To the cloud, to the CLOUD!! =P On Mon, Sep 14, 2015 at 6:13 PM, Dave G4UGM <dave.g4ugm at gmail.com> wrote:

On Mon, Sep 14, 2015 at 6:38 PM, drlegendre . <drlegendre at gmail.com> wrote: There is no cloud, just other people's computers.

On 9/15/2015 2:02 PM, Philip Pemberton wrote: I wish more things could be mirrored. Look at the loss of knowledge when the FREE web providers dropped Service. Ben.

I work for a VAR, and we've sold a fair number of IBM TS7650G ProtecTIER VTL's backed by Storewide V7000 disk. It basically emulates a bunch of LTO3 drives and cartridges, and does excellent deduplication when writing. If you have two, they'll replicate. It makes a great backup and DR solution and is a fairly straightforward shoehorn into existing tape-based workflows. Then again, there is still lots of tape out there, and as was pointed out earlier, the technology is always improving, and rapidly. Not everyone is looking to go disk-only, either. (yet) The more data you have to store, the more and more tape makes sense. -Ben ________________________________________ From: cctalk [cctalk-bounces at classiccmp.org] on behalf of Dave G4UGM [dave.g4ugm at gmail.com] Sent: Monday, September 14, 2015 4:13 PM To: 'General Discussion: On-Topic and Off-Topic Posts' Subject: RE: Is tape dead? I spoke to my former employer and they are ditching tape. They want off-site replication and if they have an off-site replica they see no need for tape..... Dave cartridges the so I 'daddy'"

Delete it on the master and have it faithfully deleted on the replica. On 15 September 2015 at 00:13, Dave G4UGM <dave.g4ugm at gmail.com> wrote: -- 4.4 > 5.4

You can't access the replica directly. Even if you could the replicas use incremental revisions, so you can go back to previous versions....

It depends on the purpose of the backup. There are two major possibilities (1) to recover from operator error (2) to recover from destruction of the system. The malware you're describing is really a variant of (2). For (1), on-system backups, or snapshots, or stuff like that are fine. The concern is a file that was accidentally deleted or clobbered, and another copy of that file on the same machine, or connected nearby, will serve. For (2) you need a backup that is isolated from whatever event clobbered your system. That event might be fire, earthquake, or malware. For all of those, distance and possibly isolation are what you need. If you care about fast recovery from both classes of issue, you will want to consider two backups, one for each kind of problem. The one for (1) can typically run much more frequently than the one for (2). paul

On Tue, 15 Sep 2015, Steven Hirsch wrote: AVG and McAfee. not necessarily the best stuff. Scan, while the malware was screwing stuff up in the background, did not find anything to complain about!

Sounds like Cryptolocker. Even the police paid the ransom to get their data back. At home I use an older Supermicro 15 bay 3u server as a NAS. I bought every disk in a pair, and usually don't buy it in advance. Even though it has a 24 port 3ware card, only the OS disk is mirrored, all the others are exported JBOD (single disk, no RAID) Then I manually run an rsync script here and there to copy the data from disk set A to B. Disk set A is the only one exported via NFS/SMB, so the only one the Windows and Mac systems can get to. Originally I was meaning to script it so the B disks would be spun down, but haven't gotten around to that yet.

On Tue, 15 Sep 2015, Mouse wrote: Google Chrome and Internet Explorer. I'm more inclined to suspect the latter, but I obviously did not have them configured the way that I thought that I did. Far too many guvmint agencies have far too much stuff "easily accessible" that shouldn't be.

On 09/15/2015 10:49 AM, Mouse wrote: I think a more important issue in backing up is "How many GENERATIONS to you keep around?" If you're just overwriting last month's backup, you could be propagating the effects of a malware or just plain error with no means of retrieval. My backups are currently done by connecting an external drive to a system, and booting with a live CD. Important stuff is also duplicated on several different machines--and when new technology obsoletes the old, carry the content forward on the newer medium. I back up my original work or valuable reference sources. No pictures or movies. When you consider how much *original* work anyone does during a lifetime, it's surprisingly little. Maybe that's changed today. I remember seeing a figure of 11 debugged lines of code per day per programmer as the average for a GSA programmer back in the 1980s. --------------- Related to the subject of backup devices, I've been seeing stupid-low prices on SSD using MLC flash. How reliable are these things? I'm still of the spinning rust persuasion, right now. --Chuck

I think the language is less important than what the code is doing. I can scribble out hundreds of lines a day when it's boilerplate or just a mechanical transcription of a well-burnt-in algorithm, but can easily drop down to the single-digit range when I'm struggling with a difficult problem. Nor do I have any reason to think I'm unusual in any respect here. Indeed, what is my lines-per-day figure when I spend all day struggling with a bug and end up fixing it by replacing four lines by two? Minus two lines per day? Note that the preceding paragraphs are entirely independent of what language I am (putatively) writing in. /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mouse at rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

On 09/15/2015 11:49 AM, Richard Loken wrote: Most likely COBOL. The GSA types were doing a lot of it back when I remember the number. Probably punched into cards. Could also have been RPG, or less likely FORTRAN. Back when I heard the statistic, I don't think that federal types had used C. Maybe JOVIAL or Autocoder... --Chuck

For many purposes, that's an important consideration, yes. There's something (small) I back up weekly for which I keep the most recent seven backups, the oldest backup in each of the most recent twelve months, and the oldest backup in any year. I'm considering something of the sort for my house backups - live replication to a backup host, with a once-a-week freeze of the replica, storing past replica drives on a scheme somewhat like the above. Pictures and movies can be original work - perhaps not for you, certainly mostly not for me (I have a few original pictures, but only a few), but I know graphic designers and photographers who have probably produced at least a gigabyte of original pictures each by now. And people into video production.... So am I. I don't trust multi-level for anything more than passing a thumb drive to a friend or the like (where failure carries a very low cost) and I mostly don't trust wear-leveling algorithms yet, so I have little use for flash, certainly not for backups. Besides, even stupid-low SSD prices are still, in my admittedly limited experience, well above spinning-rust prices. The advantages of flash (random access and physical ruggedness) aren't very important for most backup applications, so price (and burnt-in technology, to some extent) win out for my purposes. /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mouse at rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

On 09/15/2015 12:32 PM, ethan at 757.org wrote: I tend to think of pictures and movies as sui generis--they were perfectly well done in non-digital form, so I don't include them as "data" needing backup, but rather a special case of digital data masquerading as a simulacrum of analog information. Anyone remember the IBM "Photostore" setup at Lawrence Livermore? Enterprising programmers wrote tools to go through their files and "touch" them, lest they be "Photostored". In many cases, that meant "gone forever". -------------- In the case of spinning rust, what brand is most reliable? I've seen dreadful reports of DOA drives from Western Digital, fewer from Seagate, but I don't know about Hitachi, Samsung, etc. Up until now, I've confined my purchasing to 500GB drives on the hope that they're more reliable than the 3-5TB monsters. Is this a mistake? FWIW, my SA-4004 still works... --Chuck

Caveat: my experience is as a home "power user" In my experience it all depends: I have a set of Seagate "consumer grade" SATA drives that are now nearly 10 years old that have been spinning in a RAID 6 almost 24x7 w/o issues. These were bought when Seagate offered a FIVE year warranty. I've also bought a number of the same type of drives in the past five years but with only a three year warranty and the failure rate on those have been abysmal! Now days I buy "enterprise class drives" (have a RAID 6 w/ 4TB Hitachi drives and RAID 5 setups w/ Seagate ES and Cheetah drives). They cost a pretty penny but the peace of mind is worth it. The also have better warranty service. I've never bothered w/ WD drives so I can't comment on those. If I could find a cheap tape backup solution I would love it but anything capable of reasonably backing up 20TB is well out of the consumer pocket book. -Ali

On Tue, Sep 15, 2015 at 01:02:03PM -0700, Chuck Guzis wrote: Google for the Blackblaze reports. IIRC HGST and WD had the lowest failure rates, Seagate just plain sucked. IMHO, you want to buy at one generation below the current max capacity on the assumption that they ironed out the bugs on that one. Kind regards, Alex. -- "Opportunity is missed by most people because it is dressed in overalls and looks like work." -- Thomas A. Edison

On 9/24/2015 6:16 PM, Al Kossow wrote: I wonder what is happening in the clouds? Ben.

On 9/24/15 5:45 PM, Al Kossow wrote: Thinking about it, Jason/IA could say something wrt drive reliability. They are running similar chassis to us (24/36 drive Supermicros) just a LOT more of them.

On 9/24/15 4:30 PM, Alexander Schreiber wrote: Backblaze

On Thu, 24 Sep 2015, Chuck Guzis wrote: I've got a number of WD Blue 1TB drives in use. They're very good drives. You can find them for around $50 each too. g. -- Proud owner of F-15C 80-0007 http://www.f15sim.com - The only one of its kind. http://www.diy-cockpits.org/coll - Go Collimated or Go Home. Some people collect things for a hobby. Geeks collect hobbies. ScarletDME - The red hot Data Management Environment A Multi-Value database for the masses, not the classes. http://scarlet.deltasoft.com - Get it _today_!

The WDC REs are not bad drives at all but IMO Hitachi Ultrastar is the best line going right now. I have been working with them for some time from 0.5T through 3T under very high duty cycle and they are fairly bulletproof. The REs will do the work, but I have seen higher failure rates on them right out of the box and higher failure rates on them in the longer term (~3 years) versus the Hitachi. I am hesitant to trust Seagate for large scale enterprise use though I think they are fine to use at home, in lighter duty cycle applications or in USB enclosures, etc. The distinction you point out is very important particularly when selecting drives to use in constructing a RAID; it's critical to avoid those drives that attempt to spin down or sleep when idle; this confuses the heck out of RAID and will cause the admin (we, the builder) no end of misery... usually it's easy to identify these drives because they are marketed as "Green" or "energy saver" but for some of the midrange product lines ... "prosumer" ... sometimes you have to dig a little to get to the facts. I think these Green drives are the worst thing on the market since the old Quantum Bigfoot; I wouldn't recommend them to anyone. I suggest starting with the keyword "enterprise SATA" and going from there ... these drives are certified for array use 24/7 and ship without any "power saving" nonsense straight from the OEM ... the premium over a normal consumer SATA disk is not really too awful. These generally use identical or very closely related HBAs to the premium SAS disks but just with different SATA-only formatter board. No real need to go SAS unless you require multipath. Best, Sean On Fri, Sep 25, 2015 at 4:56 PM, Alexander Schreiber <als at thangorodrim.ch> wrote:

On Fri, Sep 25, 2015 at 06:55:02PM -0400, Sean Caron wrote: There's no need to swear :-) mcl

On Tue, Sep 15, 2015 at 03:32:27PM -0400, ethan at 757.org wrote: It's not simply enough to cut multiple tapes (backups) of anything you care about --- you must periodically *read* and *verify* those tapes (backups), so that you can re-copy from a known good other copy if it is damaged. -- Thomas L. Kula | kula at tproa.net | http://kula.tproa.net/

On Tue, Sep 15, 2015 at 03:32:27PM -0400, ethan at 757.org wrote: Well, I have been mulling buying a BD-writer for a while (much less now since my backups go to LTO3), thanks for the warning - looks like BD-R is another waste of time & money. Depends very much on the tape technology. Anything based on VHS tape was a complete waste of time and money. Anything based on helical scan cannot be trusted (at a previous job, we would boot/install machines from DDS3 tapes and I kept 3-4 copies of each install tape. After two years I had a nice big stack of failed tapes). Linear scan (e.g. DLT, LTO) with read-after-write is the only tape technology worth trusting. Even so, for data you care about, keep multiple copies and multiple generations around. Kind regards, Alex. -- "Opportunity is missed by most people because it is dressed in overalls and looks like work." -- Thomas A. Edison

On Tue, Sep 15, 2015 at 11:36:57AM -0700, Chuck Guzis wrote: Yeah, that is one of the canonical mistakes. I've seen tape backup setups that had _the_ backup tape, which was overwritten for every backup run. They might as well not have bothered ... Well, the backup for the digital camera archive currently _fills_ two LTO3 tapes (with the wife and me both keeping DSLRs busy, that is not a surprise). Don't trust flash for backups. With rotating rust, backups to cold storage (i.e. connect drive, write data, disconnect, store it) works pretty well. With flash ... well. The deeper one digs into how flash storage is done these days, the less one tends to trust it. Do you really trust what is effectively down to a handful of electrons trapped in a tiny cell with a very thin insulator hanging around forever? Yeah, me neither. There is a reason my workstation has a hybrid RAID1 for the rootfs: spinning rust + flash drive, with the spinning rust set to 'write mostly'. Gets me the speed of flash and the more proven long term stability of spinning rust. Kind regards, Alex. -- "Opportunity is missed by most people because it is dressed in overalls and looks like work." -- Thomas A. Edison

On Tue, 15 Sep 2015, ethan at 757.org wrote: Detection while running is no big deal. But, after the damage is done, recovery is extremely difficult. Not worth asking John Gilmore for help building Deep Crack. Would anybody really trust the miscreants to provide the key after the ransom is paid?

Depends on why you have backups - that is, what kind of trouble the backups are intended to defend against. I've identified three basic reasons to keep backups: (1) "Oops." This includes things like running rm in the wrong directory by mistake. (2) "History." This is things like "okay, this was working last May, what was in the config file then?" and "when was it we installed that?". (3) "Oh no." This includes things like disk drives dying. If anyone has a fourth reason to suggest, I'd be interested. (Note that getting cracked is basically a cross of (1) and (2) - and, indeed, (1) can be viewed as a very-short-term form of (2).) The thing is, for (3) you _want_ live replication, because you want your replica to be as up-to-date as possible when a drive fails - but for (1) and (2) live replication goes against the whole point (though depending on how it's done it can be tool in the service of such backups; historical backups can be done by freezing live replication and saving offline copies of the replicas). Thus, the answer to "are live replicas good backups?" is actually a definite "maybe, depending on your threat model". /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mouse at rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

Yes. Offsite backups are an important piece for many threat models. But that's mostly orthogonal to whether they're designed to defend against my (1), (2), or (3); any of them can be done offsite. /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mouse at rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B

I think there are other pressures, but I am sure there is an element of this. The other one is that they are under pressure to save money and often there can be short term "savings".. .. and once there is no internal system to compare with then senior management can't dispute the costs That's why you need to be able to go back generations as well. I also see that the rise in "ransom ware" in a corperate environment.. Because if it is done properly it looks like lots of individual companies and hacking them all can be a problem. I also believe that cloud companies are less likely to take short cuts with security.... what I still needed. I have a selection of tapes but DLT2 is my biggest... Dave > -----Original Message----- > From: Dave G4UGM > Sent: Monday, September 14, 2015 7:13 PM > To: 'General Discussion: On-Topic and Off-Topic Posts' > Subject: RE: Is tape dead? > I spoke to my former employer and they are ditching tape. They want off- > site replication and if they have an off-site replica they see no need for > tape..... > Dave > > -----Original Message----- > > From: cctalk [mailto:cctalk-bounces at classiccmp.org] On Behalf Of > > Richard Loken > > Sent: 14 September 2015 23:56 > > To: General Discussion: On-Topic and Off-Topic Posts > > <cctalk at classiccmp.org> > > Subject: Re: Is tape dead? > > > > The death of tape is greatly exagerated. > > > > My duties today included removing 50 tapes from our Storagetek 500 > > tape library and sending them offsite for safe storage. These are > > LTO-5 > cartridges > > that can hold up to 30 Tbyte of compressed data and it can be written > > on > the > > tape with astonishing speed. > > > > On the other hand, my employer is bored with 40Gbyte DLT cartriges > > because 40Gbyte is little more than a SSSD 5-1/4" floppy disk these > > days > so I > > now can have them for free. I have a DLT4000 backing up my toy > > OpenVMS box here at the office and another DLT4000 backing up my toy > > FreeBSD box at home. > > As for cartidges, I am overwelmed with media - including three > > unopened cartons of NIB DLT-IV cartridges. > > > > -- > > Richard Loken VE6BSV, Unix System Administrator : "Anybody can be a > > father > > Athabasca University : but you have to earn > > Athabasca, Alberta Canada : the title of > 'daddy'" > > ** richardlo at admin.athabascau.ca ** : - Lynn Johnston > --- > This email has been checked for viruses by Avast antivirus software. > https://www.avast.com/antivirus

On Mon, 14 Sep 2015 ethan at 757.org wrote: Well in this case they claim 15Tbyte uncompressed but 100Gbyte uncompressed is still a lot more than a DLT-IV will hold. -- Richard Loken VE6BSV, Unix System Administrator : "Anybody can be a father Athabasca University : but you have to earn Athabasca, Alberta Canada : the title of 'daddy'" ** richardlo at admin.athabascau.ca ** : - Lynn Johnston

On Mon, Sep 14, 2015 at 4:56 PM, Richard Loken <richardlo at admin.athabascau.ca> wrote: That's one heck of a compression ratio, given that the native (uncompressed) capacity of an LTO-5 cartridge is 1.5 TB. Usually the vendors claim 2.5:1 compression, but obviously the actual compression achieved is entirely dependent on the entropy of the data. I'd be really surprised if the tape drive can stream at a high enough rate to store 30 Tbyte; the native (uncompressed) streaming rate is 140 MB/s, so for 20:1 compression it would have to transfer 2800 MB/s, which is 22 Gbit/s, more than even SAS or Fiber Channel can handle. With LTO and similar tape drives, if you can't maintain streaming, the performance goes to near zero, and you beat the heck out of the tape.