Oracle EVP John Fowler has damned disk technology, praised flash and said tape has a future in an Information Week interview and yesterday's Oracle strategy webcast. Fowler is Oracle's top dog for server and storage systems and it's worth paying attention to what he likes and dislikes, since systems embodying these views will be …
"nothing else comes close to its ability to store masses of data cheaply"
... apart from hard-drives. In today's technology, with an 800GB LTO-4 tape costing £20 and a 1.5TB SATA drive costing £60, there's not much in it.
Of course it takes a lot of juice to keep all that spinning, so you'd spin them up and down on demand, and electronically switch the SATA connections. That would be faster and much simpler than having a robot throwing physical tapes around.
Much better than I could have said
Tapes are just a pain in the backside, with multiple systems you lose an hour a day just changing the things.
Now offsite your data (and no-one better mention the c**** word).
Admittedly, SATA drives are a bit heavier per TB and perhaps more fragile than tapes, but they take less space(*). So I don't see why you couldn't unplug them and take them off site.
(*) 800GB LTO-4 tape 102.0 mm x 105.4 mm x 21.5 mm and 200g (IBM datasheet). 2TB hard drive 26.1 mm x 147 mm x 101.6 mm and 750g (Western Digital Caviar Black)
A more realistic comparison
LTO 5s are available now and (from the Sony data sheet) they are 203gm with a native capacity of 1.5TB, so about 40% of the weight per TB, although it's not realistic to compare with a raw 3.5" drive. If you are off-siting the latter (and especially if handling them in bulk) then they have to be ruggedised and put into some form of robust enclosure that protects them from knocks. RDX are probably the best known company that supplies such devices and they have 1TB units at 650gms, but they use 2.5" units which are much more suited to being shipped around in a ruggedised container (3.5" drives are much more vulnerable).
As such, the realistic commercial alternatives to LTO5s for removable storage is about 650gm/TB vs 140gm per TB for LTO 5 and 250gm per TB for LTO 4. Of course there are some possibilities with disk - like de-dupe, although you can play such games with VTLs if you so wish.
On the RDX option, throughput is limited to about 45MBps (at fastest) which compares with LTO 5 (native) at 180MBps.
Of course if you can afford the bandwidth and the right software you can avoid this by remote backup to disk, de-dupe and using incrementals (although heaven help you if you need to fully restore a 40TB database over a link designed with incremental backup volumes in mind).
That said - if your business is small/medium or if it is for domestic use and you don't mind the manual hassle then it's perfectly reasonable to use a cycle of 2TB eSATA drives and have a trusted member of staff keep copies off site. It will be cheaper than an enterprise standard tape drive, and not many small/medium businesses will have the sort of disk storage system than can feed and LTO-5 drive at full chat (with compression) as it will require minimum data feeds of 200+ MBps to avoid shoe-shining. For many medium sized companies the RDX solution will make sense, but if your backup requirements are in the PB region then it all looks rather different.
There's no doubt that tapes are more of a niche product, but having recently experienced "commodity" server type drives in arrays (no names mentioned) of 10% annual failure rates (MTBF < 100,000 hrs) then disk reliability can also be an issue.
Really, and how does this relate
to the enterprise 60-200 drives 10000 slots world, where a 800GB tape runs about $30 and there always will be some "many hundred TB" caching in front of the jukebox?
Two guys will go there to stuff in new tapes and take out the ones for the offsite storage every other week and thats it.
We got backups of about 1200 systems going to tape (via a VTL, yes, who cares, the PC servers can't feed data enough for the tapes). 8-10 hours per month are spent thinking about the tape machinery, 4x2 hrs are reserved for in-out of tapes.
So sorry, not sure if you can blame the tapes.
Why offsite on tape?
Buy a another set of SANs and live mirror the primary to the offsite secondary. Even more so if you're using stuff like the vaulting type devices from EMC and HDS. We store all our email on vaulted archive SANs, the data is locked in for 7 years and cannot be removed without calling EMC in and asking them to wipe the box. We have a couple on primary site and live-fed duplicates at the DR site about 7 miles away.
Risk to drives...
Disks risk not powering up, or damage if moved. Not an issue with tape.
EMC must love you since they can own charge you as much as you can bear over the coming years, now they have you locked into their highly proprietory system. And only they hold the key. I bet they played the Compliance Card!
It seems to me that tapes have been loosing ground to hard drives in recent times, both in terms of price and capacity.
I'm looking hastily on amazon for tape media (just the media).
The best price I can find is around $75/TB for "LTO 3 ULTRIUM 400GB".
On the other hand, I can find new drives for $47/TB.
Disks can be backed up to directly, whereas tapes may need another grand or more for a tape drive.
It looks like hard disks have an edge on tape prices. Is this right?
The characteristics of each are probably more important than the price anyways.
Nevermind the media
The US$1000 price tag on a tape drive is enough to put people off. Perhaps they should also think about lowering the price of drives.
That said, $75 for 400GB is rather reasonable if you need the media to be slightly more robust than a hard disk.
My head hurts
Disks are "old and fail a lot", but tape is good? Exsqueeze me?
sounds like he wants to bring back the 8-track tape as well.
I know tapes are cheap, but this all sounds like a step back to the dark ages. Considering disaster recovery, what preventative measures would you take? write everything to 8 tapes at once like RAID or something?
Mine's the one with magneto-optical discs in the pocket.
Tape, I thought it had died out when flash came along
Isn't there only like one company in the world that even makes tape anymore. Not talking the resellers. I mean the actual third world (too hazardous for westerners) factory that adds the horrendous chemicals together?
There was a piece in the register a few years back about another tape manufacturer giving up the game because flash was coming out.
Flash & Tape
Flash and tape do not compete - at least not in the enterprise market (tape was not ever that big in the consumer market).
Flash replaced portable disk media, very small form factor hard drives, some optimical media and encroaching into the enterprise disk and other high performance markets. Nobody is replacing large tapes with flash.
return of the tape monkey
Maybe they feel sorry for all the increasingly idle chaps in the vast tape libraries out there.
With the super low wages now on offer in this brave new world its a no brainer I suppose.
Budding vertically integrated monoploy?
Given the control that Oracle now have in the ERP, middleware and database market and that they now have a server/storage vendor which they can't yet provision, they are now in the perfect position to create a near monoploy vertically integrated business (obvious competitors are IBM except they lack much in the way of ERP packages). It's very easy to see that Oracle could produce an appliance which was optimised to run their database and middleware and make it very difficult for the specialist server and storage vendors to compete.
This move to collapse the storage and server layers into one is not desirable if you want to avoid such tie-ins.
... it does not make good architectural or business sense* ... I cannot see it flying except for 'enterprise behemoth wannabe' co's where Oracle will price their 'Silo' so aggressively and competitively combined with a healthy does of flattery salesmanship that they feel they cannot refuse ... with Oracle laughing all the way to the payback when their 'prey' is bigger and 'trapped' with them ...
I see the gap in the market, but is there *really* a market in the gap for this that is not adequately addressed today? My gut feel would have been that the course grained, low coupling of SOA architectures is more to blame for system wide latencies than the disk IO and storage? Sounds less like a problem in the market and more a problem of how Oracle will continue to grow.
* exception may be for very very VERY data intensive processing like ... err a Google?
Disks - 'old and fail a lot'
Unlike tapes which are the very latest technology and never break, stretch, snarl up or print through.
Wasn't Oracle the company...
...pushing "thin client" years after dirt cheap memory and dirt cheap processors made "fat client" correspondengly cheap?
Re: Wasn't Oracle the company
Only sort of. As you are presumably aware, the company that really went to town on this implausible idea is the one that went bust and was bought by Oracle. But yeah, fair point, well made.
Thin client rules
Thin clients are not about HW price but maintaining the infrastructure. And if you've ever worked in a Sun Ray environment you'll hate personal PC as the plague.
RE: Thin client rules
Of course, if you've ever worked in a virtualised desktop environment, say VMware's, where you're pushing out Windows desktops from centralised servers to thinner desktop PCs, then you'd appreciate not having your users moan about not being able to read Windows docs, about having a far wider an varied application stack available to your users that they are actually familiar with from home or previous jobs, and still having a centralised infrastructure for easier management. I could say the same about Linux but I don't personally see it as being as mature or user-popular as virtualised Windows desktops. I would suggest your hating personal PCs is more due to a quais-religeous devotiuon to Slowaris than any real business advantage.
...what happened to the last tape I placed in the last tape-deck I had installed in a car stereo, 20 years ago... (yes, it chewed the tape nice and easy, and spit it back outside).
And whenever people mention tape storage, it reminds me of those huge tape drives that movies love to show, when they want to situate someone on a computer-filled room, such as Air Traffic Controller or nigh-impenetrable james-bond-villain base.
The funny thing is, those movie gadgets are always revving at constant speed, either pretty slow, or rewinding-fast. I bet those are 8mm reel film drives behind the animatronics. Or the actual thing from PDP-11s, programmed to behave that way, which would be funnier and classier.
I am talking about these:
They look great on Hollywood movies...
Now, seriously, tape drives are meant for backup purposes only. Random acessing on those? You kidding? Oh wait...
"Fowler is Oracle's top dog for server and storage systems and it's worth paying attention to what he likes and dislikes"
So he was part of the merger of SUN and STK and then asleep while the value disapeared until they begged for Oracle to buy them.
Man he is so influential I will go back and read everything he ever said.
It's worse. Fowler was the Sun "genius" given charge of the Sun Platform/Software Group waaaaay back on 2002 (or thereabouts, my memory is definately not as infallible as tape!). Yes, that's the same Fowler that screwed up Java, SunONE and Slowaris and had a big hand in crafting the downward spiral that was Sun's SPARC development plan. Any number of execs from those days will recal the grandiose promises the same John Fowler made, all about how Sun products were going to be the fastest/cheapest/coollest, etc, etc, and equally fuzzy roadmaps as the ones the Reg has posted. Sun failed to deliver on Fowler's promises back then. Sorry, but anything that comes out of Fowler's mouth needs to be taken with a truckload of salt. This would have been a whole lot more believable if it had come from Larry himself, but then Larry tries to keep at least one foot in reality when he does roadmap sessions (I'll ignore the "Unbreakable" Oracle pitch for now).
Ramdom Access Tape?
Someone mentioned Random Access Tape and PDP-11 in the same post. If memory serves me correctly DEC (maker of the PDP series) had DEC Tape, which was random access. Of course it was only random in that you could 'seek' to a certain block on the tape, but, had to pass all that tape to get there.
...energy consumption? Does tape cost more or less energy per Mb written or read than disk and flash, including cooling needs? It seems like this is a very large part of the budget for a data center these days so maybe some coverage of that aspect would be good.
energy angle: disk vs tapes:
While it depends, in general tapes are a lot cheaper per MB. Since a typical disk keeps spinning regardless of whether someone is watching it or not. Something like a street lamp which is always on regardless of the number of people on the street.
.Whereas a tape spins only when needed. So if you've a good app which caches all it's working set in RAM and well provisioned system which doesn't thrash - you have a tape which consumes very little (standby) energy. (eg: a database with a relatively predictable set of queries).
But if you've an extremely IO oriented app which needed a lot of random storage access and ran for a long time... (eg: a web server) - a disk might turn out to be better (& faster).
He does make a lot of sense
1. When flash reaches high enough capacity for home use at low enough prices, the market will slowly abandon spinning drives. With less traditional drives sold, losing economies of scale will slowly hike the price of HDDs, closing the gap even further in a positive feedback loop. As consumer drives go up in price, so will enterprise drives. This does not affect tape, which was always niche compared to disk.
2. Bit density on tape drives still has ample room to grow. T10K cartridges have surface area of about 75,000 cm^2. Compare this to about 456 cm^2 maximum for 4-platter 3.5" disks (I'm assuming 3.5" platter diameter with 1" diameter hub). The bit density for T10KB-formatted tape is about the same as of a 6 GB disk. There's ample room for growth. Assuming four-platter 2 GB disk bit density, a typical (4x5x1") cartridge could hold over 150 TB of data.
3. T10KB has 240 MB/s native throughput, not 120 as in the article (that's the throughput of the original T10K). A 20 TB cartridge will store data at 20 times the density. Assuming there would be 144 tracks (compared to 36 of T10KB), linear bit density is 5 times higher, so 1.2 GB/s throughput should be achievable. Assuming 100,000 slots means 10 connected SL8500 libraries with 64 drives each, that 1,380 TB/hour translates to almost precisly 600 MB/s (given rounding, it's insignificant).
4. As opposed to LTO, Storagetek drives maintain backward and forward compatibility with the same cartridges usable on various generations of equipment (based on the formatting), regardless of technology or format changes in between. It can be expected that the T10K cartridge will be usable on T10KC or T10KD drives, depending on their underlying technology. Obviously, Fowler may have meant 20 TB compressed capacity, which makes it perfectly viable -- 10 terabytes in 2015 seems almost like a breeze. Assuming a 2 TB T10KC is released before May 2011, 4-5 TB T10KD in 2013, 10 TB T10KE is certainly possible in 2015. 20 terabytes native is significantly more involved and would possibly require Storagetek to break backwards compatibility.
5. At some point, it may be possible that flash becomes significantly cheaper (although it's doubtful that progress would be notably faster than Moore's observation suggests, though 3-bit MLC could allow flash to overtake Moore's, as could 3D cells suggested by some people), and tape storage will be on the way out, possibly replaced by switched SATA/SAS in a MAID (zero spin-up time could make it possible). This of course assumes that the high-density storage is indeed cheaper to make and that there will be people willing to pay for lower tier (slower, but higher capacity and/or significantly cheaper) SSD storage.
My overriding memory of tape drives, when you have a lot of them, is the endless cycle of repairs they need. If you didn't get one failed tape drive per week you weren't really trying. I for one have never missed getting rid of tapes.
I read the title as "Oracle hates disco, loves tapas". Maybe I need to increase the font size in my browser.
Spinning tapes are good, but don't forget clattering relays and blinking lights to complete the mysterious supercomputer look.
The tragedy of it all
Tape, rightly, SHOULD still be the default, mainstream option for anything that isn't required to be online.
It SHOULD be massively cheaper than disk, for everything beyond a primary, online copy.
It is inherently much more flexible, automatable, reliable, scalable and manageable than any system that bundles the electronics and read-write mechanisms with the media package.
(it's also massively much more dense, requiring less physical estate)
Somehow, the tape manufacturers missed out on the opportunity, They let the disk guys sell to management on the basis of lower cost per GB, without challenging all the downsides; when they should have been offering a much lower cost, with all the advantages.
It LTO had been price pitched correctly, and staged file aging been readily pushed onto the desktop, every home user would have one - you want to watch a movie you downloaded three years ago and haven't seen since - no worries - two minutes later it's back on disk.
You want to have every document your organisation has ever scanned readily available on every desktop - no worries, within seconds of your request the robots either got it online, alerted the op to get it from the near-online racks, or issued a request to the offsite facility.
You want to have triple redundancy of everything, ever - no worries, the robot spends it's evenings defragging and replicating everything you think is important for multiple offsite copies, and requesting back tapes that might be aging a bit to reduplicate.
etc etc etc.
Tape should, by all rights, rule the off-line data world. Sure it requires a bit of management, but mostly, the tape vendors have no-one but themselves to blame.
Fowler doesn't make much of a play on the benefits of de-dupe which have a massive upside on long term data storage but realistically need random access. Nice fat SATA disks are lovely for it, particularly if you can build de-dupe pools and then spin then down. You can regularly health check the disks (which you can't really do with tape)... Can you imagine the access time on a 20TB tape? Or the shoe-shone? Horrible... Can you image the impact of losing an unprotected 20TB tape? Nasty... Folks will have to start RAIDing tape somehow (probably double write). All this is pretty obvious to the industry and not to Oracle.
I have to say that this announcement appears to be more about hyping Oracle assets than delivering a sensible product strategy and we've not even started talking about whether the industry actually wants Oracle's bespoke stack approach at all. Perhaps Fowler has forgotten what OpenSystems is all about.
Has Anyone Here Actually Seen and LTO 5 Machine in Operation?
Assuming one can stage the data to be streamed, this drive is staggeringly fast at backing up large volumes. Removing a tape for offsite archiving has saved many organisations over the years who have had to rebuild systems following catastrophic primary site failiure. Tape was never positioned as the 24/7 availability solution, but sometimes recovery points in minutes (or hours) are 'good enough' and cost effective.
Sure you can spend 100s of 1000s of dollars on multiple site, multiple vendor, replication strategiesat HW, SW and DATA levels, and VMware are doing very well at peddling ever complex virtual back up and recovery stories with EMC. When the building is down and data centre burned buried (think Bunsfield). That little tape monkey referred to on this thread, can still, get those tapes out of afiresafe somewhere, drive them in his van, do a bare metal rebuild, especially if he uses boot from tape, and get a system going from scratch in a reasonable amount of time, when all else may appear lost - and the SAN/NAS fusion gurus are looking for one neck to choke in their multi vendor overly solutions that looked so good on the white board...
LTO Tapes have a tested minimun 30 year data life and onboard physical encryption that meets mots security reqs and regs, as opposed to fudging it with proprietory HDD access software for teh non tape vendors. I am not aware of any HDD manufacturer, especially SATA drives that offer any data life guarentee at all.
I cant coment for the STK formats but LTO has good backwards compatibility.
HP and IBM sell several billion $ tape each year, and this has declined/levelled in some sectors, mainly SME (never was in consumer) who have yet to experience a total meltdown. That Oracle are re netering the fray means tape is not dead.
Tape has done and seems set to remain a great fallback for the more experienced data centre managers who have seem every conceivable failiure. You won't find too many about to rely soley on spinny cheap sata HDDs for Backup/DR.
Oracle will have and end to end solution from the application to the 1s and 0s on some sort of media, and that is a good story.
Reality may of course difffer from the white board topography scetch and that financial scenario that the salesman so skillfully sold to the CFO....
No one will get fired for using simple tape backup.
- Boffins attempt to prove the UNIVERSE IS JUST A HOLOGRAM
- China building SUPERSONIC SUBMARINE that travels in a BUBBLE
- Review Raspberry Pi B+: PHWOAR, get a load of those pins
- Experimental hypersonic SUPERMISSILE destroyed 4 SECONDS after US launched it
- That 8TB Seagate MONSTER? It's HERE... (You'll have to squint, 'cos there are no specs)