Pages: [1] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Mar'Dur Taren
Gallente Priory Of The Lemon Atlas Alliance
|
Posted - 2010.05.07 17:03:00 -
[1]
I've been testing running processors in parallel. I have a Storage Facility linked to two identical processors which both link to a new storage facility. I've done this pattern with basic and advanced processors and noticed some bad routing logic from the first storage facility.
Starting with the basic Processors. I had one processor (1) working on a batch and holding 73% of the next batch ready. The other processor (2) was working on its own batch but had no stored reserves for the next batch. Then the extractors emptied a load of raw material into the front storage facility. Instead of filling up the almost full processor (1), the routes sent the load to the empty processor (2). Now I have two processors which between them have enough for another run but its spread over the two so neither works.
Things get worse when you look at the advanced processors. Each needs two different ingredients to work. In the storage facility, there was enough of each product to set one processor running (40 units item A and 40 units item B). I set up the links and they go live. The routes send 40 units of item A to one Adv Processor and 40 units of item B to the other processor. Again I have enough for one processor to work but its spread between two nodes. Now I have two processors waiting for parts.
The routing logic needs to be a bit smarter. If more than one processor is fed by a storage facility, then the material should be routed based on a set of priorities like these;
1. If a processor is not working and has stored material. 2. If a processor is not working and has no stored material. 3. If a processor is working and has stored material. 4. If a processor is working and has no stored material.
This system should fill each processor once then come back and fill up the stores on each again if there is surplus. Further there should only be one processor with a partial store of materials. The others will either be full or awaiting material. Also this tries to get the processors all working at once which is what you want. The current system has processors sitting around waiting for a random ration of material.
|
Kaon Vortex
Gallente Middleton and Mercer LLP Consortium.
|
Posted - 2010.05.07 21:19:00 -
[2]
/signed |
Cadde
Gallente 221st Century Warfare
|
Posted - 2010.05.07 23:42:00 -
[3]
Edited by: Cadde on 07/05/2010 23:47:31
Originally by: Mar'Dur Taren ... words
The other option is to make sure you have enough material to keep BOTH processors running non stop. It's better to have excess than to have too little.
Imagine this situation:
You have 6 extractors all producing 550 units every 15 minutes for 24 hours. After 30 minutes you have 6,300 units in storage. 6,000 units goes to processor leaving 300 units in storage. After 20 cycles (5 hours) you have enough produce to run another (extra) cycle of your processor. After the full 24 hour period you have enough to run 4 extra cycles with your processor (another 2 hours of processing)
Now after the 24 hours are up, lets say you set up another extraction for 24 hours and you are still producing 550 base material per extractor per cycle. After 48 hours you now have enough material to run 9 extra processor cycles. (4 hours, 30 minutes extra) Now do another 96 hours (6 days in total that is) and you have enough to run 28 processor cycles. (14 hours)
Let's assume it would take you 30 minutes to set up a new extraction farm. You started the first extraction at 5 pm in the day so after 6 days your extractors would run out at 8pm on day 6. You now have enough base materials in storage to run until 10am the next day.
Now, lets assume you you started this whole process on a Saturday at 5pm, on Saturday a week later at 10am you will have used what was in storage on friday for 14 hours. (Go out, have a beer on Friday night. Your PI empire is doing just fine) At 10am on Saturday morning you set up a 5 hour cycle to keep your farm busy. Your extractors stop on 3pm, your storage runs out at ~5pm. Rinse, repeat, profit!
At most, your storage unit will hold 6,912 m3 of base materials (345,600 units) so you won't have to worry about running out of storage space.
Don't take what i said the wrong way, it's just a tip to make sure you would get the MOST out of your production chain as opposed to leaving one processor idle at any time because you can't produce enough to keep it running.
Also, even if BOTH are idling it doesn't mean you are running slower, it just means you don't have enough produce to keep both running and they will still output the same amount of units every hour as they would if one had proi for base mats over the other.
EDIT: Btw, this might sound an awful lot like spreadsheets online. However, i just did some rough calculations in my head and used the calculator to give me the accurate numbers. You could use a spreadsheet or you could do the math in your head rounding off as you go. One 30 minute cycle more or less isn't the end of the world in PI. However, spending that 600,000 isk extra on another processor that will idle most of the time is a waste of isk as the extra speed doesn't make it any more profitable.
My opinions belong to me, you can't have them!
|
Mar'Dur Taren
Gallente Priory Of The Lemon Atlas Alliance
|
Posted - 2010.05.09 15:38:00 -
[4]
Seems to me that what you're talking about is basically keeping a single processor full at all times. Personally the output from on processor is not enough for me. I was experimenting with trying to run 1.5 processors or two if I could.
The basic reason for this is to increase production. Basic Processors are limited to 48 cycles a day. Advanced Processors are limited to 24 cycles a day. Call me greedy but that's not enough for me. I found this issue when trying for the "One and a bit" processors working model.
I still think its a valid issue. From my observations it appears that the routes supply processors based on which can take the whole available supply. Haven't had the situation where both processors were nearly full. A little hard to engineer that situation.
|
Cadde
Gallente 221st Century Warfare
|
Posted - 2010.05.10 05:58:00 -
[5]
Edited by: Cadde on 10/05/2010 06:03:00
Originally by: Mar'Dur Taren Seems to me that what you're talking about is basically keeping a single processor full at all times. Personally the output from on processor is not enough for me. I was experimenting with trying to run 1.5 processors or two if I could.
The basic reason for this is to increase production. Basic Processors are limited to 48 cycles a day. Advanced Processors are limited to 24 cycles a day. Call me greedy but that's not enough for me. I found this issue when trying for the "One and a bit" processors working model.
I still think its a valid issue. From my observations it appears that the routes supply processors based on which can take the whole available supply. Haven't had the situation where both processors were nearly full. A little hard to engineer that situation.
Hmm. So what you are saying is that you'd rather make less isk (possibly no isk at all) as long as it's slightly faster isk?
The bottom line is, if you can't produce enough raw materials to keep 2 processors running non stop you are most likely going to be better off running one processor non stop and use the excess materials for times when you can't visit your PI network right on time.
The other option is to find a planet that has enough raw materials to supply 2 basic processors non stop.
The planet i am working right now (0.1 sec) can provide enough raw material from 4 extractors to keep one processor running non stop on 24 hour extraction intervals. That means i can run at least another 3 chains just like that with an elite command center. All of that with excess material going into storage. It's about finding that balance between number of processors and number of extractors to not waste money on building infrastructure that isn't gonna pay for itself in due time.
However i still think CCP should drop the whole processor storage buffer and draw from the storage facility instead. That doesn't mean running 1.5 processors is going to be the optimal solution.
EDIT: And even if both your processors are idle just as you look at them it doesn't mean you are producing less.
That is, if you are extracting exactly 216000 units of raw materials over 24 hours then your 2 processors will have processed that material as soon as they can even if they by chance happens to be stuck on 99% by the time the last extraction cycle has finished.
They will run exactly as many cycles when competing for materials as they would have if the second only started running when there was enough excess to make that run one cycle.
My opinions belong to me, you can't have them!
|
Janus Ovellian
Minmatar Calpolli Namtz' aar K'in
|
Posted - 2010.05.10 12:32:00 -
[6]
/signed
Interesting times await... |
Tres Farmer
Gallente Federation Intelligence Service
|
Posted - 2010.05.10 12:41:00 -
[7]
yeah.. routing and especially feeding several procs from same buffer-storage seems to be to dumb atmo. I wish procs did have a switch to get rid of the inbuilt storage.. this way we would get rid of this problem fast as a whole.
*) a switch, cause there are for sure peeps out there who want to run straight setups from extr to proc.
|
|
|
|
Pages: [1] :: one page |
First page | Previous page | Next page | Last page |