+ Post New Thread
Page 4 of 5 FirstFirst 12345 LastLast
Results 46 to 60 of 75
Hardware Thread, Why Tape Backups are Still Necessary in Technical; Originally Posted by Arthur If you are backing data up to enterprise-class SSDs, you might not want to leave them ...
  1. #46

    localzuk's Avatar
    Join Date
    Dec 2006
    Location
    Minehead
    Posts
    17,879
    Thank Post
    518
    Thanked 2,486 Times in 1,928 Posts
    Blog Entries
    24
    Rep Power
    838
    Quote Originally Posted by Arthur View Post
    If you are backing data up to enterprise-class SSDs, you might not want to leave them powered off for too long.

    What kind of temperatures are those for any enterprise setting?! My server SSDs are currently reporting in 25 degrees C, powered on! The CPU is only at 35 degrees, powered on...

    55 degrees screams "bad cooling" to me, as it is only 5 degrees off the upper temp limit for HP enterprise SSDs too.

    Those appear to be "requirements" of some form of standard anyway, not what companies actually sell...
    Last edited by localzuk; 13th May 2014 at 08:48 AM.

  2. #47
    zag
    zag is offline
    zag's Avatar
    Join Date
    Mar 2007
    Posts
    3,829
    Thank Post
    918
    Thanked 422 Times in 355 Posts
    Blog Entries
    12
    Rep Power
    88
    We use the 1tb Samsung SSD's here which can be bought for under £500 quid.

    No problems with data retention so far.

  3. #48

    Join Date
    Oct 2005
    Location
    hey hey hey, stay outta my shed. STAY OUT OF MY SHED.
    Posts
    1,047
    Thank Post
    246
    Thanked 202 Times in 156 Posts
    Rep Power
    109
    Quote Originally Posted by zag View Post
    We use the 1tb Samsung SSD's here which can be bought for under £500 quid.

    No problems with data retention so far.
    So are you saying that you don't leave them powered off for very long, or that you know better than the manufacturers themselves when it comes to data retention?

    Because if you're exceeding their guidelines then you're betting against the casino on a loaded wheel, and so far you've just been lucky.

  4. #49

    localzuk's Avatar
    Join Date
    Dec 2006
    Location
    Minehead
    Posts
    17,879
    Thank Post
    518
    Thanked 2,486 Times in 1,928 Posts
    Blog Entries
    24
    Rep Power
    838
    Quote Originally Posted by Roberto View Post
    So are you saying that you don't leave them powered off for very long, or that you know better than the manufacturers themselves when it comes to data retention?

    Because if you're exceeding their guidelines then you're betting against the casino on a loaded wheel, and so far you've just been lucky.
    Thing is, companies selling SSDs don't appear to be telling people that. They stick a 5 year warranty on them and that's it. I've never seen those sort of data retention figures, and I've bought hundreds of SSDs so far!

  5. #50

    Join Date
    Oct 2005
    Location
    hey hey hey, stay outta my shed. STAY OUT OF MY SHED.
    Posts
    1,047
    Thank Post
    246
    Thanked 202 Times in 156 Posts
    Rep Power
    109
    Quote Originally Posted by localzuk View Post
    Thing is, companies selling SSDs don't appear to be telling people that. They stick a 5 year warranty on them and that's it. I've never seen those sort of data retention figures, and I've bought hundreds of SSDs so far!
    Even if we assume the 5-yr warranty is golden that doesn't help if you have 7-yr retention requirements on some items of data (which we do).

    Backups are surprisingly difficult, when you dig into that level of detail.

  6. #51

    localzuk's Avatar
    Join Date
    Dec 2006
    Location
    Minehead
    Posts
    17,879
    Thank Post
    518
    Thanked 2,486 Times in 1,928 Posts
    Blog Entries
    24
    Rep Power
    838
    Quote Originally Posted by Roberto View Post
    Even if we assume the 5-yr warranty is golden that doesn't help if you have 7-yr retention requirements on some items of data (which we do).

    Backups are surprisingly difficult, when you dig into that level of detail.
    As @tmcd35 said, its about individual needs though. We don't have a 7 year retention period. Many schools don't.

  7. #52


    Join Date
    Feb 2007
    Location
    51.403651, -0.515458
    Posts
    9,065
    Thank Post
    232
    Thanked 2,717 Times in 2,005 Posts
    Rep Power
    795
    Quote Originally Posted by localzuk View Post
    Those appear to be "requirements" of some form of standard anyway, not what companies actually sell...
    Manufacturers can exceed the standards set by JEDEC to a certain degree (like they do for RAM), although it would be foolish to dismiss it completely.

    If you read some of AnandTech's enterprise SSD reviews, the issue of data retention gets mentioned fairly frequently.

    Btw, the image I posted above is linked to a PDF that contains more details.
    Last edited by Arthur; 13th May 2014 at 09:31 AM.

  8. #53

    seawolf's Avatar
    Join Date
    Jan 2010
    Posts
    969
    Thank Post
    12
    Thanked 285 Times in 217 Posts
    Blog Entries
    1
    Rep Power
    175
    Quote Originally Posted by tmcd35 View Post
    Very little data in a school environment is actually that mission critical.
    Really? What about the school administration or finance systems? Most countries have a 7 year data retention requirement and I'm pretty sure not being able to successfully restore these systems would probably result in the loss of a techs job or his managers job. I would certainly fire them in such an instance. I would fire myself for that. And go ahead and lose the entire Moodle system (or your LMS of choice) and see if the teachers wouldn't want to have your head on a (disk) platter.

    No, I think schools have some pretty critical data - to them.

    You post assuming yours is the only right way. In truth, as with many solutions, the answer a little less black and white. I find tape in the capacities required to be costly, slow, and a pain to manage. I can see why using it is necerssary when other more user friendly alternatives exist.

    The most important thing to note about our data center - we're not Google and don't have Google's back requirements.
    There are three very important things that at least one copy of your backups must achieve or they do not cut the mustard, whether you are Google or not:

    You must have at least one backup that is off-line (cloud backups don't qualify) and off-site (another building doesn't qualify), and they must be archival quality storage to ensure the backups are usable when needed (SATA drives aren't archival quality). There is currently only one reliable way to achieve all three of these that I am aware of at the present time, and that is with tape. You may choose not to use it, but it means your DR plan has a serious flaw in it. That's not just according to me, but to people with decades of backup/DR experience who know what can happen when this advice is ignored.

    Tape storage libraries are relatively cheap these days and cost per GB there isn't a cheaper storage medium than tape, so it's not some esoteric, costly backup method. Even VEEAM Backup now has a tape backup feature for archiving offline copies of VMs. Tape is dead, long live tape...

    FYI - the URE rate on LTO tape is 10^16 vs 10^15 for SAS/FC disks or 10^14 for SATA drives. Tape is an order of magnitude better than SAS/FC as an archival storage medium and even more so than a SATA drive.

  9. #54


    Join Date
    Feb 2007
    Location
    51.403651, -0.515458
    Posts
    9,065
    Thank Post
    232
    Thanked 2,717 Times in 2,005 Posts
    Rep Power
    795
    Quote Originally Posted by seawolf View Post
    the URE rate on LTO tape is 10^16 vs 10^15 for SAS/FC disks or 10^14 for SATA drives.
    I have some WD Re and WD Se SATA HDDs in a server at school that have URE's of 10^16 and 10^15 respectively.

    Seagate's Enterprise Capacity 3.5 SATA HDDs are 10^15 too.
    Last edited by Arthur; 13th May 2014 at 08:18 PM.

  10. #55

    localzuk's Avatar
    Join Date
    Dec 2006
    Location
    Minehead
    Posts
    17,879
    Thank Post
    518
    Thanked 2,486 Times in 1,928 Posts
    Blog Entries
    24
    Rep Power
    838
    Quote Originally Posted by seawolf View Post
    Really? What about the school administration or finance systems?
    That would seem to fit within the "very little" definition mentioned...

    Most countries have a 7 year data retention requirement and I'm pretty sure not being able to successfully restore these systems would probably result in the loss of a techs job or his managers job. I would certainly fire them in such an instance.
    You'd fire someone for working with the environment they've being able to build with the requirements placed on them and with the budget available to them? Pretty sure most places have a disaster recovery policy that is signed off by the head and governors...

    You must have at least one backup that is off-line (cloud backups don't qualify) and off-site (another building doesn't qualify), and they must be archival quality storage to ensure the backups are usable when needed (SATA drives aren't archival quality).
    You've just specified 3 absolutes which have only one solution, without providing reasoning. Why offline? Pretty sure cloud backup providers who have invested millions into their infrastructures would disagree with you.

    Any not a different building? What's different about 2 buildings 1 km apart on the same site compared to 2 buildings 1 km apart but on different sites? Is there a specific distance we should work with? To put it bluntly, if something could take out all like the buildings on our site in one go, we'd have much larger things to worry about than losing data...

    Why aren't disks, or SSDs acceptable as disaster recovery mediums? Surely that depends on the earlier mentioned data retention policy? As specified earlier, in our case, that is 1 year of data backup in electronic form...

    To get to the crux of the matter, you are still trying to force your organisation's rules and requirements on different organisations, making out that everyone else is wrong without providing any actual real justification for your view.
    Last edited by localzuk; 13th May 2014 at 08:50 PM. Reason: spelling

  11. Thanks to localzuk from:

    sparkeh (13th May 2014)

  12. #56

    mac_shinobi's Avatar
    Join Date
    Aug 2005
    Posts
    9,807
    Thank Post
    3,320
    Thanked 1,056 Times in 977 Posts
    Rep Power
    365

    Why Tape Backups are Still Necessary

    Quote Originally Posted by mac_shinobi View Post
    SANDisk Enterprise SSD ( 4tb to start with ) and there is a road map for 8tb and 16tb Enterprise SSD's, not sure it relates to backups other than large capacities ??

    It's here, the 4TB FLASH drive: SanDisk rips sheet from the Optimus MAX

    Does that mean you will have to use the larger tapes such as per this article :

    http://www.itpro.co.uk/storage/22183...-185tb-of-data
    I was referring to the size of SSD's increasing in comparison to SAS drives and that SSD's run cooler and possibly faster etc and with the size of SSD's increasing will most likely need larger capacity backup tapes in order to backup that amount of data ( system state plus user data etc ) ???

  13. #57

    seawolf's Avatar
    Join Date
    Jan 2010
    Posts
    969
    Thank Post
    12
    Thanked 285 Times in 217 Posts
    Blog Entries
    1
    Rep Power
    175
    Quote Originally Posted by Arthur View Post
    I have some WD Re and WD Se SATA HDDs in a server at school that have URE's of 10^16 and 10^15 respectively.
    WD's URE specs for their SATA drives are a bit of a dodge because they state <10 in 10^16 - which means less than ten failures in 10^16. The common benchmark for 10^16 is <1 failure in 10^16. The marketing team at work here. The WD Re and Se drives are very good though. Just not quite that good.

    Seagate's Enterprise Capacity 3.5 SATA HDDs are 10^15 too.
    Yes, some of the true enterprise grade SATA drives are rated at 10^15. I doubt that is what most people who've commented on this post are using with their NAS arrays or archival backups as a enterprise grade drive such as the WD Re or Seagate enterprise drive is generally 2x more expensive than a standard drive and much more expensive than an equivalent size tape drive.

  14. #58

    seawolf's Avatar
    Join Date
    Jan 2010
    Posts
    969
    Thank Post
    12
    Thanked 285 Times in 217 Posts
    Blog Entries
    1
    Rep Power
    175
    Quote Originally Posted by localzuk View Post
    That would seem to fit within the "very little" definition mentioned...
    If your school believes that their school administration (that being the system that contains all of the information about every student, staff member, and parents including reports, medical information, emergency contact info, discipline history, etc) and finance system is worth "very little" then sure, I guess you're correct. Or, if you deem the importance of data by the physical size of it whereas a school admin database and finance system can be small in size (but large in importance) - then sure, I guess you're correct again.


    You'd fire someone for working with the environment they've being able to build with the requirements placed on them and with the budget available to them? Pretty sure most places have a disaster recovery policy that is signed off by the head and governors...
    If they had requested the equipment (hardware/software) needed to provide a backup solution that would provide a reliable DR solution and they were denied the funds and made it clear to management in writing that they were placing the organisations critical data at risk and that was signed off by the school management - then no I would not fire them. In that case it is the school management that should get the axe for failing to properly manage risk. If they did not either state the need for such a DR solution, request money for it, and inform the management of what could occur if they decided to forgo these recommendations - then yes I would.

    You've just specified 3 absolutes which have only one solution, without providing reasoning. Why offline? Pretty sure cloud backup providers who have invested millions into their infrastructures would disagree with you.

    Any not a different building? What's different about 2 buildings 1 km apart on the same site compared to 2 buildings 1 km apart but on different sites? Is there a specific distance we should work with? To put it bluntly, if something could take out all like the buildings on our site in one go, we'd have much larger things to worry about than losing data...

    Why aren't disks, or SSDs acceptable as disaster recovery mediums? Surely that depends on the earlier mentioned data retention policy? As specified earlier, in our case, that is 1 year of data backup in electronic form...

    To get to the crux of the matter, you are still trying to force your organisation's rules and requirements on different organisations, making out that everyone else is wrong without providing any actual real justification for your view.
    You've made assumptions about risk and the importance of your organisations data that I believe overlook risks, underestimate the importance of your data, and set the bar too low.

    Why offline? Because online backups can become corrupted or accidentally overwritten with garbage just as easily as on-site backups can. Especially if you are simply using secondary copy to the cloud backup service from your local backups (as most do) rather than taking separate backups directly to the cloud. If you are doing the latter and the cloud backup provider is backing up to tape or other archival storage media (e.g. you are outsourcing your offline backups) then cloud backups would meet the requirements. However, most people do not have a fast enough connection to a cloud backup service to make that leap or can afford it. It is certainly much more expensive than using tape to meet this objective.

    Why not a different building? Because the same disaster that strikes one can easily strike the other. If there is a local flood, cyclone/hurricane, tornado, or bushfire even having a building 1km away will do little to protect you. Another building will protect you from a local building fire, but not much more. And actually, the loss of data means that the school could not quickly bring temporary buildings or lease out another facility and bring their services back online quickly with no school admin data, finance system, or IT infrastructure (e..g ability to recover data and systems to new servers). So, a data loss can greatly hinder rapid resumption of services more than loss of facilities in many cases. Organisations without good contingency/DR plans including data recovery very frequently fail - that is go bankrupt, shutdown, fold up, expire, and all such descriptors in the near future. How is that for a reason.

    Hey, if you're management is cool with a one year data retention policy and is willing to take those risks then so be it. But, really is that the bar that we should be setting as IT professionals? Shouldn't we try to go above and beyond the bare minimum that our (sometimes ill-informed) management places on us? Managers who do not have an understanding of IT, backups/DR, etc. often have no idea what it is they are agreeing to or asking until they are facing the consequences of their bad decisions made out of ignorance (even when you try to explain it to them). It's my job to protect my organisation from that as best I can. Going the extra mile. That's what being a professional is all about.

  15. #59

    tmcd35's Avatar
    Join Date
    Jul 2005
    Location
    Norfolk
    Posts
    5,713
    Thank Post
    858
    Thanked 904 Times in 749 Posts
    Blog Entries
    9
    Rep Power
    330
    I appreciate the point you are driving at @seawolf and if money was no object, as ever, I'd probably agree. If money was no object I'd be running a mirrored fibre channel SAN with D2D2T backup in a remote building using an auto tape loader and probably add in some method of cloud based backup to boot. Alas money is very much an object and thus compremises have to be made.

    For bet or worse tape wouldn't work here as a backup medium. It's proven to be the most unreliable of backup medium. And, honestly, that is 99.999% down to the human element (we forget to change tapes). It's also slow and expensive, especially for the terrabytes of storage involved and the rotation strategey required.

    We use what works best for us based on careful analysis of the problem. That's an onsite NAS finally enough running virtual tape backups, and an offsite NAS taking a full termly site snapshot.

    Whats the absolute worse that can happen? A fire/tornado/flood manages to take out both the server room and the backup NAS, and to top is off 2 drives in a rarely used NAS go bad at exactly the same time. Possible? yes, but I'm sure the odds of winning the lottery are better. So what then? Well, there's no school so all bets are off anyway until some, even temporary, provision for buildings is made. Heads of Department may have to contact local schools to get hold schemes of work. The exams officer will have to phone round the various examboards to get special despentation for our students. Finance will probably be the worst hit having to worry about paying invoices and wages.

    It'd be bad, crap will definately hit the fan, but the school will get through it. Life will go on, the school will rebuild. There is nothing that mission critical about any schools data. We're not a multi-national corp that'll face going out of business or lossing millions of pounds.

    In all likelyhood, it's doubtful any one event will take out both buildings. If it does we lose upto a terms worth of data. Should we back up offsite more often? Probably, and I'm actively looking in to it. Do we need tape? No.

    As for keeping data only for a year. That's just the main rotation. Were needed we have archives that are backed up within that rotation designed to retain key data longer. There's honstly not that much of it. Most of our backups are of staff and students MP3 collections. It's a great service we offer the music industry
    Last edited by tmcd35; 14th May 2014 at 08:08 AM.

  16. 2 Thanks to tmcd35:

    elsiegee40 (14th May 2014), sparkeh (14th May 2014)

  17. #60

    localzuk's Avatar
    Join Date
    Dec 2006
    Location
    Minehead
    Posts
    17,879
    Thank Post
    518
    Thanked 2,486 Times in 1,928 Posts
    Blog Entries
    24
    Rep Power
    838
    Quote Originally Posted by seawolf View Post
    If your school believes that their school administration (that being the system that contains all of the information about every student, staff member, and parents including reports, medical information, emergency contact info, discipline history, etc) and finance system is worth "very little" then sure, I guess you're correct. Or, if you deem the importance of data by the physical size of it whereas a school admin database and finance system can be small in size (but large in importance) - then sure, I guess you're correct again.
    You seem to be being deliberately obtuse. The original comment was that schools have very little critical information. Within that "very little", I would place MIS data and finance data. "Little" is a quantitative word, so yes, it is a measure of physical size.

    If they had requested the equipment (hardware/software) needed to provide a backup solution that would provide a reliable DR solution and they were denied the funds and made it clear to management in writing that they were placing the organisations critical data at risk and that was signed off by the school management - then no I would not fire them. In that case it is the school management that should get the axe for failing to properly manage risk. If they did not either state the need for such a DR solution, request money for it, and inform the management of what could occur if they decided to forgo these recommendations - then yes I would.
    You pretty obviously don't work directly in a school. My last budget here for backup, for 5 years, was £2500. Including all hardware, software, training, etc... You aren't going to be able to fulfil both "quick day to day data recovery" and "tape based archival backup for offsite storage" with those funds.

    You've made assumptions about risk and the importance of your organisations data that I believe overlook risks, underestimate the importance of your data, and set the bar too low.
    No, I've done what we call a risk assessment. We've looked at the likelihood of various disaster scenarios. We've looked at cost benefit analyses. Our conclusion is for our D2D2D solution as it stands right now. The school up the road might well come to a different conclusion.

    Why offline? Because online backups can become corrupted or accidentally overwritten with garbage just as easily as on-site backups can. Especially if you are simply using secondary copy to the cloud backup service from your local backups (as most do) rather than taking separate backups directly to the cloud. If you are doing the latter and the cloud backup provider is backing up to tape or other archival storage media (e.g. you are outsourcing your offline backups) then cloud backups would meet the requirements. However, most people do not have a fast enough connection to a cloud backup service to make that leap or can afford it. It is certainly much more expensive than using tape to meet this objective.
    You're conflating technical implementations of technology with the technology itself. You're making assumptions that you really shouldn't. Cloud backup can be exactly the same as onsite backup. It is just in a remote location, and the whole point of it is that it scales and therefore benefits from economies of scale which small organisations cannot gain themselves. The actual way they do it is important, yes, and should be taken into consideration when choosing a provider. My last school had a cloud based backup solution. It was "cloud" because it was offsite and "black box" to the school. We installed a client on every server, and every night a differential backup was performed via a dedicated 100MBit fibre to the provider (who happened to be a local college with hundreds of TB of data to backup, and the fibre was used for other things during the day by them). Their solution backed up D2D2T*2 - so there were 2 identical sets of tapes produced of every backup, one which was left in the machine for restoration purposes, 1 which was removed to one of their other campuses...

    Why not a different building? Because the same disaster that strikes one can easily strike the other. If there is a local flood, cyclone/hurricane, tornado, or bushfire even having a building 1km away will do little to protect you. Another building will protect you from a local building fire, but not much more. And actually, the loss of data means that the school could not quickly bring temporary buildings or lease out another facility and bring their services back online quickly with no school admin data, finance system, or IT infrastructure (e..g ability to recover data and systems to new servers). So, a data loss can greatly hinder rapid resumption of services more than loss of facilities in many cases. Organisations without good contingency/DR plans including data recovery very frequently fail - that is go bankrupt, shutdown, fold up, expire, and all such descriptors in the near future. How is that for a reason.
    Likelihood of flood here? Low to medium, depending on type of flood. Sea based flood - sea-levels would have to raise by 20m to get all the buildings on the site. River? No nearby rivers, only a couple of tiny streams within 1 mile, and they've not flooded for decades, and when they did, they only flood within a few meters of the streams themselves. Water table rising? This is a real possibility in our server room, so we have flood and humidity detectors in the server room which message out to me as soon as there is a change, and I come in and check what's going on. Water tables don't rise quickly enough to be a threat, as there is 2 foot between the floor and servers. Cyclone/hurricane - we don't get these in this country. Tornado - possible, we do get light powered versions of these in the UK occasionally. They usually take a few tiles off rooftops, so wouldn't get to our server room due to its location. Bushfire - we don't get these.

    So, our risk assessment highlights fire as our main risk factor, and the likelihood of buildings at opposite ends of our site being lost are slim to none due to the spacing. Therefore, the outcome of our risk assessment is that our far building is a suitable location for a backup solution to be hosted. We then also have the removable hard disks as offsite.

    Hey, if you're management is cool with a one year data retention policy and is willing to take those risks then so be it. But, really is that the bar that we should be setting as IT professionals? Shouldn't we try to go above and beyond the bare minimum that our (sometimes ill-informed) management places on us? Managers who do not have an understanding of IT, backups/DR, etc. often have no idea what it is they are agreeing to or asking until they are facing the consequences of their bad decisions made out of ignorance (even when you try to explain it to them). It's my job to protect my organisation from that as best I can. Going the extra mile. That's what being a professional is all about.
    It is nothing to do with bar setting or being professional. It is to do with funding and time. It is to do with what we are actually capable of doing with what we've got! You'd scream out loud at our last backup solution, which was specced up for a 3 year life, and ended up running for 5 years, on equipment out of warranty... As there simply wasn't the money to do anything about it.

    Also, I want to touch on data retention, as the discussion seems to be confusing retention of actual data with backup retention. Our data, as backed up, is over a decade of data. We've got finance information from 2001 on it, for example. That is backed up with everything as expected. However, we only retain backups for 1 year.

  18. 4 Thanks to localzuk:

    AButters (14th May 2014), elsiegee40 (14th May 2014), sparkeh (14th May 2014), zag (14th May 2014)

SHARE:
+ Post New Thread
Page 4 of 5 FirstFirst 12345 LastLast

Similar Threads

  1. No tape backup support without 3rd-party software
    By SteveMC in forum Windows Server 2008
    Replies: 32
    Last Post: 9th January 2009, 04:05 PM
  2. Replies: 12
    Last Post: 19th September 2008, 07:45 AM
  3. Disk-to-Disk-to-Tape Backup
    By enjay in forum Hardware
    Replies: 30
    Last Post: 23rd November 2007, 03:21 PM
  4. RM Disk to Disk to Tape Backup Solution
    By Chris in forum General Chat
    Replies: 0
    Last Post: 2nd July 2007, 10:14 AM
  5. Disk to Disk to Tape backup. How do you do it?
    By trekmad in forum How do you do....it?
    Replies: 0
    Last Post: 30th May 2007, 07:49 PM

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •