#internetarchive.bak 2017-12-23,Sat

↑back Search

Time Nickname Message
00:10 🔗 ez Senji: enterprise grade backups are entirely different price range, yea
00:10 🔗 ez the 100PB number is a bit arbitrary, as it assumes a lot of redundancy the system being super unreliable by itself
00:11 🔗 ez Senji: if we just assume 100% reliability its still something to the tune of 20pb. IA estimate is $2M/PB/eternity.
00:12 🔗 ez so we're talking about 40M $ in terms of raw market cost for iabak
00:14 🔗 Senji I think our figures work out closer to $1M/PB/etermity; but....
00:14 🔗 ez the eternity thing is the keyword here, as it assumes constant rotation of failing media
00:15 🔗 ez tapes dont last that long.
00:15 🔗 ez eternity figures simply assume cost of storage gets cheaper over time (15%, at ever slightly decaying rate matching current sigmoid trend)
00:15 🔗 Senji We still achieve 90% recovery from tapes written in the 1960s and not stored by idiots :-D
00:16 🔗 Senji (Data recovery and transcription is our _main_ business, archiving is a more recent secondary thing)
00:16 🔗 ez but yea, i can see how tapes could definitely have an edge
00:16 🔗 ez then again, thats modern super-dense tapes
00:16 🔗 ez which kinda push the limitis same way hard drives do
00:16 🔗 ez floppies from the 60s can be read by hand and magnetic probe
00:17 🔗 Senji Our $25/TB figure is based on last generation tapes and updating them every two generations
00:18 🔗 Senji We have some hand-built hardware for reading 5-track VAX tapes that has to be lovingly operated every few years when one of our customers finds a box at the back of warehouse somewhere :-D
00:18 🔗 ez Senji: $25/TB is not bad at all
00:18 🔗 ez thats definitely very competetive compared to jurry rigged consumer hw
00:19 🔗 ez i mean $10/TB is roughly the price you get a used 1/2TB drives for
00:19 🔗 ez and thats used drives, which definitely have far more reliability issues compared to tape
00:20 🔗 Senji I think thats cheaper than I can get them here in the UK :-D
00:21 🔗 ez Senji: so youre saying the tapes are fairly cheap on their own
00:21 🔗 ez one needs just super expensive gear to read/write those?
00:22 🔗 Senji And you either need really expensive robots or lots of man hours
00:22 🔗 Senji Both of our robots are second hand and still cost six figures of dollars
00:22 🔗 ez theres no need for real librarian robot
00:22 🔗 ez just one which takes pile of tapes, take one, put in recorder, pull it, throw on another pile
00:23 🔗 ez no need for random access as usual tapebots do
00:23 🔗 Senji We don't keep most of our tapes in the robots; but it saves a lot of time being able to shove them in once a week rather than every few hours
00:24 🔗 Senji We have little baby 10-18 tape robots for transcription work; but that's often like 50GB/tape so even so an operator has to change the robot every few hours
00:24 🔗 ez yea
00:24 🔗 ez i can see viability only if the tape is 1TB+
00:24 🔗 ez preferably to the tune of 10TB
00:24 🔗 ez 50GB is laughable, its cheaper to go optical unless that tape costs few cents each
00:24 🔗 Senji We're using 1TB tapes for our cheaper clients and 4TB tapes for our prestige client :-D
00:25 🔗 ez and that 1TB costs $25?
00:25 🔗 ez i can p much imagine that in raw manufacturing cost, but tape vendors tend to charge insane markups
00:25 🔗 ez the market being overspecialized and all
00:26 🔗 Senji I don't know if the $25 figure is for the 1TB or 4TB tapes (plus operator time to write them)
00:26 🔗 Senji I just happened to be in a conversation with my boss the other day where he was saying one of our competitors on a bid was taking the piss asking for $100 !
00:26 🔗 ez the jukebox bot can be rigged for fairly cheap
00:27 🔗 ez its really just two piles, or even human could do it if the tapes are 1TB+
00:27 🔗 ez its just 20000 tapes
00:28 🔗 ez Senji: yea, i've been far from ent ops for few years now
00:28 🔗 ez but back then, those tapes costed arms and leg
00:28 🔗 ez and those were like 100-400GB
00:28 🔗 Senji Oh eek, it's half past midnight, I should be asleep :-D
00:48 🔗 Somebody2 Interesting conversation, thanks!
05:18 🔗 Mateon1 has quit IRC (Read error: Operation timed out)
05:18 🔗 Mateon1 has joined #internetarchive.bak
10:10 🔗 db48x has quit IRC (Read error: Operation timed out)
10:18 🔗 db48x has joined #internetarchive.bak
11:13 🔗 db48x has quit IRC (Ping timeout: 255 seconds)
22:00 🔗 sep332 has joined #internetarchive.bak
22:06 🔗 sep332 has quit IRC (Read error: Operation timed out)

irclogger-viewer