PRPNET & Phrot discussion

[masser]

A PM I received from Lennart, recently...

Topic: PRPNet & phrot

Hello masser.

I have not done any timetest on PRPNet & phrot.

But what i have seen on bases higher then 2 it use to be faster with phrot.

The big thing with PRPNet is that you use cllr.exe in win and sllr in Linux. It is only to replace the files if it comes a new version. Thats why i like this better then LLRNet.

I can setup a testserver with some candidates if there is some that like to compare LLRNet and PRPNet.

PRPNet can also be setup todo doublecheck.
It will send out 2 candidates and compare residue.


Happy prime hunting

/Lennart

[masser]

...to which I replied:

Hi Lennart,

That sounds like a terrific idea. I have about 1600 Riesel tests we could use as a preliminary test. I am a little busy right now with real world issues, but hopefully in a day or two I can send you the set of tests I am thinking of.

Currently, my main reluctance for migrating to PRPnet is that our LLRNET server resides with our stats server. This allows our stats to be near real-time. If we use the PRPnet server, we will have to communicate large data sets over internet to our statskeeper. This is just a slight inconvenience.

The only other drawback to PRPnet that I can think of is that PRPnet is relatively new, and hence has not been significantly tested by the number-crunching community at large.

Do you mind if I post our pm exchange in the forum?

Thank you for taking the initiative on this issue.

[mdettweiler]

FYI, I've been running a PRPnet server over at Conjectures 'R Us for the past few weeks and, so far, it has run quite smoothly and reliably. Of course, it doesn't have any SQL features, so it can't integrate quite as tightly with SR5's existing stats database as LLRnet can, but the speed gains are so great that IMO, it's worth the extra admin hassle.

Also, you may want to keep in mind a system similar to what we do at NPLB for our LLRnet stats system: our LLRnet servers are not directly integrated into the database, and in fact are on completely separate servers; however, the servers copy off their results files daily and put them up for download, and scripts over on the database server then pull them down and impor them automatically. In fact, we've just recently improved things so that the DB server imports the results hourly.

[masser]

I uploaded a small file to this thread and reserved it for PRPNET. Lennart has or will shortly post info for his prpnet server.

Depending on how quickly this first file is tested, we will reserve the other available Riesel files on that thread for PRPNET.

Please be patient with us while we develop a stable PRPNET system.

[Lennart]

First i have done some archive that you can download if you not have PRPNet.

There is client's for Windows, Linux & Mac.

You can download it here http://samband.mine.nu/prpnet/sr5

It uses LLR for base=2 and Phrot for bases >2.

//**************************************************************** 
//email= is a REQUIRED field (either here, or on the command line)
// Comment the next line out, and fill in your email address.
// ***************************************************************
email=your@email.scom
 
//****************************************************************
// This value differtiates clients using the same e-mail ID
//(not critical)
//****************************************************************
clientid=comp_1
 
//****************************************************************
// server= are the servers that you want to get work from.  The first
// number is a percentage of work to get from that server.  The second
// parameter is a unique identifier appended to file names created by
// the client for that server.  The third is the server IP.  The fourth
// is the port for that server.  It is normally 7101.  Using a percentage
// of 0 will get work from that server only if the client could not
// connect to any other servers.  Please read the prpnet_servers.txt
// in this directory for information on some known PRPNet servers.
//****************************************************************
server=100:a:samband.mine.nu:7001
 
//****************************************************************
// maxworkunits= is the maximum number of workunits that a client can
// request at a time.  If a client requests too many the chances that
// some workunits timeout is greater.  The default is 5.  The server
// also has a limit, so the actual limit will be the less of this
// value and the value on the server. 
//****************************************************************
maxworkunits=1
 
//****************************************************************
// This is the name of LLR executable.  On Windows, this needs to be
// the LLR console application, not the GUI application.  The GUI 
// application does not terminate when the PRP test is done.
// On some systems you will need to put a "./" in front of the executable
// name so that it looks in the current directory for it rather than 
// in the system path.
// LLR can be downloaded from http://jpenne.free.fr/index2.html
//****************************************************************
llrexe=cllr.exe
//llrexe=./sllr
 
//****************************************************************
// This is the name of Phrot executable.  
// On some systems you will need to put a "./" in front of the executable
// name so that it looks in the current directory for it rather than 
// in the system path.
// Phrot can be downloaded from http://home.roadrunner.com/~mrodenkirch/
//****************************************************************
phrotexe=phrot.exe
//phrotexe=./phrot
// Default idle status is 1 (idle)
idle=1
 
// Timeout on communications errors
// (default is 60 minutes, minimum is 1 minute if not specified here...)
// Note that the actual used in the client is anywhere from 90% to 110% of this value
errortimeout=5
// Size limit for the prpclient.log file...
// 0 means no limit.
// -1 means no log.
loglimit=5000000
// Set the debug level for the client
//    0 - no debug messages
//    1 - all debug messages
//    2 - output debug messages from socket communication 
debuglevel=0

This is a test to compare how PRPNet works on sr5 work.

server=100:a:samband.mine.nu:7001
Port is 7001

I will post a new post here when it is up and runing.

It should be up in about 1 hr.

/Lennart

[Lennart]

Now it is up and running.

Deadline 3 days.

/Lennart

[ltd]

As I am the guy who will have to import the results into the stats I have a problem. LLRNET bases the results on a "nickname" and therefore the DB also uses the nicknames to identify the contributor when results are imported.

As I can see in the configuration description of prpnet it looks like the client is identified by the mail account only. As this information is not known to me at the moment I have no chance to give credit to the correct user. My user table has already a column to store the mail information so with the additional information it will be easy to modify the importer.

Another request to Lennart. As soon as you have some results can you send me an example result file so I can start to implement a parser for this file type.

I will earliest be able to start this beginning of februar as I am out of town next week.

[Lennart]

Here is a file. You cant change User but we can use nickname as clientID for now.

/Lennart

[ltd]

It will be easy to create a parser for this file format.
I also think we should use the clientID to store the nick.
Maybe it is possible to make an extension to prpnet to have an extra field for
a nickname as it is not a good idea to show the mail address on a stats page.
Every bot searching for mail addresses would find it and start to spam.

[mdettweiler]

What I do over at CRUS is use the computer ID value as the username in my script that converts PRPnet results to more traditional LLRnet-format results--that way, no need to worry about any clunky user/email databases or anything like that.

For example, I would set my email address as normally, and set my computer ID on all my clients to "mdettweiler". From what I've seen the server doesn't need these to be unique so that shouldn't be a problem.

I know it's not quite a perfect solution, and at CRUS it isn't even quite necessary to get the usernames quite right because we're not doing any stats over there at the time, but I figured we'd need some sort of system like this for when we start using PRPnet at NPLB, since there we have an existing LLRnet stats system that we'll need to integrate with.

[masser]

Any news on the prpnet/phrot server? I have been busy with day-job issues, but was curious if the parsing issues have been worked out. Have we completed the first test set?

best regards,