The number of TCP connections was pretty small. I’m pretty sure it was a DDoS of a different port/protocol, though I’m not sure what kind because my router does not show this.
Usualy DDoS is just lot of connections from different sorses, because it not posible to overload from 1 place. So if router is HOME router, then it need not so much for overload with peers.
Chia is also made introdusers, that exchange peers ip, as peers with port forwarding is not os much, so some peers get lot of load. I seen that i had also 20-40 Mbit trafik from and to chia network
1 Like
I’m wondering if it’d profitable to mine chia on storage space hosted on Storj 
Who’s up for renting 1PB for Chia mining? Storj node operators are ready to welcome your plot data! 
… Well I guess it wouldn’t be profitable ^^’ It’s a shame, this would bring a lot of data to the network 
harvester need to read plots in some seconds, it is not posible to do it on storj
Is it not ?
Accessing a file on the storj network is supposed to be pretty fast as they can host websites on it.
So why wouldn’t it fit chia mining?
website is 1-50mb, 1 plot is 100GB do you see the diference?
decryp so big files on fly i think not posible, and need to reed 1000 plots per second
Also it wil be not profitable, every 18s need to read 1Pb of data, it is very expencive, and i seen post that guy with 1.3 Pb win every 3-5 days only
1 Like
You’d need random access to parts of files. Storj so far only has a “download the whole file” mode. You could probably fake it splitting a plot into many small files, but that’s probably a lot of work to implement.
Edit: had outdated information, sorry.
1 Like
I wouldn’t be so sure about that. It doesn’t ever read entire plots, just very specific parts. Read amounts are actually really small. My 11TB worth of plots generates just 5GB of reads per month.
I don’t see why using s3fs in conjunction with the gateway mt wouldn’t work for finished plots. I don’t really have an always on Linux system other than my NAS. But I might try to find a windows solution for this and just test with a few plots to see if it’s viable.
Edit: I guess the real question is whether the entire chain of patchwork services can properly deal with the partial reads and only download those stripes that are needed. Right now I’m looking into whether I can use s3fs in WSL and access the location from the windows host. So that would use Chia Client > SMB > s3fs > Storj Gateway MT > Storj Network. To be clear, I don’t think this is a great idea. I’m just checking for educational purposes what the effect would be. I’m already uploading 2 test plots which seems to take about 2 hours per plot, so that’s already kind of a limitation. It would really be great if there were a way to map a Storj bucket as a network drive directly in windows (or linux for that matter).
Edit2: s3fs on WSL is a no go. Not supported. I should have guessed.
2 Likes
Another problem is the plotting, you would burn through many nvmes and need gigantic amount of processing power to generate more plots than the average network growth. So even if storing plots on storj works, you’d spend a fortune to create them and it wouldn’t be very profitable.
2 Likes
If that is the case how is streaming accomplished? For example this one:
http://link.tardigradeshare.io/s/1N2isb7JVSzs8s9w91f2f8XMBfdkqinvSWrP9poPYQWbAuUhrb6TX7V3YbLCsu5vJkq4oHwkz6QP4mpJZswF3nZPzC3nUbvFWRfuXjdzJVLfHMvSVDPRZH2KxegRNpTKPcf2XbRYmD7yfmkuZzuhVyS3A12BemXRF9w7pYutQPbBJgbiaWzh1PxpkbrhR3hzwUs4QUPNX3PgT5psNo2G6mYUf7rxrftK81WYE9jSTBh7c8FoDVyYg3KFzRsZveSSHHHodNtWqLxJ4jgR4Wf685aS7KEESMVqpgBB5PNvfgEEWRF49/fireplace/Fireplace%2010%20hours%20full%20HD.mp4?wrap=1
If you jump back and forth, the file starts to plays pretty quick, so I doubt the full almost 15 GB have been downloaded in total. I have no idea if the implementation of Chia would allow it to use Storj, but it looks like at least some random access is possible. But I have no idea really.
This is absolutely not true. You can download at the stripe level (similar size as audits use). So the question is only whether the software/services chain you’re using for access all supports jumping into specific parts of a file only.
An update on my little experiment. After trying and failing to use several tools, I found something that works. But there are some pitfalls to avoid.
So what didn’t work
- As I already mentioned, s3fs won’t work within WSL
- TntDrive refused to connect to the gateway MT (I don’t fault Storj for this as it’s the only thing I’ve used that doesn’t connect to the latest version of the gateway)
- ExpanDrive seemed to work but then just started throwing read errors when trying to access part of the files.
- S3 Drive with default settings attempted to cache entire plots on a drive that didn’t have enough room
What eventually worked was using S3 drive, but making sure the caching options are turned off. This isn’t immediately clear how to do, as you just have to remove the path for the caching location. Removing the caching location prevents it from trying to cache entire plots. When adding the new plot location to Chia it did take a few seconds to find the plots, but it seems to be perfectly happy now.
EDIT: Also use read only mode for the mapping and mount to an NTFS folder instead of a drive letter to prevent windows from trying to write stuff to the read only volume and causing constant popups.
I’ll keep an eye on egress, but I might need to ignore egress from today as I think it tried to cache an entire plot, which may have generated 100GB of egress right away. I’ll keep you posted, but so far this seems very doable. Network traffic used by S3 Drive is minimal, so I don’t think there is a lot of egress traffic involved.
Please be aware that different S3 mapping tools may work completely differently. Some may always cache the entire file, others may cache relatively large chunks and may not allow you to disable that. My experience only applies to the tools I used. That said, I think this should probably work with s3fs as well as it allows you to disable caching as well.
6 Likes
Cool, thank you for that experiment 
I would be happy to get an update how this works out in the next days
thanks for this @BrightSilence. Ive also wondered if this would be possible (although probably not the best idea) and if Chia would be able to read the plot quick enough from Storj network.
Do let us know any updates.
One short update. The biggest challenge is finding the right software to map S3 storage as a network drive. I know that Storj is perfectly capable of handling the access patterns for this use case and I’m pretty sure that s3fs is pretty solid at this point. But on windows, it really isn’t nearly as simple. I can already say that the performance of S3 Drive is horrible. I tried uploading a plot through it and after running about 30 seconds at 3MB/s, it just plain fails the transfer.
Even weirder, something I did has just removed one of the 2 plots I was using for this test. I am now reuploading that (using the new object browser in the satellite interface, which is actually really fast). And I will try mounting the bucket as read only next to prevent this from happening again.
If anyone knows of good software for windows to map S3 buckets to a local or network drive, please let me know.
EDIT: Another small update, I’m up and running again, but something started attempting to write to the location and instead of just denying that, S3 Drive pops up a message to say the write was blocked every time, which becomes incredibly annoying and spammy. I think this was just windows going rogue and trying to put stuff in the root of a new drive like it usually does. So instead I now mounted the location to an NTFS folder, so windows would keep its hands off. Seems to work so far.
1 Like
Oh, sorry then, it seems I had outdated information.
It doesn’t really matter, but this has always been the case. This feature was there when the network went live to support both streaming video and of course audits as well as other use cases. Though I don’t think they had Chia in mind, haha. You had to use libuplink though as I don’t believe the CLI interface can do this. Not entirely sure about that part though.
Alright, I have some early results. After about 24 hours of use, my 2 plots generated about 2.4GB of egress. So we can go with 1.2GB per plot. To make it a little easier, lets take a hypothetical 1000 plots for roughly 100TB worth of plotted space. that would lead to roughly 1.2TB of egress per day at $7. Lets just call that $10. 100TB of storage for a day would cost about $13. So costs between $20 and $25. Those plots would at the moment earn about $58 per day. So technically profitable right now.
But:
- Even with 100TB it would take on average a month to win a block
- I haven’t tested whether plots passing the filter can solve problems fast enough to actually claim a block reward
- Even if they could with my 2 plots, that may be a lot different when the connection is being hammered by 1000 plots.
- Higher bandwidth may put pressure on stability of the solution (I know for sure S3 Drive would not be able to handle that, maybe s3fs would)
In short, I have no idea whether this scales.
But you gotta admit, it’s pretty cool that this works on small scale to begin with!
So…
Technically short term it might be. At least for another week or so. 
But disregarding how bad of an idea it is, this
isn’t true, at least at small scale and would possibly even work at larger scale. Which is pretty awesome.
3 Likes
But hey, @StorjLabs, if you feel like putting some more interesting test data on our nodes and taking a chance on winning some Chia as a bonus. This might be a cool project for testing purposes!
5 Likes
It will be very interesting if Storj start to mine chia
1 Like