This is turning into a typical/. discussion. Guys who haven't had experience verus those who have. For those who want to use disk backups: Try that when you are backing up 3TB+ a day. It's all a question of scale.
Tape won't die becuase it's alredy been used for 30+ years and there is a load of data out there that needs to be kept. There is a significant market just keeping this alive. Although the tape manufactureers would love us to do it, it is very rarely viable to pull back and refresh all the media you generate. Net result is that large backup and achival systems, once in are there for keeps. This somewhat explains why initial capital costs are so high, as refreshes for SMEs are rare. In orgs that have rampant out of control data growth ( this is most believe me - sub rant:.PSTs are evil) refreshes come along at 18 or 24 month intervals. This coincides nicely with the doubling in capacity that tape drive manufacturers produce.
The real killer though is not the hardware but the software, keeping old backup systems alive to restore your old media is a nightmare. Anyone tried to restore an Novell system recently?
This is from practical experience - I'm a backup admin and I test backup technologies for million dollar installs.
AIT-2 Native 6MB/s Real 9MB/s AIT-3 Native 12MB/s Real 20MB/s LT0 Native 15MB/s Real 23Mb/s
LTO2 looks like boing a big win at a native 30Mb/s but the issues with it will be that it will be very tricky to get it to stream effectively at 50Mb/s. You gonna need a fast server to keep it supplied at that rate.
Sounds obivious doesn't it, but in reality space is an issue. There are no jukebox technologies that can store DVDs as densly as tapes, when you measure in TB/sq ft. You'd have to build a big ass robot for DVDs - I mean HUGE.
There is also the issue of speed. DVD etc just cannot match a LTO drive going at 25Mb/s full whack (and that's a medium speed these days)
Point in case - I work in the finance industry and run their backups (5TB/day). I could fill a DVD-R in under a minute and I'd need 000's per week. I'd spend my entire life just unwrapping the buggers.
However we are all looking forward to a holgraphic disc technology, and have been for the last 2 years. No signs yet...
Exactly. MTBF stands for Mean time BEFORE failure not BETWEEN. This is because a failure in these components can usually only happen once and it is usually fatal.
Tape media technology is having a hard time keeping up. We use lots of AIT-2 drives (40+) in a big silo to keep up with data growth. Tape technology follows Moore's Law but at a slower rate. It seems to double every 18 months in capacity. There's no way it can keep up with this level of growth in disk capacities.
Some tape transfer rates: DLT8k 5MB/s 35GB AIT-2 6MB/s 50GB AIT-3 12MB/s 100GB LT0 11 MB/s 100GB DTF2 20MB/s 200GB
The problem is never usually the tape drives, but the network interface of the backup server, if you want to get more than 1 server per drive. SAN attach backups are still a little immature. Gbit ethernet is OK for now.
The future of backups IMO is to stage to big piles of cheap disk and then stream the data from there to tape. You don't need very high performance disk to do this, just good sequential transfer rates.
Mean Time BEFORE (or BETWEEN it means the same) Failure
i.e. not if but when it fails
Sure IBM has screwed up some batches, but if you haven't backed up to tape or CDR/whatever and you're complaining then nobody is impressed.
Our production standard for data is 1. Mirror RAID-1. either in software or hardware, but you must do it. hot swap of course make this a lot less painful. 2. Backup daily off machine and offsite. You can use a replicator technology, now they are becoming so much better.
Anything less has to be a little foolish. Downtime is money remember.
but disk GUIDs are needed in some form
on
CPRM Smokescreen
·
· Score: 1
I think people are thinking a little small on this issue. If you've ever had to try setting up any reasonable Fibre Channel SAN you will quickly recognise that a GUID for a disk is a good thing.
All the zones and WWNNs etc really start to confuse when you have 00's of them.
The problem comes when the high spec. Enterprise devices filter in to the consumer channel which always happens eventually. So if you want to complain then go exert some pressure on the vendors from that angle, not as a consumer.
What I am wondering is whether anyone has done a similar distro for PCs? What would be ideal for a small office is to take a old PC and make it into a headless 'qube-a-like'. All the tools exist, it really should just be matter of glueing a good web interface to the front.
I disagree with this logic. Sure paper lasts longer than magnetic media, but until they invent the matterfax(tm), digital is a safer bet in the long run. Digital copies are easy.
What I'd rather see is an immediate start to the submission of all books to the libary in a digital form as well as the hardcopy. The cost of this should be bourne by the publisher. After all don't the vast majority already use DTP? The cost to each publishing house is a copy of Adobe Acrobat or similar.
Offline storage for these books is going to get huge though. That is a problem and more than likely terrifies the LOC.
Top drawer comment. I applaud. I suggest that the irc admins do further than a rather weak ban. I'd say you have to teach the lesson 1st time or it's never going to get learnt. Consider the situation if IRCnet shuts down, the kiddies shunt elsewhere and then ech net shuts down until all is quiet. Then the IRC admins could relly get people's attention at this point. It's like standing dead silent waiting to say something - you do get people's attention more quickly than you'd expect. The network may be able to restart after that or a self enforced silent period. For all the genuine users this should not be a hardship. A move to a webchat room should be acceptable for the short term. As this really is a free and now fairly unreliable service due to the DDoS, anyone who is betting their business on it is foolish in the extreme. Who would loose the most? I think the kiddies would.
Slashdot Mirror
This is turning into a typical /. discussion. Guys who haven't had experience verus those who have. For those who want to use disk backups: Try that when you are backing up 3TB+ a day. It's all a question of scale.
.PSTs are evil) refreshes come along at 18 or 24 month intervals. This coincides nicely with the doubling in capacity that tape drive manufacturers produce.
Tape won't die becuase it's alredy been used for 30+ years and there is a load of data out there that needs to be kept. There is a significant market just keeping this alive. Although the tape manufactureers would love us to do it, it is very rarely viable to pull back and refresh all the media you generate. Net result is that large backup and achival systems, once in are there for keeps. This somewhat explains why initial capital costs are so high, as refreshes for SMEs are rare. In orgs that have rampant out of control data growth ( this is most believe me - sub rant:
The real killer though is not the hardware but the software, keeping old backup systems alive to restore your old media is a nightmare. Anyone tried to restore an Novell system recently?
You laugh but the Russians did it first using real dogs in the battle for Stalingrad.
t ory21/
They trained dogs with high explosive strapped to them to go an look for food under German tanks.
Linkage: http://community-2.webtv.net/Hahn-50thAP-K9/K9His
Just you wait until a user calls you up and says their "UN9" box has stopped working!
It happens...
This is from practical experience - I'm a backup admin and I test backup technologies for million dollar installs.
AIT-2 Native 6MB/s Real 9MB/s
AIT-3 Native 12MB/s Real 20MB/s
LT0 Native 15MB/s Real 23Mb/s
LTO2 looks like boing a big win at a native 30Mb/s but the issues with it will be that it will be very tricky to get it to stream effectively at 50Mb/s. You gonna need a fast server to keep it supplied at that rate.
Sounds obivious doesn't it, but in reality space is an issue. There are no jukebox technologies that can store DVDs as densly as tapes, when you measure in TB/sq ft. You'd have to build a big ass robot for DVDs - I mean HUGE.
There is also the issue of speed. DVD etc just cannot match a LTO drive going at 25Mb/s full whack (and that's a medium speed these days)
Point in case - I work in the finance industry and run their backups (5TB/day). I could fill a DVD-R in under a minute and I'd need 000's per week. I'd spend my entire life just unwrapping the buggers.
However we are all looking forward to a holgraphic disc technology, and have been for the last 2 years. No signs yet...
Exactly. MTBF stands for Mean time BEFORE failure not BETWEEN. This is because a failure in these components can usually only happen once and it is usually fatal.
Tape media technology is having a hard time keeping up. We use lots of AIT-2 drives (40+) in a big silo to keep up with data growth. Tape technology follows Moore's Law but at a slower rate. It seems to double every 18 months in capacity. There's no way it can keep up with this level of growth in disk capacities.
Some tape transfer rates:
DLT8k 5MB/s 35GB
AIT-2 6MB/s 50GB
AIT-3 12MB/s 100GB
LT0 11 MB/s 100GB
DTF2 20MB/s 200GB
The problem is never usually the tape drives, but the network interface of the backup server, if you want to get more than 1 server per drive. SAN attach backups are still a little immature. Gbit ethernet is OK for now.
The future of backups IMO is to stage to big piles of cheap disk and then stream the data from there to tape. You don't need very high performance disk to do this, just good sequential transfer rates.
Mean
Time
BEFORE (or BETWEEN it means the same)
Failure
i.e. not if but when it fails
Sure IBM has screwed up some batches, but if you haven't backed up to tape or CDR/whatever and you're complaining then nobody is impressed.
Our production standard for data is
1. Mirror RAID-1. either in software or hardware, but you must do it. hot swap of course make this a lot less painful.
2. Backup daily off machine and offsite. You can use a replicator technology, now they are becoming so much better.
Anything less has to be a little foolish. Downtime is money remember.
I think people are thinking a little small on this issue. If you've ever had to try setting up any reasonable Fibre Channel SAN you will quickly recognise that a GUID for a disk is a good thing.
All the zones and WWNNs etc really start to confuse when you have 00's of them.
The problem comes when the high spec. Enterprise devices filter in to the consumer channel which always happens eventually. So if you want to complain then go exert some pressure on the vendors from that angle, not as a consumer.
bof.
Doh! not dreaming just asleep at the wheel. I really must read then post. Appologies. Comment withdrawn.
See:
http://www.netmax.com/
http://www.syrinex.com/
http://www.e-smith.net/
Pretty toy but way too pricey for my tastes.
What I am wondering is whether anyone has done a similar distro for PCs? What would be ideal for a small office is to take a old PC and make it into a headless 'qube-a-like'. All the tools exist, it really should just be matter of glueing a good web interface to the front.
A single CD SOHO soln? Friday dreaming no doubt.
I disagree with this logic. Sure paper lasts longer than magnetic media, but until they invent the matterfax(tm), digital is a safer bet in the long run. Digital copies are easy.
What I'd rather see is an immediate start to the submission of all books to the libary in a digital form as well as the hardcopy. The cost of this should be bourne by the publisher. After all don't the vast majority already use DTP? The cost to each publishing house is a copy of Adobe Acrobat or similar.
Offline storage for these books is going to get huge though. That is a problem and more than likely terrifies the LOC.
Bof.
ooops - Sorry. Silly formatting.
Top drawer comment. I applaud. I suggest that the irc admins do further than a rather weak ban. I'd say you have to teach the lesson 1st time or it's never going to get learnt. Consider the situation if IRCnet shuts down, the kiddies shunt elsewhere and then ech net shuts down until all is quiet. Then the IRC admins could relly get people's attention at this point. It's like standing dead silent waiting to say something - you do get people's attention more quickly than you'd expect. The network may be able to restart after that or a self enforced silent period. For all the genuine users this should not be a hardship. A move to a webchat room should be acceptable for the short term. As this really is a free and now fairly unreliable service due to the DDoS, anyone who is betting their business on it is foolish in the extreme. Who would loose the most? I think the kiddies would.