Is E@H kinda grabby?
27 Jan 2014 13:28:29 UTC
Topic 197359
(moderation:
Hi,
back when updating to BOINC I'd allotted 2GB disk space for it...
Recently I noted quite some network traffic of my machine during idling...
Incidentally checking the [transmission] tap of BOINC just during that stage I found a_full_page of E@H WUs being downloaded (HUH?!)
Checking the disk usage revealed that by now E@H has puked 1,6GB into there, literally_starving_all other projects like S@H...
I've now stopped E@H from receiving new tasks and don't you know it, all over sudden S@H is able to receive a noticeable number on new WUs again...
Am I alone with this phenomenon?
cheers
Is E@H kinda grabby?
)
For most projects, this would probably be more than adequate but if you wish to participate in the gravity wave (GW) search, it's probably not. For just one task, you need quite a large number of large data files. There will be literally 1000s of tasks that can use the same set of large data files so once you have made the initial download, you can look forward to lots more tasks that use the same set of data. E@H uses 'locality scheduling' which tries to achieve just this.
Over the life of the full run, you will get additional large data file sets from time to time, particularly in the latter stages of the run. A safer limit would probably be in the order of 5GB or so. The data files should be removed automatically when all tasks depending on them have been completed and verified. If maintaining 5+GB of data represents a problem for you, you could use your preferences to exclude the GW run - S6 directed search (CasA). There is a new search for Gamma Ray Pulsars (FGRP3) which has just started which seems to have a very modest data file size.
That's not WUs, it's just a complete set of large data files. Your machine would have been idling because you can't start crunching a new task until all the data is downloaded. You never see the actual task appearing on the transfers tab. It's an extremely small set of parameters that is passed to the client during the scheduler request/scheduler response exchange. The good news is that subsequent tasks should use the same set of data for quite a while.
Yes, you will need quite a lot of space to store the sets of large data files you will receive over the full run. No, this will not affect your ability to download and process work for other projects, provided you have allocated sufficient space to store everything. If you exceed the limits you have set, BOINC will refuse to download new work/data but you should certainly see the complaints about that. If BOINC isn't complaining, your ability to get work for other projects will not have been impaired.
I doubt this has anything to do with E@H's large data files (unless you have started exceeding your disk space limit). The issuing of a work request has nothing to do with E@H and E@H has no control over it. BOINC decides what work should be requested and where to request it from. Of course, if you set NNT for one project, you are bound to get all subsequent tasks from the other one.
Cheers,
Gary.
RE: back when updating to
)
As your host is a Windows 7 i7-drive one, it seems likely to be pretty modern.
If you use a hard drive to hold BOINC data, then at the going price of about US$100 for a 1 Terabyte drive, your 2GB allocation is literally about US 20 cents worth. If you kick that up to a couple of dollars worth (20gB) I think you'll find yourself unlikely to experience disk space limitation on the SETI/Einstein combination.
If you are using an SSD drive, of course, I suppose it might be on the order of 240-256 GB worth roughly US$200, in which case your 2GB allocation is already something like US$2, and is already using up 1% of your available space.
If you have an unusually small SSD, or have a primary use for your machine which is unsually disk-hungry, then I could imagine allowing the BOINC applications to use what they need to pose a problem. But for the great majority of people running machines of your class here, this can only be a self-inflicted problem by setting too low a restriction to allow effective participation in the project.
The choice is yours of course, but I hope others reading this exchange with the same problem will think it reasonable to up their resource allocation. For most of them it will be zero drive expense, as simply allowing the use of vacant space already purchased.
I do agree that the current GW application, interacting with boinc and the server behaviors, can easily lead to remarkably large numbers of GW data files consuming a remarkable amount of space, compared to previous history. My personal case involved a modern machine which I had constrained to running a single GW CPU job at a time (down from two a month or two ago). It had 8 Gigabytes of GW data files when I reviewed it a couple of weeks ago.
I checked my disk. In the
)
I checked my disk.
In the program data/boinc/projects directory I have all my projects listed and their data stored.
The size of my Einstein project directory is 3.230.563.162 bytes (~3GB); this is without other projects and the boinc overhead.
So I mean the best thing you could do is increase the allowed disk space to > 3GB.
To have some numbers to compare:
Milkyway ~ 66MB
gpugrid ~ 102MB
LHC classic ~ 3MB
pogs ~ 302MB
simap ~ 500kB
collatz ~ 12.1MB
Albert ~ 262MB
RE: To have some numbers to
)
Which are nice numbers, but are of no use without telling what the connect to/additional value is set to when the global computing preferences are used through Einstein@Home, or the Minimum/Additional values are set to when the local preferences are set through BOINC Manager 7 (or the global preferences at the Seti@Home project webpage).
RE: RE: To have some
)
Oohh I forgot ...
min workbuffer 0.01 days
add. work buffer 0.00 days
Currently running Einstein only on my multiuse pc, no pending wu's from other projects.
running 3 nvidia, 3 amd, 3 intel gpu wu's + 3 cpu wu's
currently 21 wu's active/ready to run/ready to report
Hope this is complete now.
Alex
Thanx for the feedback so
)
Thanx for the feedback so far...
Well, its a 300(289)GB disk, so I followed the suggestion and increased the for BOINC avail HD space to 5GB and set [receive new WUs] again, time will tell...
But there is a slight lack on info about how much disk usage the (various) projects should get assigned...
cheers
RE: Thanx for the feedback
)
Einstein takes ALOT of disk space as compared to alot of other projects, but that may also depend on what units you are crunching. I am only crunching the BRP-4G gpu units, 3 at a time in about 1.3 hours each, with a cache setting of 0.50 and 0.25 days and Einstein is using 1gb of space for itself. I have my Boinc settings set to use most 1000gb of drive space, leave at least 0.10 gb free space and to use at most 90% of disk space. I used the default install parameters so mine is sharing the C: drive with Windows etc, but if you have a secondary harddrive in the machine you can install Boinc there and it will run just fine, you just need to do it during the install process, not later on.
RE: ...if you have a
)
Slightly hard to realize on an i7_64 notebook ;-) (whilst dual disk units getting common these days, SDD for system/OS, conventional IDE/SATA for data storage...)
But maybe the concept of BOINC and the project concepts were designed while having primarily multi drive desktop towers and server housings in mind...
cheers
If disk space really does
)
If disk space really does matter, a usb storage device with 8GB may be a solution. The boinc data directory can be placed there and the setting for the usage of disk space can be set to 100%.
A notebook usually has also a slot for a sd-card, which is also an option.
RE: RE: ...if you have a
)
Yes and no, as Alex said a usb stick can easily be used and since lots of people are using laptops as desktops these days, and we ALL know how important backups are, alot of people have portable hard drives plugged into their laptops. Lots of options out there including even getting a new and bigger hard drive for the laptop, assuming space is a problem. I bought mine last year and it has a 1tb drive in it, it is an i7 with HT so can use all 8 cores for crunching and still have plenty of drive space for any size cache. It currently has 841gb of free space on it, I could not find the one I wanted with a smaller drive in it.
Yes Boinc was designed for desktops in mind but it has long since evolved into supporting most operating systems and devices. You can crunching on desktops, laptops, tablets, cellphones, the Raspberry Pi and almost anything else you can think of except the Apple i-devices. Apple is propriety and I am guessing hasn't let Boinc in yet. Here is a link to a webpage of lots of Distributed Computing projects http://www.distributedcomputing.info/projects.html
The Boinc ones are noted, so if you have a device that you want to crunch with check it out, there should be at least one project that will let you use it.