|
Biowarfare posted:Once the plots are generated it's read many. Green eco crypto! (that will landfill a ton of SSDs). If you had at least 384GB of RAM, you could plot in a ramdisk, which would be far less henious
|
#
?
May 8, 2021 07:13
|
|
|
|
| # ? May 26, 2024 09:25 |
|
|
Moey posted:I thought all 3tb drives were ticking time bombs? Anecdotally, I have a 3TB HGST NAS Drive which has been in 24/7 use for about 6 years with zero reported bad sectors.
|
|
#
?
May 8, 2021 14:30
|
|
|
Same here, my 3TB HGST HDN724030ALE640 has been flawless.
|
|
#
?
May 8, 2021 18:57
|
|
|
KozmoNaut posted:Same here, my 3TB HGST HDN724030ALE640 has been flawless. Yep, exact same drive 
|
|
#
?
May 9, 2021 01:47
|
|
|
13 GB NVME??
|
|
#
?
May 9, 2021 06:22
|
|
|
Biowarfare posted:13 GB NVME?? M.2 Optane device for cache.
|
|
#
?
May 9, 2021 07:22
|
|
|
So I've got the DS1821+ setup using three 14tb drives. Now I have data from the old Windows server that's on 3x 4tb drives and 3x 2tb drives that I need to start copying over. My original plan was to install the drives from the Windows box into the Synology and copy everything locally, but it doesn't seem the Synology will mount NTFS drives internally, only via the USB interface. I got that from googling around, so please tell me if I'm wrong. I have to imagine it will take 1000x longer to do ~16tb of data via USB than doing it internally, but I'm new to Synology so I don't know. How would y'all go about transferring this much data over?
|
|
#
?
May 9, 2021 18:24
|
|
|
Do you have a decent network connection between the old windows server and the new Synology? (Or just a Windows machine if the old server has died). Transferring that way will likely be as fast as the drives can do.
|
|
#
?
May 9, 2021 18:40
|
|
|
For those that accept it, does Synology really care what m.2 drive you stick in there? I know they sell their own.
|
|
#
?
May 9, 2021 18:46
|
|
|
Rooted Vegetable posted:Do you have a decent network connection between the old windows server and the new Synology? (Or just a Windows machine if the old server has died). Transferring that way will likely be as fast as the drives can do. It's gigabit LAN. I mounted the Synology share on the old Windows machine and did a test, about 20 minutes for 85gb. So around 3 days worth of copying stuff that way, if I'm doing the math correct. Guess I'll be babysitting it for awhile.
|
|
#
?
May 9, 2021 18:53
|
|
|
EC posted:It's gigabit LAN. I mounted the Synology share on the old Windows machine and did a test, about 20 minutes for 85gb. So around 3 days worth of copying stuff that way, if I'm doing the math correct. Guess I'll be babysitting it for awhile. Only if you watch. Go do something else.
|
|
#
?
May 9, 2021 18:58
|
|
|
EC posted:. So around 3 days worth of copying stuff that way, if I'm doing the math correct. True but that's within range of the maximum throughput of the drives in question. I don't think LAN transfer is slowing you down vs. locally connected read. You're gaining a lot of not worrying about filesystems etc and are using both devices as intended.
|
|
#
?
May 10, 2021 03:24
|
|
|
KKKLIP ART posted:For those that accept it, does Synology really care what m.2 drive you stick in there? I know they sell their own. It doesn't but it is only supposed to be for caching so a high endurance one is best.
|
|
#
?
May 10, 2021 03:28
|
|
|
buffbus posted:It doesn't but it is only supposed to be for caching so a high endurance one is best. You're probably rewriting less than you think. Are you tracking total bytes written?
|
|
#
?
May 10, 2021 04:05
|
|
|
Depends on what you're doing and if you have it as a read-only or read/write cache Ours burned out an Intel SSD DC S4500, it was being used as a backup target in Veeam and it had write-caching turned on. Derp. My particular synology has a slot for either a 10gb card or cache SSD, but not both  So we use SATA SSD's for cache drives.
|
|
#
?
May 10, 2021 12:29
|
|
|
Biowarfare posted:Once the plots are generated it's read many. Doesn't it do all that in RAM though?
|
|
#
?
May 10, 2021 13:14
|
|
|
no it's all disk io
|
|
#
?
May 10, 2021 18:08
|
|
|
Rooted Vegetable posted:True but that's within range of the maximum throughput of the drives in question. I don't think LAN transfer is slowing you down vs. locally connected read. You're gaining a lot of not worrying about filesystems etc and are using both devices as intended. It's going pretty well so far. I've got all the data off 3 of the 4 2tb drives. I'm worried about the last one as I was having some issues accessing it last week, but it seems to be working fine so far. Any good tutorials for learning and setting up Docker for stuff like SAB/Sonarr? I have a real basic, surface level understanding of what Docker is and why it's good to use it. According to the Sonarr discord there's no real way to migrate an install from a Windows box to a Synology, and I am dreading having to go in and adjust all the series and replicate everything. SAB should be easy, although it will be sad to lose the 11 years worth of history on the thing.
|
|
#
?
May 10, 2021 18:24
|
|
|
EC posted:According to the Sonarr discord there's no real way to migrate an install from a Windows box to a Synology, and I am dreading having to go in and adjust all the series and replicate everything. SAB should be easy, although it will be sad to lose the 11 years worth of history on the thing.
|
|
#
?
May 10, 2021 20:07
|
|
|
wolrah posted:It is definitely possible to edit the database backup directly to keep the existing data with new paths, but whether it's worth the trouble is up to you and depends on your comfort level with SQL. I have zero experience with SQL, unfortunately.
|
|
#
?
May 10, 2021 20:36
|
|
|
Biowarfare posted:Once the plots are generated it's read many. And that'd take a while on spinning rust so people just buy cheap SSDs to burn, right? Guess it's a good thing I got one of the fancy new PCIE 4s before the rush 
|
|
#
?
May 10, 2021 21:06
|
|
|
Yes, that's the point. It sucks up both slow and fast drives. You need the slow drives for long term storage, so all large storage is gone. You need as many fast NVME or SSDs as you can to trash through to write as much as possible, as fast as possible (to generate the bingo boards to put into the rotational drives), so all fast storage is gone. The min size requiring 300G+ means that 128/256GB SSDs are largely untouched, though
|
|
#
?
May 10, 2021 21:33
|
|
|
I'm decommissioning my old Unraid server which has a bunch of still working, if old disks. Is there a small enclosure or something similar I can get to either use these as a huge USB array or JBOD? Or just on the network is fine as a backup/no frills file storage device. I already have a Synology DS1819+ so that is doing all the heavy lifting for me already.
|
|
#
?
May 10, 2021 21:41
|
|
|
Gyshall posted:I'm decommissioning my old Unraid server which has a bunch of still working, if old disks. I looked for cheap jbod enclosures recently but they were surprisingly expensive.
|
|
#
?
May 10, 2021 22:15
|
|
|
Biowarfare posted:Yes, that's the point. It sucks up both slow and fast drives. You need the slow drives for long term storage, so all large storage is gone. You need as many fast NVME or SSDs as you can to trash through to write as much as possible, as fast as possible (to generate the bingo boards to put into the rotational drives), so all fast storage is gone. The min size requiring 300G+ means that 128/256GB SSDs are largely untouched, though A few minutes thought while I was on the Thinking Throne a while ago (because I forgot my phone) and I figure it'd be cheaper to build/buy a server with ~1TB of RAM once then have it generating plots on platter drives that you migrate to, IDK, external USB enclosures hooked up to Raspberry Pis. Then you only pay for one piece of plotting hardware that doesn't wear out and you could even have hot-swap bays for zero downtime plotting. Maybe that USB enclosure thing doesn't strictly work with throughput constraints but it'd be what I'd do if I was serious about touching the poop.
|
|
#
?
May 10, 2021 22:45
|
|
|
Biowarfare posted:Yes, that's the point. It sucks up both slow and fast drives. You need the slow drives for long term storage, so all large storage is gone. You need as many fast NVME or SSDs as you can to trash through to write as much as possible, as fast as possible (to generate the bingo boards to put into the rotational drives), so all fast storage is gone. The min size requiring 300G+ means that 128/256GB SSDs are largely untouched, though So they design a whole new "thing" and instead of making it less insanely (with chunks that can be derived and signed in a standard chunk of memory and written once to disk) they've done it so it far exceeds anything but large servers, instead choosing to use basically large scratch disks - chewing through flash? RAM is built to constantly be re-written, and plenty of nerds these days have 16-32GB in their desktop, or computers which could be upgraded to that amount. Why not make it in 4 or 8gb "plots" ? I assume it's just so they don't have to store as much metadata on the blockchain, but you could also then have them built to be stitched together as they're written once to the disk. Sorry this just blows my mind that they saw a problem, thought "we could fix this!" and then just made it dumb in a different, less renewable way.
|
|
#
?
May 10, 2021 23:11
|
|
|
H110Hawk posted:Sorry this just blows my mind that they saw a problem, thought "we could fix this!" and then just made it dumb in a different, less renewable way. They don't care, they have a shitcoin that they pre-mined 21 million of and are hoping the noveltly will let it take off so they can cash out.
|
|
#
?
May 10, 2021 23:28
|
|
|
Chia is a scam.
|
|
#
?
May 10, 2021 23:29
|
|
|
redeyes posted:Chia is a scam. Aren't they all?
|
|
#
?
May 11, 2021 03:17
|
|
|
Volguus posted:Aren't they all? Yes, the oceans will burn.
|
|
#
?
May 11, 2021 03:27
|
|
|
Inept posted:They don't care, they have a shitcoin that they pre-mined 21 million of and are hoping the noveltly will let it take off so they can cash out. Right you said it was crypto.
|
|
#
?
May 11, 2021 03:40
|
|
|
We could always tie the people to their mining equipment, and throw both into the ocean.
|
|
#
?
May 11, 2021 11:07
|
|
|
Just put in a tree planting tax per gigabyte of storage purchased via any system. Clearly that’ll fix all externalities
|
|
#
?
May 11, 2021 14:39
|
|
|
Nitrousoxide posted:Doesn't it do all that in RAM though? I've seen someone run chia mining in a ram drive. It seems to run the plots fine if you have 512 GB of RAM.
|
|
#
?
May 12, 2021 07:35
|
|
|
Devian666 posted:if you have 512 GB of RAM. So then what's the expected break even ROI date for THAT?
|
|
#
?
May 12, 2021 22:46
|
|
|
AlternateAccount posted:So then what's the expected break even ROI date for THAT? I bought 512GB of ddr3 server ram for under $400, so probably decent (unrelated to mining, this is for a database server)
|
|
#
?
May 12, 2021 22:51
|
|
|
AlternateAccount posted:So then what's the expected break even ROI date for THAT? Less than a quarter if the rumors of chia destroying SSDs within weeks are true. As is usually the case with cryptocurrencies, what software engineers think is a limiting factor is something that accountants will work around. Bitcoin limited by energy turns into "arbitrage electricity prices, stealing energy outright where possible." If Bitcoin existed contemporaneously with Enron, they probably wouldn't have gone bankrupt. Chia is operating in the space that every cloud storage and big data company has already optimized down to slivers of margin. Once you know how to game opex/capex, amortization, and depreciation you can make very expensive things very cheap as long as the value of your product stays above a certain amount. xarph fucked around with this message at 23:22 on May 12, 2021 |
|
#
?
May 12, 2021 23:15
|
|
|
AlternateAccount posted:So then what's the expected break even ROI date for THAT? In one of the goon discord channels someone has been test mining chia to understand it. Calculations on the current price indicate if you buy 50TB of drives you could have them paid for in about a week. Now I've said that we'll see how many hard drives get destroyed. People are mostly concerned with SSD failures but what's going to happen to WD's SMR drives that are still out in the wild (and cheaper than the CMR versions so miners will buy them).
|
|
#
?
May 13, 2021 06:19
|
|
|
Devian666 posted:In one of the goon discord channels someone has been test mining chia to understand it. Calculations on the current price indicate if you buy 50TB of drives you could have them paid for in about a week. Admittedly I won't shed a single tear for a dead SMR drive.
|
|
#
?
May 13, 2021 14:12
|
|
|
|
| # ? May 26, 2024 09:25 |
|
|
The load on the storage drives isn't that high, is it? You generate the plots on the SSD or in RAM, then write that info to the storage drives. Then you just have to read from them, to check for matches, right?
|
|
#
?
May 13, 2021 14:28
|
|