an vision how our first production System p could look
M6: our old production server
M7: a low cost server (high GHz single HyperThreading core (=semi-dual core))
M8: a low cost gaming PC (low GHz dual core)
Speedtest 1.0 source code (all-bit, all platforms) is here:
#include "stdio.h"
#include "time.h"
class testclass
{
public:
int x;
testclass(void) {
x=1;
}
~testclass(void) {
x=0;
}
};
int main(int argc, char **argv)
{
double n=0;
long i=0;
long t1=0;
long t2=0;
long t1t=0;
long t2t=0;
t1t=clock();
printf("Speedtest 1.0 (c) 2008 Siipi\n");
printf("Counting 10 billion floating points...\n");
t1=clock();
while(n<100000.0)
{
n+=0.00001;
i++;
}
t2=clock();
printf("Done. i=%ld, n=%f, time=%fs.\n",i,n,
(double)(t2-t1)/CLOCKS_PER_SEC);
printf("Creating and deleting 1 billion class objects...\n");
t1=clock();
i=0;
while(i<100000000)
{
testclass *a=new testclass();
delete(a);
i++;
}
t2=clock();
printf("Done. i=%ld, time=%fs.\n",i,
(double)(t2-t1)/CLOCKS_PER_SEC);
t2t=clock();
printf("Total time=%fs.\n",(double)(t2t-t1t)/CLOCKS_PER_SEC);
return(0);
} |
To compile and run Speedtest 1.0 under AIX/Linux/MacOSX/Sun/Cray/IRIX/AmigaOS, enter the following commands:
g++ st.cpp
time ./a.out
To compile and run Speedtest 1.0 under Windows, enter the following commands:
(Launch Visual Studio 2005 C++, select: File/New/Project From Existing Code/Console Application)
(Select Release instead of Debug)
(Choose Build/Build Solution)
\programs\rktools\ntimer release\st.exe (ntimer.exe comes with the Windows 2003 Resource Kit Tools)
You don't necessarily need to run Speedtest 1.0 with time/ntimer, I just used it to double-check that my time measurement was working fine.
The program output will look like this:
Speedtest 1.0 (c) 2008 Siipi
Counting 10 billion floating points...
Done. i=1410063201, n=100000.000003, time=15.078000s.
Creating and deleting 1 billion class objects...
Done. i=100000000, time=12.703000s.
Total time=27.781000s.
Unfortunately I don't have all systems which exist, but here are some precompiled binaries:
The source code can be also downloaded, to make it a bit easier on systems which have no web browser installed: ftp://ftp.siipi.com/st.cpp.
I would be glad to know if someone's server/workstation can run this under 21s.
I managed to get the POWER5 code twice faster with the following compiler options:
g++ -O3 -fomit-frame-pointer -fstrict-aliasing -mcpu=power5 st.cpp
The Linux code went also more than twice faster with these options:
g++ -O3 -mtune=pentium4 st.cpp
Furthermore, the AIX programs might also run faster when compiled with IBM VisualAge AIX 6.0 C++ (especially with the XL C/C++ addon) or Intel C++, although Linux GNU C++ beats Microsoft Visual Studio 2005 C++ in optimized machine code speed, which also surprised me, let's say vastly.
| | | | | | |
| | | My First SuperComputer Thursday, 03 January 2008 21:20:39 EET | | | | | For the first time in my life I got to see how a SuperComputer is. Well, it's just the smallest IBM System p server on the market, but it still beats the IBM DeepBlue '97 (the one which played chess against Kasparov, and was 2 racks in size) by miles. I'll test soon how it compares to our fastest IBM System x servers (they have 2x Intel Xeon Quad Core 2.66GHz), and I think it will beat them all too, maybe even all together in performance.
The server was even easier to install than my previous x3650, well it was preinstalled. It seems IBM takes better care of their own babies than their Intel servers, which is of course understandable. The CPU is from IBM (btw, XBOX 360 CPUs were also done by IBM, it's a 3-core PowerPC CPU), and of course the rest of the hardware too. Harddrives and RAM might be from other suppliers, I haven't checked yet.
The first thing which baffled me with this machine was that it was missing a video card! A quick call to the reseller solved this little "problem", as they explained that I have to use a serial cable to connect to the server with a terminal program. Alright, I checked on Google for more details about this setup, and found out that it is running on 19200,N,8,1,Hardware protocol parameters (IBM should really put this info in the AIX 5L manual, as it's kinda essential to get started). Fortunately the manufacturing machines service guys had one such cable to borrow, so I was able to make a first contact with this SuperComputer.
Interestingly, the server is alive at the serial interface even when it's powered off (just the power cable connected to the power supply). And turning the power on is not just like pushing the power button, well it works too if you wait long enough, but internally it's still running a boot up sequence before the actual server with harddisks and stuff starts.
After the first contact via serial terminal, I managed to setup the IP address and connect to it via telnet. Once I downloaded the VNC Server, I was even able to connect with the graphical VNC Viewer from my Windows XP workstation.
So far so good, but another problem arised when I tried to install the GNU C++ compiler: the /opt mount was too small. However a quick Google brought the solution again:
chfs -a size=+4 /opt
and at once the size of the /opt mount was large enough for the g++ installation.
Now I'm still missing some bos.adt.libm (that's IBM AIX slang, so mind me :) stuff to compile my C++ programs, but that came on a CD too, so I will continue my journey tomorrow.
| | | | | | |
|
| |
Hot Topics
|