What does this mean?
Dragstk
Syracuse, N.Y.
Sometimes when I get a new WU, in the middle of the download speed, I get this:
Conversation time very short, giving reduced weight in bandwidth avg
While I know it's nothing critical, I was just wondering what it means. Is there some giant Folding book of definitions I could look in, or do I have to keep buggin' you guys :bigggrin:
Thanks;
Dragstk
Conversation time very short, giving reduced weight in bandwidth avg
While I know it's nothing critical, I was just wondering what it means. Is there some giant Folding book of definitions I could look in, or do I have to keep buggin' you guys :bigggrin:
Thanks;
Dragstk
0
Comments
Keep on asking. We will know most of it but I have never heard of that one yet. Plus once someone kinds the answer, the rest of us get to find out as well.:)
Would that be on dialup?
I'll take a run over to the Folding site, see if I can learn anything. Here's a little expanded view of what I was asking about.
[19:33:33] Loaded queue successfully.
[19:33:34] - Receiving payload (expected size: 16796)
[19:33:34] Conversation time very short, giving reduced weight in bandwidth avg
[19:33:34] - Downloaded at ~32 kB/s
[19:33:34] - Averaged speed for that direction ~121 kB/s
[19:33:34] + Received work.
No, I'm not on dial up, I have cable.
Thanks;
Dragstk
OH, client has a parm called lifeline. It is a variable used to determine in part how often the client tries to reconnect and how well you connect on average affects the lifeline value. Client will check for failed sends less often if it is a very good value. so, since they are weighing the connect length to figure out how clients are really connecting after some folks had real issues with connects off and on, they talk about weight. MOSt of this weighing is done at Stanford, BTW, they use it to find weak net links in their network, double-check on their bandwidth providers, etc.
BUT, if they have an area in remote location that is serrved by one ISP, they can send stats to that ISP as to poor linking often. Stanford actually was one of the first on the web-- it was poart of the first four universiites to be part of DARPA's littel universisty net that became the web. As a result, they have influence as part of participating in FOUNDING the web in US. NOTE, Europe's USENET was around about the same time, that is why I say US web.
In my case, I looked up at the client call lines, after and before that set of lines, noted that the lifeline value had changed, for my two folding clients here-- I am also on cable. Lifeline tracks an average of number of retries needed to get work turned in and receive work, and how long on average communications is happening, etc. I get those things for 5-6 connects out of first 20-30 work units, then client is happy and does not change lifeline much at all unless I get something unusual that makes for real bad or good connect and conversation between client and Folding servers (that is the conversation part, software to software between client and server-side folding software at Stanford's end).
John D.-- who freely admits he is of the old school that believes in "checking up on the box" and mines data and program and O\S logs for interesting patterns...
The client is attempting to estimate your download speed. If you've got a very fast connection and a very small WU, the resolution of the clock is poor enough that you can't get a good speed measurement. The number will be averaged in with many other downloads, but it receives a reduced weight, in an effort to base the average on the more accurate data.
Learn something new every day...I hope
True-- but in my case the average became what happened after the very fast rate. This happened with a BIG WU, then two more scattered later of which one was a small one (which fits the explanation) and then another big one. Each result was just out of range for previous average for that client install. So, rather than toss the average way out, the actual single numbers were not calced into the average download as if they were typical.
Ever heard of a bell curve??? Some downloads might happen when Folding has low load and the ISP is not hyperbusy, others might happen when one or both are busy. Average lets folding predict what is needed, and from their end the numbers versus server load let them look and see if it is their end that is busy-- if lots of folks have problems downloading from one server from many clients located in may places, then they look deeper into the functioning of the server.
Let's take a WU from the Barton box, which I am using right now to post. Lifeline is 2332. WU downloaded at ~289KB/sec this is a project p936_fkfe2_all worth 107 points (found that out today). NOT a small WU. Average download is ~170 KB/sec right now for the Barton box. There were some WU downloads that came in at ~110KB/sec when Comcast or Folding was busy. At time of download, the Lifeline went to 2324.
So, they use weighing to get a true tracking average. Some will be slower, some higher. The ones that are REAL high versus average are NOT tossed out completely, they simply do not weigh into the average calcs as if they were typical.
Now, Linux is more efficient with pure upload and download networking the way I have it set up. Lifeline for the Linux folding client is 3287 right now. The download for the Linux client is averaging ~330KB/sec as of now. Lifeline goes up with effective downloading, down with less effective downloading.
My two computers are and have been getting lots of big WUs.
John D.