I'm adding a new thread dedicated especially for new application versions - test versions.

I will send new information on this topic.
Here it will be possible to exchange thoughts/testing problems.

At the beginning I add a test version of PERD application for Windows:
https://root.ithena.net/usr/apps_test/ithena_perf_v1.03_bin_windows_intelx86.zip

You can start this application by running the file: ithena_perf_v1.03_windows_intelx86.exe

The application should work properly on 32-bit and 64-bit systems.

I would like to ask you to test the application (on different systems).
The application should not "freeze" anymore.

You can send the results to my email address: lswierczewski at cybercomplex.net

I was supposed to add the application to BOINC faster, but there were some problems/errors.

Thanks a lot! ;)

Win7 Pro SP1 x64

tuxliketimeout_v1.0.exe - System Error
The program can't start because api-ms-win-crt-runtime-ll-l-0.dll is missing on the computer.
Try reinstalling the program to fix this problem

required Visual C++ 2015 Redistributable (x86 и x64)
found and installed in addition to previous versions, maybe then a 2017 version will be required - for the full collection ;o)

At the beginning I add a test version of PERD application for Windows:
https://root.ithena.net/usr/apps_test/ithena_perf_v1.03_bin_windows_intelx86.zip

You can start this application by running the file: ithena_perf_v1.03_windows_intelx86.exe

The application should work properly on 32-bit and 64-bit systems.

I would like to ask you to test the application (on different systems).
The application should not "freeze" anymore.

You can send the results to my email address: lswierczewski at cybercomplex.net

I was supposed to add the application to BOINC faster, but there were some problems/errors.

Thanks a lot! ;)

Win XP 32 bit
I am probably not doing this correctly, but when trying to run the programme I get the message "The procedure entry point inet_pton could not be located in the dynamic link library WS2_32.dll"

I am not sure when to extract this to, so I tried a separate folder and also from the Boinc folder.

Conan

---

I've updated the DLL libraries.

https://root.ithena.net/usr/apps_test/ithena_perf_v1.03_bin_windows_intelx86_R1.zip

@Conan: Can you check with the new version of R1 now?
@Sergey Kovalchuk: You can test R1 on Win7 Pro SP1 x64 without Visual C++ Redistributable?

@Sergey Kovalchuk: You can test R1 on Win7 Pro SP1 x64 without Visual C++ Redistributable?

I have other projects there requiring library versions up to 2010
Version 2015 was not required before - I installed it only now

I tried again with the new R1 download file.

The error message changed to "api-ms-win-core-crt-l1-1-0.dll was not found"

So I went out and downloaded a version of it.

Re-ran and then got the error "api-ms-win-core-crt-l2-1-0.dll was not found"

Located that and downloaded that as well

Tried again and then got the error message that

"The procedure entry point EtwGetTraceFlags could not be located in the dynamic library ntdll.dll"

I have ntdll.dll installed in /windows/system32/ folder so it should be OK, unless a newer version is required.

It is now 1.35 AM here in Australia so I am off to bed and I will try again later today.

Conan

---

@Conan: Maybe adding ntdll.dll directly to the application directory will help?
https://www.dll-files.com/ntdll.dll.html

as I said - now the program needs Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019 ;o)

Probably the easiest and fastest solution will be to add Microsoft Visual C++ Redistributable for Visual Studio to client system requirements...

Is that acceptable?

Is that acceptable?

this is a normal requirement, it sometimes requires a really strange ;o)
here the main thing is to determine exactly the version you need - it depends on the compilation mode in Visual Studio
now requires specified in the post above 2015-2019

yesterday I was looking for a solution - I saw a hint that you can choose the compilation mode as in version 2010.
then the requirement for the version 2010 redist is very likely to be “closed” by another BOINC project or some new office suite in general.
in addition, complete sets of such packages are included in Windows updates, for example, the full set in Windows Update KB2999226

PS when you proceed to compile the application on Linux, a similar question will arise regarding the requirements of the version of libraries - either too old (forbidden in new distros), or vice versa too new for most users. here is an example

@Conan: Maybe adding ntdll.dll directly to the application directory will help?
https://www.dll-files.com/ntdll.dll.html

G'Day Rysiu

No that did not change anything,

I tried two different and newer versions of ntdll.dll but I still get the same error message that there is a problem with the entry point EtwGetTraceFlags and the library ntdll.dll.

Nothing really stood out as helpful doing a quick look on Google.

I did try re-registering ntdll.dll and then found out that you can't as it is not a function that is supported by ntdll.dll, so that is why it failed to register, worth a try.

I have not added the newer downloaded versions of ntdll.dll to my /windows/system32/ folder as I don't think my installed version of ntdll.dll is corrupt and I don't want the newer versions to bugger up my Windows install, so I have only placed the newer files in the iThena folder I created for the test.

Conan

---

I'm sending links to new trial versions of applications:

Windows (x86/x64): https://root.ithena.net/usr/apps_test/ithena_perf_v1.03_R3_bin_windows_intelx86.zip
Linux (x64): https://root.ithena.net/usr/apps_test/ithena_perf_v1.03_R3_bin_x86_64-pc-linux-gnu.zip

Unfortunately there is no version for Linux 32-bit yet.

The applications should be run from the wrapper level.

For Windows:

wrapper_ithena_perf_v1.03_windows_intelx86.exe

For Linux:

wrapper_ithena_perf_v1.03_x86_64-pc-linux-gnu

The application can run much longer.
In my tests the application has been running for about 360 minutes (Windows/Linux the same execution time).

The "freezing" problem has been removed.

Tasks will run longer.
However, credits will be equivalent.

The application on Windows platform requires Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019.

Please, test the app and send the results (result_* and stderr files) to email: lswierczewski at cybercomplex.net

You got any problems? ;)

Linux (x64): https://root.ithena.net/usr/apps_test/ithena_perf_v1.03_R3_bin_x86_64-pc-linux-gnu.zip

Now the choice of application / subproject is essentially unambiguous: CNode only for Linux, PERF only for Windows (NRT is actually abandoned)

With the launch of the new version, the situation will change - two applications for Linux. Will it be a user-controlled choice? Or the choice will be up to the server, then what is the algorithm?

doubling the number of applications — additional processes on the host, more host resources (will there be conflicts between tasks?), more load on the network — traffic in GB/s is not as worried as the number of connections (for NAT) and their intensity (Mp/s on a router / firewall)
on farms where there are many (hundreds) instances, doubling tasks can have disastrous consequences for the user location and, in fact, a DDoS attack on the project site.
providers can use mechanisms to suppress such activity

again, there will be a question about the equivalence of the assessment for both tasks, and the bonus for the double load on Linux hosts
solution to the problem with the credits for the task, in turn, will lead to the issue of badges - it's time to think about their design ;o)

the application has been running for about 360 minutes

6 hours continuously and without checkpoint?
it is not important for continuous 24/7 work
but for work in "business time" 8/5 - there is almost no chance for completion, especially with the option "network required"

[upd] about the test.
> 2 hours progress 0.00000
stderr - a pack of such messages

warning: Ignoring nonsense TCP MSS 4297216
WaitForSingleObject failed. (ERROR 0)
warning: Ignoring nonsense TCP MSS 4297216

result.dat several times "error - unable to receive control message: Connection reset by peer"

@Sergey Kovalchuk The test application has ended its execution?
How long did you finally wait?
Information in stderr quite standard.

In fact, I can introduce shorter tasks.
But I think 2 hours would be acceptable?

In relation to questions:

1) Do you know how the BOINC server currently handles the selection of applications as standard? Does it work correctly on the server/project side? In the past (quite a long time ago) there were problems with it... The choice of application preferences is probably an open problem here.

2) The network load I think will be regulated over time. I can implement the option for the user to define how "difficult" (very big network load? or small?) tasks will be downloaded by a specific client.

@Sergey Kovalchuk The test application has ended its execution?
How long did you finally wait?

I sent the result - check e-mail

17:19:35 (5636): Can't open init data file - running in standalone mode
17:19:35 (5636): wrapper (7.7.26016): starting
17:19:35 (5636): wrapper: running ithena_perf_v1.03_windows_intelx86.exe ()
warning: Ignoring nonsense TCP MSS 4297216
. . . . . .
warning: Ignoring nonsense TCP MSS 4297216
23:16:09 (5636): ithena_perf_v1.03_windows_intelx86.exe exited; CPU time 23529.693230
23:16:10 (5636): called boinc_finish(0)

local time UTC + 3

by the way, your mailer does not like ZIP files, some messages with previous results were not delivered

In fact, I can introduce shorter tasks.
But I think 2 hours would be acceptable?

the task can be of any duration if at least once an hour there is a checkpoint.
it’s better if the real result is saved: processed one server from the list - save the checkpoint

In relation to questions:

1) Do you know how the BOINC server currently handles the selection of applications as standard? Does it work correctly on the server/project side? In the past (quite a long time ago) there were problems with it... The choice of application preferences is probably an open problem here.

2) The network load I think will be regulated over time. I can implement the option for the user to define how "difficult" (very big network load? or small?) tasks will be downloaded by a specific client.

I read somewhere, sort of like a "local scheduler": 10 levels of difficulty, matching the CPU model (parsing the line), etc.
IMHO, too difficult for such a simple task.

another idea is an endless task, once an hour packing the result and sending it to the server with a trickle message (search keyword)
an example is climate prediction or QCN (sensor and continuous)
build a distributed network of measuring agents following the example of a network of seismic sensors :o)
PS then crediting is not for the task (it is endless, you can just kill it), but for each sending of the result (each trickle) - an idea from the DHEP project

In fact, I can introduce shorter tasks.
But I think 2 hours would be acceptable?

the task can be of any duration if at least once an hour there is a checkpoint.
it’s better if the real result is saved: processed one server from the list - save the checkpoint

In relation to questions:

1) Do you know how the BOINC server currently handles the selection of applications as standard? Does it work correctly on the server/project side? In the past (quite a long time ago) there were problems with it... The choice of application preferences is probably an open problem here.

2) The network load I think will be regulated over time. I can implement the option for the user to define how "difficult" (very big network load? or small?) tasks will be downloaded by a specific client.

I read somewhere, sort of like a "local scheduler": 10 levels of difficulty, matching the CPU model (parsing the line), etc.
IMHO, too difficult for such a simple task.

another idea is an endless task, once an hour packing the result and sending it to the server with a trickle message (search keyword)
an example is climate prediction or QCN (sensor and continuous)
build a distributed network of measuring agents following the example of a network of seismic sensors :o)
PS then crediting is not for the task (it is endless, you can just kill it), but for each sending of the result (each trickle) - an idea from the DHEP project

It's not BOINC related, but that is precisely what the RIPE Atlas project is doing. I have had one of their probes active for almost six years.


https://atlas.ripe.net/about/

---

I'll probably solve this at the project scheduler level.

The panel for each host will be able to define specifications.

Additionally, we will add other options here: https://root.ithena.net/usr/forum_thread.php?id=29

The form in the panel will have a structure (prototype):

-----------------------------------------------------------
Applications executed on HOST ID XXX:

[APP NAME]: [checkbox = yes/no]
[APP NAME]: [checkbox = yes/no]
[APP NAME]: [checkbox = yes/no]


-----------------------------------------------------------
HOST ID XXX location info:

Continent: [input - list]
Country: [input - list]
City: [input - text]
Postal Code: [input - text]
Latitude: [input - double]
Longitude: [input - double]

Accuracy of location: [input - double]
Mobile point: [yes/no]


Connection Type: [input - list below]

    Hardwired broadband access:
        Dial-up access
        Integrated Services Digital Network (ISDN)
        Digital subscriber line (DSL, ADSL, SDSL, and VDSL)
        DSL Rings
        Fiber Optics (Fiber-to-the-x (FTTx))
        Power-line Internet
        ATM and Frame Relay
    Wireless broadband access
        Satellite broadband
        Mobile broadband
            2G
            3G
            4G
            5G
        WiFi Class

ISP Name: [input - string]

Theoretical bandwidth (upload) [kbps]: [input - double]
Theoretical bandwidth (download) [kbps]: [input - double]