Your first-tier storage is hard drive based, right? Tape is only a backup. How can hard disk not be fast enough?
We do massive data dumps on a regular basis, and I was typing quickly. I probably should have said, "For our needs, hard disks are extremely inconvenient and their throughput is too slow individually to suit.". Good catch.
$32 a TB never count compression, 8.5TB with the newer tapes/heads and max "feature" not sure one price as it's a sole vendor product I wont go near it unless it's the only option.
One vendor for the drives -- Oracle -- but many, many vendors for the tapes, including several aftermarket licensees.
For fast, gigantic storage at 250Mbyte/sec/drive or faster, at present the StorageTek T10K series sure seems to be the only option.
So 8.5TB raw for $160 for the same media = $18.83/TB. A bit less than three bucks a terabyte more. Compared to hard drive cost in 4TB drives of about $100/TB, that's still a hell of a deal!
4TB hard drive: around $400 5TB tape: around $160 8.5TB tape (same media as 5TB, newer drive): still about $160
Cost per terabyte of disk: about $100. Cost per terabyte of tape: about $19
I'm ignoring the cost of the tape drive, just like I'm ignoring the cost of the head(s) involved in NAS/SAN storage.
To fix your quote to be in line with reality:
Glacier is cold storage; the drives are only spinning when they are filled, when retrieving, and when scrubbing / consolidating. Just like tape but at least five times more expensive.
I'd love to see a Petabyte-Scale Tape Storage System that looked something like this...
You're thinking way, way, way too small. One of my tape silos a few feet away from me holds about 135 petabytes, and if we bought some expansion cabinets it would hold enormously more than that.
IMHO, a SL8500 silo with dozens of robots shuffling tapes around under the LED lights looks way cooler than a pair of spinning reels. Here's a small sample: http://www.youtube.com/watch?v=OpVnk_GeCaw
Tape is slow, expensive, proprietary and unreliable.
The only people who still use it are those who have to, or idiots with money to burn.
Fact check on the troll.
"Tape is slow". Absolutely false for throughput; true only for IOPS. A modern tape is much faster than a modern hard drive. That's the point of the article, and my personal experience as well. Random I/O to/from tape drives is incredibly slow, but no hard drive can touch a modern tape drive's throughput. It's the reason LHC uses it.
"Tape is expensive": True only in a non-ROI sense, therefore mostly false. You'll find a modern, large tape silo of equivalent capacity to a modern, large storage appliance usually works out much cheaper both in initial cost and cost over time if you intend to use the hardware for at least three to five years. That said, the cost of admission to the world of enterprise tape is pretty high; it's the ongoing costs that are much lower than hard drives.
"Tape is Proprietary": Both true and false. LTO is an open (licensable) standard, but the fastest/largest tape drives on the planet are typically proprietary right now, because being the fastest/largest causes more sales, and therefore funds innovation in faster/larger tape technology.
"The only people who still use it are those who have to...": False. There are many, many use cases for tape where it is not a requirement, but is just more convenient, reliable, faster, and less expensive than a hard-disk solution. I could list them, but, well, you're a troll and I don't want to type much more.
"The only people who still use it are... [those] with money to burn.": False. ROI is what drives most of our tape purchases, and we save an enormous amount of money by using tape in appropriate scenarios. Hard disks are appropriate for some use cases, tapes are mandatory or just a smart purchase in others.
There are 8.5TB uncompressed capacity tapes in enterprise use right now.
I work for the company that makes those 8.5TB uncompressed drives. Even our internal use is quite limited at present; demand vastly outstrips supply, and they are quite hard to get!
For the time being, we're still using the 5TB uncompressed/10TB compressed T10K "C" drives with T10K T2 tapes. Still the fastest, largest tapes on the market outside of their newer, younger brother, and worth every penny!
Tape never died in enterprise use. Its use as a desktop backup device, though, certainly is very limited now compared to the day when many new workstations shipped with a built-in tape drive to back up the disk...
But doing backups of company data on tape is a bad idea. A bad reader can ruin tape. Its also susceptible to strong EM events. Optical is the way to go.
Have you ever tried restoring data from a DVD silo that's been in continuous use for ten years? I have, from multiple silos using different media. The rate of corruption of optical media is TERRIBLE.
Optical media is useful for certain types of storage, but historical reliability rates are awful. Meanwhile, tape tends to find the errors at write time, with far fewer incidents of "write once, read never". Everybody thinks optical is great until they work with the damned stuff. Failures are rampant, and unlike tape they tend to happen silently and undetectably until you try to read the stuff some time hence.
For tape, EM worries are obviated by the use of decent modern tape containers; any EMP source sufficiently-strong to get through the shielding on a modern, shielded tape shipping container is likely to destroy the container itself as well (read: low-level nuclear explosion). Also, if you store your data with a good off-site storage company (Iron Mountain is a fine choice), they will store your data in shielded cages that won't let a cell phone signal leak out, much less much worry about EM destruction of your data.
If you value your data, use tape. Or replicate the data like crazy globally and distribute the electrical & maintenance cost to your customers. If you don't really care about the data, optical media is just fine.
Maybe there are some cases where tape is cheaper than drives. However, Amazon has stated that they do not use tape for their Glacier service, which probably stores more data than even the mighty LHC.
Math follows. You've been warned.
A typical storage array where I work has 192 3TB drives in it, more or less. We use SSDs in hybrid storage pools, but we'll ignore that for the time being as it doesn't meaningfully change the equation. Let's leave the hefty cost of the storage appliance out of this. Let's just look at electricity alone.
Each drive consumes about 8 watts, and must be spinning continuously in order to provide reasonable response times. That's 1,536 watts per rack, just to power the drives. Ignore the shelf power consumption, the heads for the NAS array, the PDU draw and loss... we're just talking very back-of-the-envelope stuff here.
Now let's ignore the cost of your tape silo, but I submit to back up half a petabyte requires a library somewhat cheaper than your NAS device above. Typical tapes hold 10TB apiece, and are written to once. Their power cost is largely ignorable; the tape library only consumes power for perhaps a small display and some internal LED lights, and significant wattage only while running the job, which we could take to tape over the course of about a week assuming sufficient drives. Let's assume worst-case for the tape drive to hard drive comparison -- that we're not using RAID or mirroring of any sort on the storage array -- and that we're actually backing up 576TB of data from that storage array. That means we require something like 58 tapes. Media cost is going to be something like 58 * $160/tape == $9,280 in media to back up that storage array.
Typical cost for electricity is 12 cents per kilowatt-hour. 1536 * 24 * 365 & $0.12 == $1,614.64 to run your storage array every year, just in drive wattage (and that's quite conservative; most good-quality 7200RPM to 15,000RPM drives run a watt or two higher than this).
So there's a highly-simplified breakdown for the cost of tape versus disk; the library pays for its media cost compared to disk in 6 years of usage. Is that worth it? That's a great question. For our needs, hard disk just can't keep up with the data rates we require, so it's a speed/throughput thing, not a cost thing. Tape seek time is horrible, so for any application requiring IOPS, hard disks win.
There are lots and lots of ways to look at this equation. I've priced it out on purchase orders dozens of times, and every time tape wins for archival needs. You just can't beat the flexibility it offers, particularly for disaster recovery and legal hold requirements. "Here's a FedEx package containing your encrypted backup tapes" is far more convenient and an easy sell than starting the conversation with, "First, let us install a storage appliance on your property and set up a WAN connection to our data center..."
There's a balance between cost per byte and initial acquisition cost. DVDs and BDs have extremely low acquisition costs, but relatively high cost per byte (you can get started with $100). Hard drives have higher costs, but lower cost per byte. Tape has the least cost per byte, but initial costs are quite high...
Nailed it. I've never seen it expressed quite that concisely. Great work! Think I'll be using this quote regularly from now on.
Current pricing is about $60 for 1.5TB ($120 for a 3tb drive) for consumer disk vs LTO5 $30 + 2k for a tape head (LTO6 is more expensive as the takes a 3x the cost).
$160 for a Sun T10000T2 tape, uncompressed capacity 5TB, typically holds a touch over 10TB in our usage (assuming you have something that can pump 250Mbyte/sec to the drive, which requires a bit of engineering to get right). That's $16/TB, and you can bring that down quite a bit if you buy in the "tens of thousands" kind of bulk we do.
The drives themselves are quite expensive -- it's the largest, fastest mass-produced tape drive on the planet right now -- but media costs are very reasonable.
One of the best things about tape is there's no such thing as "undetectable, uncorrectable" error rates anymore. Built-in checksum validation at write time has pretty much eliminated that; if the checksum fails, the packet is re-written. We're talking about tracks just a few microns away from one another with thousands of tracks per tape. Pretty amazing, really.
"Somewhere under that pile of rubble is a LVD SCSI port, and it might even have power"
This is why you have multiple sites, and redundant equipment at each site. They don't need to be archiving the same data at each site, but at the very least they should use the same media.
And anyone still using LVD SCSI for tape storage today is really behind the times. 16Gbit Fibre Channel, TYVM!
...you need to both refresh and verify you storage...
You came pretty close with the process, but for most businesses you're not quite there. Here are a few clarifications on the process.
Typically large companies (including those, like us, with stringent HIPAA requirements) take two simultaneous copies from the original source. We don't copy a copy if it can be avoided, and we have enough tape drives to do this.
We contract out with a local storage company to grab the tapes within a few days and store for the given retention period off-site. One copy usually remains on-site as well for long-term retention and rapid restoration. With plenty of capacity in the silo (tens of thousands of tapes in an Oracle/Sun SL8500), we are not terribly concerned about retention policies. If we get tight on space, we'll just expand the silo again.
The same data usually still exists as on-disk media marked read-only, available for the legal folks who insisted we archive it in the first place. Often it also exists at a second geographical location thousands of miles (at minimum) from the first, with its own backup tapes. Plus it exists on two tapes at each site, one near-line and one off-site. Given tape reliability, three layers of data protection is typically sufficient. If "legal hold" is involved, we also insist that the disk array be kept on a valid support contract to reduce the risk of failed disks in the storage appliance.
Retention policies dictate we keep around at least a few tape drives of every generation we've ever used which has tapes archived with our off-site storage facility. Even if they are not in the silo, they're in a storage closet waiting for us to bring them to life if needed up to twenty years later.
I do this kind of thing all the time. Feel free to ping me at my easily-figured-out email address (firstname@lastname.org) if I can answer additional questions for you.
long-life optical discs fail... [store] tape in a cool, dark place...
This, this, one-thousand times this. I've worked in data centers for a decade and a half, and seen innumerable optical media go bad within just a few years (typically about 3 years) even in DVD jukeboxes in climate-controlled environments. Meanwhile, we restore from fairly ancient tapes on a regular basis.
In reality, most companies don't store tapes longer than 7 years anyway; that's the upper limit of typical audit liability. The data on the tapes may be older than that, kept indefinitely on-disk, but most large companies have a fairly aggressive destruction/over-write schedule for data on tape older than 7 years.
It's very unlikely we'll need data off a tape 20 years from now, but kept in the right conditions -- like the bat-cave of a tape silo room housing tens of thousands of 10TB tapes a few feet away from me right now -- there's a really good chance the data will be readable. While we do have plenty of tape failures (hundreds per year), they are almost always caught at write-time by the verification head.
On a modern tape drive, you usually have several dozen "heads" on any given tape drive, and there will be two sets of them each with its own mechanism to align it with a precision of just a few microns. Pretty amazing, really; if you drop by the Denver, CO area some time, the Oracle/Sun building engineers there can often arrange a tour of our tape testing facilities if you sign a NDA and represent a potential sale. Anyway, the second mechanism will be engaged on the tape in order to read what the first just wrote and verify it before it passes the "successful write" confirmation back up the fibre channel chain. This way you can guarantee you don't get "write once, read never" media.
The drives do speed matching, but they have a minimum speed, below that they start shoe-shining.
Agreed. At my work we do parallel streams to multiple Sun T10000 T2 tapes (T10K "C" drives) at 250Mbyte/sec uncompressed (500 megabytes per second compressed, more or less, usually quite a bit more). If for some reason we push less than about 120mbytes/sec, the tape rewind times cause all kinds of issues.
We make the same kind of decision when choosing Sun T10000 "B" drives instead of "C" or the new "D" drives if the source cannot push data fast enough.
I've long laughed at articles saying tape is dead. For large-scale* backup, retention, transport, and legal hold problems, there simply is no other solution that scales reasonably well.
*My definition of "large-scale" for this specific context: hundreds of terabytes or more, much of it transported thousands of miles regularly. If you don't work with hundreds of terabytes and at least dozens of petabytes on a daily basis, you may suffer from optimistic delusions regarding disk storage capabilities, one which disk storage vendors are all too glad to reinforce, to the detriment of customers faced with half-baked solutions that cannot hope to meet their throughput requirements. Given "large-scale" data, there's no replacement for tape at present; everything else is a low-throughput also-ran, typically harboring enormous and unplanned complications. We're also heavy users of VTL, replication, cloning, S3-workalikes, and various disk technologies. Tape remains vital to large enterprise operations, and those predicting its imminent death have been the butt of jokes about marketing wonks for a decade and a half.
I just started playing EVE Online in February of this year after a long hiatus from all online gaming. It has a great community, and due to the way skill trees work and the variety of places to play in (hisec, lowsec, nullsec, wormhole) it can be as casual or as hard-core as you want it to be.
I enjoy the heck out of wormholes at present!
1. get a job in NYC for $200k. 2. convince them to let you telecommute (will not be easy). 3. move to Austin TX. 4. profit.
Some of my co-workers were hired in California and moved to Salt Lake City. Many paid cash for their home in this area due to selling their home in CA.
In a large corporation, the downside of such a decision -- even though I know you meant the above somewhat tongue-in-cheek! -- is that someone paid at the high end of the bell curve is frequently passed up for raises due to the cost-of-living differential, regardless of job performance. The short-term benefit can be magnificent, but the long-term downside of promotions without attendant raises due to already having a salary "above" what the management manual suggests for the position in that area catches up with you.
Mexico does not have the time-zone disadvantage of India, and recently (due to various political and monetary factors) is very cost-competitive. Our Guadalajara office is chock-full of brilliant, dedicated people -- just like our India office -- but getting things done with their help is considerably easier due to the better time zone alignment.
Given the choice of hiring otherwise-equal software engineers in India for $25K/year vs. Mexico for $25K/year, I'd go with Mexico every time due to the time zone differential. Travel costs are cheaper. Conversations with US-based co-workers occur more frequently. Training time is quicker due to quicker turnaround on questions and feedback. You can obviously work around issues working with teams on opposite schedules -- and we do, every day! -- but it will never be as trivial as simply having most of your workers share common core hours.
Like I said at the start of the conversation: location and risk/reward are two fundamental aspects of how much one gets paid. It affects how easy it is to land a good-paying job, as well. In today's market, one's longitude matters a whole lot more than one's latitude; it's a lot easier for me to work with my great team-mates in Mexico during my normal working day than to stay up past midnight to meet with equally-talented teammates in India, Singapore, or Sydney.
Great link! Thanks for that. It shows pretty clearly something I've long suspected: that Salt Lake City, Utah is a "bellweather" area, smack-on the national average for technology pay. The more prospective employees educate themselves about these statistics, the better enabled they are to determine the right career path and negotiate fair pay.
I'd change jobs if my job consumed my life. Here's a breakdwon of a typical healthy week for a happily married father:
168 hours in a week. 15 hours spent alone with your spouse. Dates, hanging out, get a sitter for the kids, go to the gym together, go shopping together, etc. Yeah, I'm serious, if you aren't getting 15 weeks alone with your spouse you're probably 15 hours spending time with your kid(s). This can include chores together, hanging out, whatever. If no kids or they've moved out, you get this time back! 40 hours a week working (maybe 50; 60+ weeks should be a rarity, not a regular thing!) 10 hours commuting. Less if you can swing it. 56 hours sleeping (yeah, as if, but budget it anyway). 12 hours eating, shaving, showering, using the restroom, getting dressed, doing laundry & other chores, etc.
What's that leave you with? 20 hours a week that you can decide what to do. If you're working 70 hours a week, you've cut into the time with spouse and/or kids and have ZERO time for yourself.
Work/life balance matters differently to different people. Employers routinely expecting heroic effort is abusive. Luckily, I live very close to work and can work from home a good deal; the quality-of-life improvement from a reduced commute makes up for a LOT...
Two factors to consider: Location and risk/reward ratio.
The "$200,000" figure -- as mentioned earlier in the discussion -- is extremely location-dependent. Let's say you live in Salt Lake City, UT and make $100,000 per year. I'm from the area, and $90K-$110K seems to be a very typical pay rate for sysadmins and programmers with 10 or more years of experience in their field, if you include health and other benefits in the W-2 (all the employers do!). Then let's cross-reference that with the Cost Of Living calculator at http://money.cnn.com/calculator/pf/cost-of-living/ . To enjoy an equivalent quality of life, here's what that guy needs to make (these are all area averages, not the specific cities referenced):
San Francisco, CA: $171,987 Los Angeles, CA: $140,380 Seattle, WA: $123,784 Raleigh, NC: $99,154 Austin, TX: $97,991 Washington, DC:$151,479 New York, NY (Manhattan): $231,289 New York, NY (Queens): $162,684
Advice for earning the big bucks? If you want to earn a high-end "salary", move to one of the technology hubs. Get an education in finance and high-end mathematics. Or get a security clearance, shine up your resume and skills, and plan to have no life outside of work due to travel abroad. Get some solid experience, get in on some high-profile projects, and job-hop when you're at the apex of your visibility in your company for much better pay.
If you don't want to live in Silicon Valley or New York, then be your own boss. Write your own products. Innovate in a field that lacks innovation. Develop software nobody has really thought of before. Monetize it and make a living. Cash out by selling the company to a bigger fish, and start working on your next idea. My company, for instance, buys small companies who create innovative software a LOT and integrates that software into its other offerings. And that strategy works very well for continually expanding the customer base and revenue. Sometimes it works out well for the little guy, sometimes not. It's a risk of running your own business.
Working for someone in a "typical" IT or software engineering field is typically not gonna make you rich. It can make you a good living with solid pay and far less stress than running your own business or working in high-pressure financial and security markets, though. And you're less likely to lose several fortunes on your way to making a fortune if someone else takes the risk, and you get paid the salary. Risk/Reward.
For me, quality of life matters a lot. I've been on both sides of the fence (trying to make it in business for myself vs. drawing a salary), and right now in my life it makes more sense to draw a reasonable salary doing a job that I love in a great environment working with quality people and making a positive difference without feeling that my job is my life. But I know I'm probably not going to strike it rich doing so... and I'm OK with that.
But man, how atheists go on and on and on and on and on.... about how stupid it is to believe in god...
I'm an atheist. Haven't ever told anybody that they are stupid to believe in their God. Your statement is disproven.
I'm really an "apatheist": a subset of atheism that simply believes it's a waste of time & resources to debate an unfalsifiable hypothesis. I don't care if the supernatural exists, and there is currently insufficient evidence to suggest that caring about the question matters in the slightest. There are a lot of atheists like me: we don't care what you believe in or don't believe in as long as you don't try to force us to believe what you do. Most of us are not the vocal minority you apparently assume all atheists to be.
Absolutely fair comments, thanks for the information that new tapes have a higher cost on a tape drive than used tapes. I should have said "millions upon millions of feet of tape", which would have been a correct statement. I stand corrected.
A drive wears down in a few months if it is constantly writing to new tapes.
That's what the cleaning tapes in your silo are for. The heads are typically good for millions upon millions of read/write cycles as long as they are kept clean. The motors driving the reels are typically of the brushless, multi-speed variety, capable of decades of reliable operation with quality bearings. Cleaning tapes, on the other hand, must be replaced regularly as they wear out.
I frequently see bizarre claims on Slashdot, but the claim that using new tapes "wears down" your tape drive is a new and strange one. Where on earth did you get this idea?
Unless there are legal reasons to keep 5-10 years of backups, or you are dealing in more then 3-5 TB of storage to be backed up, or taking things off-site daily via courier tape is just too expensive.
I like your summary of three important reasons for tape archive. I'll restate in different terms. 1. Mid-term to indefinite data retention. 2. Large quantities of data, where "large" is a value greater than a single hard drive can reasonably store. 3. Disaster recovery planning.
But there are more.
4. "Oops".
That's the category of this KDE git issue. Recovering from an "oops". People screw up. How do you recover? I'm a big fan of having multiple layers in that onion: online snapshots, near-line replicas, and off-line tape backups are a basic three-tiered framework for figuring out how to protect the data. I'm amazed as big as KDE is, they don't have storage/backup expertise helping them keep their data secure. Makes me think I may have found my next open-source niche to fill.
5. Reliability. Contrary to the "fragile, expensive" opinion above, tape failure rates are demonstrably lower than hard drive failure rates despite regular handling. Research left to the reader; hard drives fail at a rate about fifteen times higher than their rated MTBF, which was already considerably higher than tape. Data on tape is far more resilient than data on a hard drive.
6. Cost. If you have to store data long-term, consider tape. Administrative, electrical, power, cooling, and storage requirements are all cheaper.
That's what I can think of off the top of my head; I'm sure there are more reasons for tape to be a good choice. The reality for many people that want to store their data "in the cloud" also is this:
I back up your "cloud" storage onto tape drives. Your cloud storage is only as reliable as my ability to recover it from a disaster.
It all depends on the scale. If you're talking a small project with a small budget, I agree with you: tape backups are overkill, too expensive, and kind of pointless. Your average open-source project is usually just a few gigabytes at most. Use a snapshotting, journaling filesystem, always keep each version in at least three different places, create a retention policy that makes sense for you based on the needs of your project, and you're good.
And you're right. Today's modern tapes are good for about 4,000 read/write cycles. Even if you get the tapes at a substantial discount, a 5TB+ tape is expensive to destroy!
But when you are talking large enterprise data archiving needs, high-end hard drives do not compete with high-end tape drives in the slightest. And in today's risk-averse corporate climate, a reasonable disaster recovery strategy is a MUST, and providing multiple tiers of storage -- online, near-line, and off-line -- is attractive. 9/11 showed everybody how quickly DR plans can melt.
I could go into a lot of specific numbers talking about how a few modern tape drives in a modestly-sized tape silo outperform similarly-sized hard drives for near-line storage in just about every category except random seek time by several orders of magnitude, but I'll leave that as an exercise for the reader.:-)
Slashdot Mirror
We do massive data dumps on a regular basis, and I was typing quickly. I probably should have said, "For our needs, hard disks are extremely inconvenient and their throughput is too slow individually to suit.". Good catch.
One vendor for the drives -- Oracle -- but many, many vendors for the tapes, including several aftermarket licensees.
For fast, gigantic storage at 250Mbyte/sec/drive or faster, at present the StorageTek T10K series sure seems to be the only option.
So 8.5TB raw for $160 for the same media = $18.83/TB. A bit less than three bucks a terabyte more. Compared to hard drive cost in 4TB drives of about $100/TB, that's still a hell of a deal!
Late 2013 pricing.
4TB hard drive: around $400
5TB tape: around $160
8.5TB tape (same media as 5TB, newer drive): still about $160
Cost per terabyte of disk: about $100.
Cost per terabyte of tape: about $19
I'm ignoring the cost of the tape drive, just like I'm ignoring the cost of the head(s) involved in NAS/SAN storage.
To fix your quote to be in line with reality:
You're thinking way, way, way too small. One of my tape silos a few feet away from me holds about 135 petabytes, and if we bought some expansion cabinets it would hold enormously more than that.
IMHO, a SL8500 silo with dozens of robots shuffling tapes around under the LED lights looks way cooler than a pair of spinning reels. Here's a small sample: http://www.youtube.com/watch?v=OpVnk_GeCaw
Fact check on the troll.
"Tape is slow". Absolutely false for throughput; true only for IOPS. A modern tape is much faster than a modern hard drive. That's the point of the article, and my personal experience as well. Random I/O to/from tape drives is incredibly slow, but no hard drive can touch a modern tape drive's throughput. It's the reason LHC uses it.
"Tape is expensive": True only in a non-ROI sense, therefore mostly false. You'll find a modern, large tape silo of equivalent capacity to a modern, large storage appliance usually works out much cheaper both in initial cost and cost over time if you intend to use the hardware for at least three to five years. That said, the cost of admission to the world of enterprise tape is pretty high; it's the ongoing costs that are much lower than hard drives.
"Tape is Proprietary": Both true and false. LTO is an open (licensable) standard, but the fastest/largest tape drives on the planet are typically proprietary right now, because being the fastest/largest causes more sales, and therefore funds innovation in faster/larger tape technology.
"The only people who still use it are those who have to...": False. There are many, many use cases for tape where it is not a requirement, but is just more convenient, reliable, faster, and less expensive than a hard-disk solution. I could list them, but, well, you're a troll and I don't want to type much more.
"The only people who still use it are... [those] with money to burn.": False. ROI is what drives most of our tape purchases, and we save an enormous amount of money by using tape in appropriate scenarios. Hard disks are appropriate for some use cases, tapes are mandatory or just a smart purchase in others.
I work for the company that makes those 8.5TB uncompressed drives. Even our internal use is quite limited at present; demand vastly outstrips supply, and they are quite hard to get!
For the time being, we're still using the 5TB uncompressed/10TB compressed T10K "C" drives with T10K T2 tapes. Still the fastest, largest tapes on the market outside of their newer, younger brother, and worth every penny!
Tape never died in enterprise use. Its use as a desktop backup device, though, certainly is very limited now compared to the day when many new workstations shipped with a built-in tape drive to back up the disk...
Have you ever tried restoring data from a DVD silo that's been in continuous use for ten years? I have, from multiple silos using different media. The rate of corruption of optical media is TERRIBLE.
Optical media is useful for certain types of storage, but historical reliability rates are awful. Meanwhile, tape tends to find the errors at write time, with far fewer incidents of "write once, read never". Everybody thinks optical is great until they work with the damned stuff. Failures are rampant, and unlike tape they tend to happen silently and undetectably until you try to read the stuff some time hence.
For tape, EM worries are obviated by the use of decent modern tape containers; any EMP source sufficiently-strong to get through the shielding on a modern, shielded tape shipping container is likely to destroy the container itself as well (read: low-level nuclear explosion). Also, if you store your data with a good off-site storage company (Iron Mountain is a fine choice), they will store your data in shielded cages that won't let a cell phone signal leak out, much less much worry about EM destruction of your data.
If you value your data, use tape. Or replicate the data like crazy globally and distribute the electrical & maintenance cost to your customers. If you don't really care about the data, optical media is just fine.
Math follows. You've been warned.
A typical storage array where I work has 192 3TB drives in it, more or less. We use SSDs in hybrid storage pools, but we'll ignore that for the time being as it doesn't meaningfully change the equation. Let's leave the hefty cost of the storage appliance out of this. Let's just look at electricity alone.
Each drive consumes about 8 watts, and must be spinning continuously in order to provide reasonable response times. That's 1,536 watts per rack, just to power the drives. Ignore the shelf power consumption, the heads for the NAS array, the PDU draw and loss... we're just talking very back-of-the-envelope stuff here.
Now let's ignore the cost of your tape silo, but I submit to back up half a petabyte requires a library somewhat cheaper than your NAS device above. Typical tapes hold 10TB apiece, and are written to once. Their power cost is largely ignorable; the tape library only consumes power for perhaps a small display and some internal LED lights, and significant wattage only while running the job, which we could take to tape over the course of about a week assuming sufficient drives. Let's assume worst-case for the tape drive to hard drive comparison -- that we're not using RAID or mirroring of any sort on the storage array -- and that we're actually backing up 576TB of data from that storage array. That means we require something like 58 tapes. Media cost is going to be something like 58 * $160/tape == $9,280 in media to back up that storage array.
Typical cost for electricity is 12 cents per kilowatt-hour. 1536 * 24 * 365 & $0.12 == $1,614.64 to run your storage array every year, just in drive wattage (and that's quite conservative; most good-quality 7200RPM to 15,000RPM drives run a watt or two higher than this).
So there's a highly-simplified breakdown for the cost of tape versus disk; the library pays for its media cost compared to disk in 6 years of usage. Is that worth it? That's a great question. For our needs, hard disk just can't keep up with the data rates we require, so it's a speed/throughput thing, not a cost thing. Tape seek time is horrible, so for any application requiring IOPS, hard disks win.
There are lots and lots of ways to look at this equation. I've priced it out on purchase orders dozens of times, and every time tape wins for archival needs. You just can't beat the flexibility it offers, particularly for disaster recovery and legal hold requirements. "Here's a FedEx package containing your encrypted backup tapes" is far more convenient and an easy sell than starting the conversation with, "First, let us install a storage appliance on your property and set up a WAN connection to our data center..."
Nailed it. I've never seen it expressed quite that concisely. Great work! Think I'll be using this quote regularly from now on.
$160 for a Sun T10000T2 tape, uncompressed capacity 5TB, typically holds a touch over 10TB in our usage (assuming you have something that can pump 250Mbyte/sec to the drive, which requires a bit of engineering to get right). That's $16/TB, and you can bring that down quite a bit if you buy in the "tens of thousands" kind of bulk we do.
The drives themselves are quite expensive -- it's the largest, fastest mass-produced tape drive on the planet right now -- but media costs are very reasonable.
One of the best things about tape is there's no such thing as "undetectable, uncorrectable" error rates anymore. Built-in checksum validation at write time has pretty much eliminated that; if the checksum fails, the packet is re-written. We're talking about tracks just a few microns away from one another with thousands of tracks per tape. Pretty amazing, really.
This is why you have multiple sites, and redundant equipment at each site. They don't need to be archiving the same data at each site, but at the very least they should use the same media.
And anyone still using LVD SCSI for tape storage today is really behind the times. 16Gbit Fibre Channel, TYVM!
You came pretty close with the process, but for most businesses you're not quite there. Here are a few clarifications on the process.
I do this kind of thing all the time. Feel free to ping me at my easily-figured-out email address (firstname@lastname.org) if I can answer additional questions for you.
This, this, one-thousand times this. I've worked in data centers for a decade and a half, and seen innumerable optical media go bad within just a few years (typically about 3 years) even in DVD jukeboxes in climate-controlled environments. Meanwhile, we restore from fairly ancient tapes on a regular basis.
In reality, most companies don't store tapes longer than 7 years anyway; that's the upper limit of typical audit liability. The data on the tapes may be older than that, kept indefinitely on-disk, but most large companies have a fairly aggressive destruction/over-write schedule for data on tape older than 7 years.
It's very unlikely we'll need data off a tape 20 years from now, but kept in the right conditions -- like the bat-cave of a tape silo room housing tens of thousands of 10TB tapes a few feet away from me right now -- there's a really good chance the data will be readable. While we do have plenty of tape failures (hundreds per year), they are almost always caught at write-time by the verification head.
On a modern tape drive, you usually have several dozen "heads" on any given tape drive, and there will be two sets of them each with its own mechanism to align it with a precision of just a few microns. Pretty amazing, really; if you drop by the Denver, CO area some time, the Oracle/Sun building engineers there can often arrange a tour of our tape testing facilities if you sign a NDA and represent a potential sale. Anyway, the second mechanism will be engaged on the tape in order to read what the first just wrote and verify it before it passes the "successful write" confirmation back up the fibre channel chain. This way you can guarantee you don't get "write once, read never" media.
Agreed. At my work we do parallel streams to multiple Sun T10000 T2 tapes (T10K "C" drives) at 250Mbyte/sec uncompressed (500 megabytes per second compressed, more or less, usually quite a bit more). If for some reason we push less than about 120mbytes/sec, the tape rewind times cause all kinds of issues.
We make the same kind of decision when choosing Sun T10000 "B" drives instead of "C" or the new "D" drives if the source cannot push data fast enough.
I've long laughed at articles saying tape is dead. For large-scale* backup, retention, transport, and legal hold problems, there simply is no other solution that scales reasonably well.
*My definition of "large-scale" for this specific context: hundreds of terabytes or more, much of it transported thousands of miles regularly. If you don't work with hundreds of terabytes and at least dozens of petabytes on a daily basis, you may suffer from optimistic delusions regarding disk storage capabilities, one which disk storage vendors are all too glad to reinforce, to the detriment of customers faced with half-baked solutions that cannot hope to meet their throughput requirements. Given "large-scale" data, there's no replacement for tape at present; everything else is a low-throughput also-ran, typically harboring enormous and unplanned complications. We're also heavy users of VTL, replication, cloning, S3-workalikes, and various disk technologies. Tape remains vital to large enterprise operations, and those predicting its imminent death have been the butt of jokes about marketing wonks for a decade and a half.
I just started playing EVE Online in February of this year after a long hiatus from all online gaming. It has a great community, and due to the way skill trees work and the variety of places to play in (hisec, lowsec, nullsec, wormhole) it can be as casual or as hard-core as you want it to be. I enjoy the heck out of wormholes at present!
Some of my co-workers were hired in California and moved to Salt Lake City. Many paid cash for their home in this area due to selling their home in CA.
In a large corporation, the downside of such a decision -- even though I know you meant the above somewhat tongue-in-cheek! -- is that someone paid at the high end of the bell curve is frequently passed up for raises due to the cost-of-living differential, regardless of job performance. The short-term benefit can be magnificent, but the long-term downside of promotions without attendant raises due to already having a salary "above" what the management manual suggests for the position in that area catches up with you.
Mexico does not have the time-zone disadvantage of India, and recently (due to various political and monetary factors) is very cost-competitive. Our Guadalajara office is chock-full of brilliant, dedicated people -- just like our India office -- but getting things done with their help is considerably easier due to the better time zone alignment.
Given the choice of hiring otherwise-equal software engineers in India for $25K/year vs. Mexico for $25K/year, I'd go with Mexico every time due to the time zone differential. Travel costs are cheaper. Conversations with US-based co-workers occur more frequently. Training time is quicker due to quicker turnaround on questions and feedback. You can obviously work around issues working with teams on opposite schedules -- and we do, every day! -- but it will never be as trivial as simply having most of your workers share common core hours.
Like I said at the start of the conversation: location and risk/reward are two fundamental aspects of how much one gets paid. It affects how easy it is to land a good-paying job, as well. In today's market, one's longitude matters a whole lot more than one's latitude; it's a lot easier for me to work with my great team-mates in Mexico during my normal working day than to stay up past midnight to meet with equally-talented teammates in India, Singapore, or Sydney.
Great link! Thanks for that. It shows pretty clearly something I've long suspected: that Salt Lake City, Utah is a "bellweather" area, smack-on the national average for technology pay. The more prospective employees educate themselves about these statistics, the better enabled they are to determine the right career path and negotiate fair pay.
I'd change jobs if my job consumed my life. Here's a breakdwon of a typical healthy week for a happily married father:
168 hours in a week.
15 hours spent alone with your spouse. Dates, hanging out, get a sitter for the kids, go to the gym together, go shopping together, etc. Yeah, I'm serious, if you aren't getting 15 weeks alone with your spouse you're probably
15 hours spending time with your kid(s). This can include chores together, hanging out, whatever. If no kids or they've moved out, you get this time back!
40 hours a week working (maybe 50; 60+ weeks should be a rarity, not a regular thing!)
10 hours commuting. Less if you can swing it.
56 hours sleeping (yeah, as if, but budget it anyway).
12 hours eating, shaving, showering, using the restroom, getting dressed, doing laundry & other chores, etc.
What's that leave you with? 20 hours a week that you can decide what to do. If you're working 70 hours a week, you've cut into the time with spouse and/or kids and have ZERO time for yourself.
Work/life balance matters differently to different people. Employers routinely expecting heroic effort is abusive. Luckily, I live very close to work and can work from home a good deal; the quality-of-life improvement from a reduced commute makes up for a LOT...
Two factors to consider: Location and risk/reward ratio.
The "$200,000" figure -- as mentioned earlier in the discussion -- is extremely location-dependent. Let's say you live in Salt Lake City, UT and make $100,000 per year. I'm from the area, and $90K-$110K seems to be a very typical pay rate for sysadmins and programmers with 10 or more years of experience in their field, if you include health and other benefits in the W-2 (all the employers do!). Then let's cross-reference that with the Cost Of Living calculator at http://money.cnn.com/calculator/pf/cost-of-living/ . To enjoy an equivalent quality of life, here's what that guy needs to make (these are all area averages, not the specific cities referenced):
San Francisco, CA: $171,987
Los Angeles, CA: $140,380
Seattle, WA: $123,784
Raleigh, NC: $99,154
Austin, TX: $97,991
Washington, DC:$151,479
New York, NY (Manhattan): $231,289
New York, NY (Queens): $162,684
Advice for earning the big bucks? If you want to earn a high-end "salary", move to one of the technology hubs. Get an education in finance and high-end mathematics. Or get a security clearance, shine up your resume and skills, and plan to have no life outside of work due to travel abroad. Get some solid experience, get in on some high-profile projects, and job-hop when you're at the apex of your visibility in your company for much better pay.
If you don't want to live in Silicon Valley or New York, then be your own boss. Write your own products. Innovate in a field that lacks innovation. Develop software nobody has really thought of before. Monetize it and make a living. Cash out by selling the company to a bigger fish, and start working on your next idea. My company, for instance, buys small companies who create innovative software a LOT and integrates that software into its other offerings. And that strategy works very well for continually expanding the customer base and revenue. Sometimes it works out well for the little guy, sometimes not. It's a risk of running your own business.
Working for someone in a "typical" IT or software engineering field is typically not gonna make you rich. It can make you a good living with solid pay and far less stress than running your own business or working in high-pressure financial and security markets, though. And you're less likely to lose several fortunes on your way to making a fortune if someone else takes the risk, and you get paid the salary. Risk/Reward.
For me, quality of life matters a lot. I've been on both sides of the fence (trying to make it in business for myself vs. drawing a salary), and right now in my life it makes more sense to draw a reasonable salary doing a job that I love in a great environment working with quality people and making a positive difference without feeling that my job is my life. But I know I'm probably not going to strike it rich doing so... and I'm OK with that.
I'm an atheist. Haven't ever told anybody that they are stupid to believe in their God. Your statement is disproven.
I'm really an "apatheist": a subset of atheism that simply believes it's a waste of time & resources to debate an unfalsifiable hypothesis. I don't care if the supernatural exists, and there is currently insufficient evidence to suggest that caring about the question matters in the slightest. There are a lot of atheists like me: we don't care what you believe in or don't believe in as long as you don't try to force us to believe what you do. Most of us are not the vocal minority you apparently assume all atheists to be.
Absolutely fair comments, thanks for the information that new tapes have a higher cost on a tape drive than used tapes. I should have said "millions upon millions of feet of tape", which would have been a correct statement. I stand corrected.
That's what the cleaning tapes in your silo are for. The heads are typically good for millions upon millions of read/write cycles as long as they are kept clean. The motors driving the reels are typically of the brushless, multi-speed variety, capable of decades of reliable operation with quality bearings. Cleaning tapes, on the other hand, must be replaced regularly as they wear out.
I frequently see bizarre claims on Slashdot, but the claim that using new tapes "wears down" your tape drive is a new and strange one. Where on earth did you get this idea?
I like your summary of three important reasons for tape archive. I'll restate in different terms.
1. Mid-term to indefinite data retention.
2. Large quantities of data, where "large" is a value greater than a single hard drive can reasonably store.
3. Disaster recovery planning.
But there are more.
4. "Oops".
That's the category of this KDE git issue. Recovering from an "oops". People screw up. How do you recover? I'm a big fan of having multiple layers in that onion: online snapshots, near-line replicas, and off-line tape backups are a basic three-tiered framework for figuring out how to protect the data. I'm amazed as big as KDE is, they don't have storage/backup expertise helping them keep their data secure. Makes me think I may have found my next open-source niche to fill.
5. Reliability. Contrary to the "fragile, expensive" opinion above, tape failure rates are demonstrably lower than hard drive failure rates despite regular handling. Research left to the reader; hard drives fail at a rate about fifteen times higher than their rated MTBF, which was already considerably higher than tape. Data on tape is far more resilient than data on a hard drive.
6. Cost. If you have to store data long-term, consider tape. Administrative, electrical, power, cooling, and storage requirements are all cheaper.
That's what I can think of off the top of my head; I'm sure there are more reasons for tape to be a good choice. The reality for many people that want to store their data "in the cloud" also is this:
I back up your "cloud" storage onto tape drives. Your cloud storage is only as reliable as my ability to recover it from a disaster.
I agree with you, except for this part:
It all depends on the scale. If you're talking a small project with a small budget, I agree with you: tape backups are overkill, too expensive, and kind of pointless. Your average open-source project is usually just a few gigabytes at most. Use a snapshotting, journaling filesystem, always keep each version in at least three different places, create a retention policy that makes sense for you based on the needs of your project, and you're good.
And you're right. Today's modern tapes are good for about 4,000 read/write cycles. Even if you get the tapes at a substantial discount, a 5TB+ tape is expensive to destroy!
But when you are talking large enterprise data archiving needs, high-end hard drives do not compete with high-end tape drives in the slightest. And in today's risk-averse corporate climate, a reasonable disaster recovery strategy is a MUST, and providing multiple tiers of storage -- online, near-line, and off-line -- is attractive. 9/11 showed everybody how quickly DR plans can melt.
I could go into a lot of specific numbers talking about how a few modern tape drives in a modestly-sized tape silo outperform similarly-sized hard drives for near-line storage in just about every category except random seek time by several orders of magnitude, but I'll leave that as an exercise for the reader. :-)