Google gives away 100 PETABYTES of storage to irritate AWS

Google has flicked the switch to take its “Nearline” archival cloud storage service live, and tossed in an offer of 100 petabytes of free storage to set the snowball rolling. Nearline is Google's competitor for Amazon Web Service's Glacier. Unlike Glacier, which is thought to rely on Blu-Ray, Nearline uses disk and flash*. …

[Nate Amsden]

need more bandwidth

1PB of data will take a long time to import, even if I had 1PB of data to import to their cloud, even at gigabit speeds(on a tier 1 ISP) the latency alone I believe would ensure it takes longer than 3 months to copy that data(at least without massive parallelization).

[Lusty]

Re: need more bandwidth

Latency makes no difference to throughput if you know what you're doing, although the majority in our profession are sadly afflicted by poor throughput with latency.

The other two clouds offer 10Gb links via MPLS so although I've not looked specifically at Google I'm assuming they would too. a quick calc (365 / 2 x 24 x 60 x 60 x 1.25 /1024 /1024) suggests you could upload 18PB in 6 months with this arrangement, although it certainly would require more investment than free would suggest.

[Anonymous Coward]

Re: need more bandwidth

You must import 1PB within 3 months to benefit from the free offer.

If you don't have 1PB of data, /dev/urandom will do nicely.

Assuming they are using power-of-ten units like disk drive manufacturers (1PB = 1000^5 bytes), then to upload 1PB over 91 days you will need to push 1.02Gbps continuously - not counting TCP/IP framing overhead.

That's pretty awesome. So you'll need to get yourself a better-than-gigabit connection to Google - or at least, run your /dev/urandom generator inside Google Compute.

You'll be using a good chunk of data centre resources. Even if Google manage to squeeze 45 x 4TB drives into a 4U chassis, and they only store two copies of your data, that's about two racks that your 1PB of random data will be occupying.

But better cancel it before the free trial runs out, 'cos after that you'll be paying $10,000 per month.

[Doctor_Wibble]

Re: need more bandwidth

I'm guessing this storage would be most useful (or used as encouragement for) people whose businesses are almost entirely cloud based (also at google, obv.) and/or running large web-based data collection, storage and processing type businesses or even video streaming startups - who should have the sense to initially populate their free storage by means of a truck-load of tapes.

Or they want everyone with a DVD collection to upload them all and free themselves from the burden of having control over their own stuff...?

Or it's sponsored by the NSA who just want to trawl through your pr0n collection.

[TheVogon]

Re: need more bandwidth

What this really means is that hardly anyone is using Google's cloud so they are throwing about ever increasing abouts of free stuff in the deperate but increasingly unlikely hope of replicating the sucess of AWS and Azure...

[Andy Roid McUser]

Re: need more bandwidth

"Replicating the success of Azure"

Have you ever had to battle with Azure.. It makes windows 10 look like a finished product. AWS and Google Compute are vastly superior in terms of actually building a supportable cloud platform. A little off topic but those five little letters get my Friday morning blood boiling.

[crediblywitless]

...And then in half a dozen years' time, Google changes its mind about running the service, shuts it down, and gives everyone 30 days to remove their data from it and go away.

[Roland_Bavington]

Wot no Oracle?

The article doesn't mention us here at Oracle StorageTek, we launched an cloud archive service last month at $1/TB/Month as an extension to the raft of other products we already sell from Database or VM to a service all the way to ERP.

[future research]

Re: Wot no Oracle?

The article doesn't mention us here at Oracle StorageTek, we launched an cloud archive service last month at $1/TB/Month as an extension to the raft of other products we already sell from Database or VM to a service all the way to ERP.

That is very interesting, I had to go an confirm those prices, for two reasons, normal oracle don't publish list prices (if you have to ask you can't afford it) and the fact they are 10x cheaper than AWS. Oracle are not normally known for under cutting the competition, (although the ZFS file storage servers are very competitively priced).

It can only be a loss leader for oracle (like milk at supermarkets). It seam the industry as a whole is trying to ultimately go with the SaaS model. (e.g. Office 365) consistent revenue stream and to opportunity to hike the price of the on premise license to really milk those who don't want to / can't move to the cloud.

[Lusty]

Re: Wot no Oracle?

They probably didn't mention you because no-one cares, a little like Oracle Linux. They never mention VMware in cloud conversations either. If you were a serious competitor you'd be on the list. By serious competitor I don't mean offering a service which competes, I mean offering a service which competes and has the customer base, scale, and brand recognition in the sector. Oracle are big but not recognised for their cloudy products. I hate to say it but I would count Oracle in the "me too" category alongside HP where cloud is concerned. Perhaps I'm wrong, if so I didn't mean to cause offence it's just how I see the cloud landscape right now.

FWIW how often do you see Amazon mentioned in database articles? They have some extremely capable products there...

[Jon Massey]

Re: Wot no Oracle?

that and it's only available in North America

[Anonymous Coward]

Re: Wot no Oracle?

They probably didn't mention it because it had its own Register article already.

http://www.theregister.co.uk/2015/06/23/ellison_aws_price_war_amazon/

[Graham Cobb]

Re: Wot no Oracle?

I went and looked at the Oracle offer after the earlier Reg article. Unfortunately there seemed to be two problems:

1) You don't seem to want personal customers (and probably not SMEs either -- the signup process was very corporate-oriented, unlike Amazon's).

2) You don't seem to allow for import/export using harddisks -- I use Glacier and I just send a copy of my backup disk to Ireland to import a terabyte or so of data. Uploading that over my DSL connection would be hopeless!

Of course, I'm not particularly surprised: Oracle's customers are mainly large businesses, and your very low price is unlikely to be worthwhile for Oracle for people who only need a few TB.

[Anonymous Coward]

Considering that the calculations people have done above are talking about at least 90 days of flat out use of a 1Gbps connection to qualify for this offer how many companies are there that really could qualify? I can imagine large companies having 1Gbps connection but dedicating them to just the backup, I don't see it. Perhaps my view is skewed but we provide services to some large companies and they are precious about a 100Mbps connection and would probably shout as us if we filled it for more than a few seconds at a time.

[bigtimehustler]

"Glacier is optimized for infrequently accessed data where a retrieval time of several hours is suitable" - When they say this, i really think they have forgotten what the word optimize means. Nothing has been optimized here at all, they are just using cheaper storage methods where speed doesn't matter. I would hardly say using cheap components and providing a slow access speed is an optimization! It be perfectly acceptable and what people want, but optimized, no.

[MatthewSt]

Surely using cheap components is optimising for cost?

[TeeCee]

Hmm, I wonder which is actually cheaper. Enough BD jukeboxes and (re)writable media to handle that much, or a metric fucktonne of el-cheapo rotating rust with a bit of flashy cacheing......?

[Lusty]

Ultimately the offline will end up cheaper as dedupe rates begin to suffer with scale. The more data you have the more often you hit the same hash for different data and eventually will get to the point where storing as dedupe actually takes more space than storing native data (because all the pointers etc. need recording alongside the data). BD wouldn't be deduped so as long as ultimate capacity is cheaper that way then it's a winner.

Sadly, I have always seen Bluray as a dead technology thanks to the Internet so have no idea what a blank disk costs for comparison

[naive]

It is all great, but the telco's do not play ball

Offerings like this are great, watching the pain of backup/restore with a multitude of products on daily basis, cloud based backup and archival sounds like heaven, offering great potential for increases in efficiency.

However, bandwidth to the internet is an issue, limiting usability and access to cloud services.

Telco's are lagging, and desperately concentrate on milking the "last mile" cow for decades, instead of enabling innovation. The year when Gigabit connections to Google or AWS bitbarns become affordable, is the year of the cloud.

[Lusty]

Re: It is all great, but the telco's do not play ball

Simple answer - don't connect to your cloud with the Internet. Cloud doesn't mean Internet connected, you can use MPLS to all major providers.

[Phuq Witt]

Good News, I Think

Not because I'm planning to move my AWS backups to Google. But I'm expecting enough people will, that Amazon will be forced to further reduce their own prices in the next tit-for-tat round of 'Cloud Wars'.

I'm sure said war will all end in tears. But, while it lasts I'm not complaining about my bills getting smaller.

Yay for ruthless, cut-throat capitalism!

[Bloakey1]

Re: Good News, I Think

Putting your data in the cloud is like lending your wife to the bloke down the pub.

You can't guarantee her integrity or his.

you are still legally responsible for her.

you do not know if she will come back when you need her.

Will she come back in a timely manner if at all.

When and if she does come back she might be corrupted.

She might not come back in the right format (pregnant).

She might have been interfered with and no longer be useful to a native app.

The permutations are endless, I keep my data close because I care.

[Amorous Cowherder]

Amazon charge for deletion?

Don't Amazon charge for deletions over a certain amount in a calendar month on Glacier? So even if you do try to "move" aren't you going to get stung when you delete everything from the old Amazon host?

[Anonymous Coward]

Amazon Glacier is not BluRay

I know that's widely repeated, but the origin seems to be one guys speculative "how could they possibly be doing that?" opinion column.

I can tell you for certain it's not BluRay. I can't tell you what it is for confidentiality reasons, but it's not optical media.

[chris 252]

Legit interesting and only a little bit ahead of the curve

More and more of the organizations I work with are making 10GbE connections to the outside world which means this sort of petabyte-scale storage offering is within the realm of possibility.

And the public Internet is not the only way there -- AWS and Google along with MS and others all peer at 100GbE speed with high speed research networks like Internet2 (in the US). A 10GbE connection to Internet2 gets you a fast path to multiple IaaS clouds on a network optimized end-to-end for high speed parallel data transfers. I2 is more preferable or at least seems more purpose-built for the use case than some telco trying to sell me the same thing on MPLS

[Puffin]

Half a million a month?

If my maths are right, 100 Petabytes at 1cent per gigabyte per month is more than half a million quid a month? Nice business if you can get it.

[Federal]

It's not cheap

A 4TB WD NAS rated (Red) drive goes for $150.

At 0.01 / per gb/ per month for 5 years, Google will sell you the same drive for 0.01 x 12 x 4,096 x 5 (cents times months times gb times years) = $2,457.60

Plus I/O charges etc. and what if they decide to get out of the nearline storage business?

For nearline storage, the drives don't have to be kept powered up. And you need to keep a sysadmin to make sure the stuff gets to google or gets to a local drive and to keep track of what's where.

The "savings" are smoke and mirrors.

[ragge]

Why should the tape vendors fear?

It is still 150 times as expensive as tape if you keep your tapes for 15 years (both StorageTek (10K) and IBM (Jaguar) state that their tapes are good for 30 years).

(Tape drives and library not included, you may add a tenth to a fourth of the cost or so, depending.)

Why do you think that StorageTek can do it as a service for one tenth the price?

Tape may be dead - in the heads of those that do not do the math.