Plz help sb1.25 mismatch stats

[LtZombie]

When I use the sb1.25 client on my p4 2,53 ghz my stats will not go above 89,0 cems/sec. while the client itself states above 500 k cems/sec. I dont know why. When i install the client on a different pc with the same username it works fine.

What am I doing wrong? A p4 2,53 ghz with 1gb ddr mem should be able to score above 500 k.

Plz help.

LtZombie

[LtZombie]

I am currently testing if v1.20 gives the same troubles. (Mismatch between stats and client cems/sec)

LtZombie

[Nuri]

Do you mean stats at client vs. stats at the web site?


The stats at the web site are based on actual progress. If you stopped the client for some time, it will naturally drop.

[LtZombie]

Yep indeed I mean webstats vs client stats. When I run the v1.25 client it sais 500+k on the client but on the webstats it is not higher than 90. I dont mean the numbered stats on the web, coz I know that they drop and rise, but the graphical representation gives me the indication that there is something wrong. The grapihics dont go above 90 on the web while my client is at 500 +.

I am now tryin v1.20 and it seems that it is working correctly. So the webgraph is indicating the same as my client.

My user name is LtZombie if anyone wants to see what I mean. I run the program 24/7 so there should be a 24/7 indication, but that isnt the case. In my team there is also BobDeBouwer as user. It is also me on a different location, there v1.25 is working correctly also 24/7

[Joh14vers6]

Originally posted by LtZombie
Yep indeed I mean webstats vs client stats. When I run the v1.25 client it sais 500+k on the client but on the webstats it is not higher than 90. I dont mean the numbered stats on the web, coz I know that they drop and rise, but the graphical representation gives me the indication that there is something wrong. The grapihics dont go above 90 on the web while my client is at 500 +.

I am now tryin v1.20 and it seems that it is working correctly. So the webgraph is indicating the same as my client.

My user name is LtZombie if anyone wants to see what I mean. I run the program 24/7 so there should be a 24/7 indication, but that isnt the case. In my team there is also BobDeBouwer as user. It is also me on a different location, there v1.25 is working correctly also 24/7

What is possible is that the client is not transmitting intermediate blocks, due some circumstances. Normaly when a client reports back within 24 hours the gap is filled again, but it will take longer than 24 hours it will show a gap. Look at your logfile if the client submitted all the time blocks.

[LtZombie]

The log files show no problems. everything looks fine. At the moment I am still running v1.20 to compare. It seems that this client is working fine. No mismatches between web and client. Also this log looks fine. Tonight I try the v1.25 again to see what i can find.

Strange, I use 4 pcs in which three of them are running correctly. Only my main pc gives problems. All are running v1.25. Same install, coz i did it myself.

[Mystwalker]

From your stats, it seems like you have a test running for at least 1.5 months. If this test is on your main PC, then the low stats come naturally, as the performance is averaged over a whole test, not over block intervals.

To illustrate it graphically:
The performance of 500 kCeM/s is shown, but it is spread across the whole time interval of the test.

When you try it on other computers or use another client for testing, you most likely have a new test requested, haven't you? If yes, these fresh tests of course show "Client speed = stats performance", as long as they are run 24/7.

[LtZombie]

Look at the lifetime user rate graph of LtZombie, the gaps are my problem. Howcome the client is running and there are gaps? Thats my question. More clearly I cant say. That the rate varies fromm 400 till 600 i can understand of pc usage, but the gaps to 70 or 80 dont make sence to me....


The overall rate is about 88, while the programs runs 24/7 on a p4 2,53 ghz.

With the 1.20 client i am running now things seem correct.

[Mystwalker]

The overall rate shows 90k CeM/sec for the time since your signup 172 days ago. If you had the client running 24/7 all this time, then you would have maybe 500k CeM/sec in this field. But I guess the client wasn't running for some time.

According to the graph, there was an inactivity of about 1.5 months lately, plus some more in the past (though those do not relate to the 1.25 client, as it hasn't been out then...).

When you get a new test with the 1.25 client and expire the old one (which I wouldn't do if I was you, as this wastes a lot of work), I most likely will work as you think, too.

[LtZombie]

Now we are getting somewhere. coz the large gap from 05/01 till 06/01 reads as you can see some activity. Well it should be about 500 in stead of 80 or 90.
And that is what i dont understand.

Also in the 24 user rate their is a gap visible which should not be there. In this case the about 800 k is from my pc together with a second machine running the same username. The 500 rate is from my pc with the 1.20 client which I installed to test. The gap is the mystery part. It was solely my pc running without help from the second pc (see 800 rate) and with the 1.25 client running.

Maybe you think why ask now when there is a gap of a month. Maybe you think that in that month the pc was not running. Well the truth is, it wasnt bothering me that much, but now it was goin to irritate me. I use 6 machines on different locations and from different friends to join this project, so the credits should also be recognized.


Also the teamrates give the idea that something is not ok.... The most part 2 machines are running 24/7 (including mine) With the change from 1,25 to 1,20 client the rate goes up from below 1500 to just below 2000. So with the 1.20 client the rates are properly registered.

Why not use the 1.20 instead, well the 1.25 client makes better use of the proc and mem, so there is a performance gain. I just want the rate also to be correct.


LtZombie

[kugano]

YOUR CLIENT'S ACTUAL RATE

500 --| .-----.
| |#####|
250 --| .-----. |#####`-----.
| |#####| |###########|
0 --|_|#####|_____|###########|__
| | | | |
Mon Tue Wed Thu Fri
o o
| |
`-- test assigned |
|
test finished --'
Compare with...

WHAT YOUR STATS WILL CORRECTLY
LOOK LIKE ON THE WEBSITE

500 --|
|
250 --| .-----------------------.
| |#######################|
0 --|_|#######################|__
| | | | |
Mon Tue Wed Thu Fri
o o
| |
`-- test assigned |
|
test finished --'

[Nuri]

LtZombie,

If understand correctly, you're saying that the PC worked 7/24 for the period of April 15th to June 7th, then there might be something wrong.


Did you ever check the log file of your client?

If there's something unusual on the client side, I guess you might find traces of that in the log file.

What comes to my mind is, may be (and somehow) the client could not connect to the server for a very long time (could not resolve hostname).

[LtZombie]

That could be the case, but why the 80 rate than?

[Mystwalker]

Just to make sure: Which '80' do you mean?

The "Work rate - overall" field of the statistics table or the low graph 'altitude' in the diagram from 04/15 - 06/05?

[Nuri]

If you mean the statistics, it's normal.

It's simply total work done divided by the time since sign up.

[LtZombie]

Ok, I have tested it for a while....

Here are the results:

When I do a first install of my sb1.25 client and I start it up, it runs fine. That means "The Last Day Work Rate" and "24-hour user rate" is correctly stated in the stats. Its about 600K cEMs/sec. When I stop the programs and start it up again then "The Last Day Work Rate" and "24-hour user rate" stats show no more then 5K cEMs/sec. While my client gives the normal rate of 600K

I did the test for a full day each time I tried a configuration. If u want a clear understanding of what I mean, plz refer to Teambob. Ltzombie at the moment is running fine, but bob the bouwer and ukkie001 give problems. They give incorrect readings in the "The Last Day Work Rate" and "24-hour user rate" stats. Look at ukkie001 for instance. It starts of with above 500K and drops down to 4 to 5 K in the lifetime user rate.

How I know this: They are all three p4 2,53ghz pc's with 1gb ddr 2700 mem installed. So they should give the same readings or at least almost the same. There is no work done on any pc while testing. They all run the same, with the same client.

Plz give me an explanaition....

[Archemedes]

I am seeing the same result here.

The graph of the last 24 hours is showing the line at ~100 while the client is saying ~500.

If the graphs are showing an average, it is pretty useless. It should correspond to the client reading at that particular date & time, ie; the work done that hour.
It seemed to work that way when I started, periods of inactivity showed up zero on the graphs and matched the client after restarting. But now it doesn't.

[royanee]

There is nowhere near enough server capacity to store that much information. Or if there is, it would be a much bigger waste. It's not really useful enough to multiply the stored data by 72 or more (considering 3 days per test or so).

[Archemedes]

Hmm, ok. Your saying the data is the running average cEM per test.
There seems to be data points stored already as the graph is not flat. The 24 hr graph could easily show the actual cEM while it would make sense to show the average on the overall graph.

Not a huge issue, just wanted to understand what I was looking at ;-)

[royanee]

Originally posted by Archemedes
Hmm, ok. Your saying the data is the running average cEM per test.

That is my understanding. If you run multiple computers, you will see interesting breaks in the graph. One thing is that when you haven't returned a progress report in a while, there will be a big zero from the last one until the current time. So if the computer has been off, and it sends another progress report, the graph will drop in height.