Amazon Offers Paid Web Database Service

firepoet writes "Amazon has released a new web-services based storage engine that looks an awful lot like a directory service: SimpleDB. While not supporting SQL per se, they offer several simple operations — CREATE: to make a new domain. GET, PUT, DELETE: to manipulate your domain. QUERY: to find things within the domain. Data is stored in cells that contain multiple attributes. A single attribute may contain multiple values. For example: (name, bob), (favoriteFruit, apple), (favoriteFruit, banana). Another interesting tidbit is the cost structure; you pay for how much data you store, how much you transfer, and how much CPU time the database uses while manipulating your data. 'Amazon SimpleDB is designed to store relatively small amounts of data and is optimized for fast data access and flexibility in how that data is expressed.'"

Google.

[Seumas (6865)]

It still seems weird that Amazon is doing all of this stuff, instead of google.

It's almost getting to the point where you might be able to code, run and host an entire online service or site on Amazon without any colo or shared hosting. Except for the price, of course.

Re:

[TrappedByMyself (861094)]

It's almost getting to the point where you might be able to code, run and host an entire online service or site on Amazon without any colo or shared hosting.

I think except for DNS service, people already do this

Re:Google.

[Baricom (763970)]

It already is that way. Several startups run almost entirely on Amazon.com's web services platform. The CEO of one of them, a sort of mash-up between YouTube and Ustream, claimed [roughtype.com] that "we don't own a single server."

For many people, the price is lower on Amazon's platform than elsewhere. I'm looking at doing ad-hoc webcasts through an Amazon S3 cluster with an open-source Flash video server.

Re:Google.

[l-ascorbic (200822)]

I think many people are misunderstanding what SimpleDB is for. It's about plugging the gap between S3 (Simple Storage Service) and EC2 (Elastic Computer Cloud). The former provides simple file storage, the latter hosted Linux images with no persistent storage. This means that while you can run web servers on EC2, databases are a no-no. While some people have hacked together persistent EC2 filesystems on top of S3, these are not really appropriate for databases. This new service means that sites with simple database requirements can run entirely on AWS.

Incidentally, I run a site [clevr.com] that uses S3 for all of the storage and delivery of media, and it's fantastic. Works totally seamlessly, is very cheap and reliable, and is about as scalable as it's possible to be. It means that the bandwidth used by our main server is minimal, despite being quite a high traffic, image-heavy site. This in turn means that we can have it colocated with a smaller host that's local to us.

Re:Google.

[vidarh (309115)]

It's only cheaper than elsewhere if you have really huge daily peaks and relatively small amount of data you need persistence for, or if you occasionally need large amounts of processing. I did the math for a setup of about 240 Xeon 2GHz cores, and I could easily find providers that would cost down to about 35-50% of Amazon with managed hosting with a much more manageable setup and a lot of functionality Amazon just doesn't offer. Ultimately I ended up looking at a setup involving using another service for the basic load, and then handling peaks for the highest 2-4 hours of the days by spilling over some backend processing to EC2 as needed.

Personally I don't think most people who think EC2 is cheaper for them have actually looked around at the alternatives. Most people I've actually talked to who think EC2 is so fantastic are acting like fanboys because of the API and haven't bothered doing the maths or looked at the competition.

If you have very small setups there are a ton of VPS providers that can beat them on price. If you have setups over say about $1000/month, managed hosting can beat them unless your traffic varies far more than normal over the course of a day (probably about 4x-5x above your base load, unless your base load is very low) and you're willing to scale your EC2 setup up and down over the course of the day, and even then you can beat pure EC2 by putting your base load and any servers needed for more than about 6-8 hours a day onto another provider.

If you're spending more than $50k-100k, leasing or buying the servers start getting even cheaper.

The one area where EC2 is attractive is when you need compute farms for temporary use ( (At the moment Softlayer [softlayer.com] is my favorite for managed hosting, btw.)

The advantage...

[SanityInAnarchy (655584)]

The advantage I see to EC2 is scalability. That's the reason startups like it -- when you get Slashdotted, you just fire up more instances (virtual servers). You can do this automatically, because of those APIs.

Also, EC2 and S3, and now SimpleDB, work well together -- specifically, EC2 for things that are really dynamic, and S3 for things which are only sort-of dynamic (you want to base them on a database so they're easy to change, but it's still going to stay the same for weeks at a time). And I'm not sure

Re:

[RAMMS+EIN (578166)]

``It still seems weird that Amazon is doing all of this stuff, instead of google.''

Why would that be weird? Just because everybody is obsessed with Google doesn't mean that other companies don't do cool stuff.

Re:Google.

[kripkenstein (913150)]

It still seems weird that Amazon is doing all of this stuff, instead of google.

It's almost getting to the point where you might be able to code, run and host an entire online service or site on Amazon

That is exactly why Google is not doing it, and Amazon is.

New startups running entirely on Amazon are, in general, Google competitors. After all, Google is in the business of constantly creating new apps and services, many of them of the sort that would use Amazon's services. There is therefore no incentive for Google to sell services that make this easier for others to do. The same goes for Yahoo and Microsoft.

Amazon, on the other hand, has a core business that is not competing with new startups of the kind mentioned above. And since Amazon anyhow has built for itself a highly-scalable architecture, there is no reason not to monetize it as they are doing.

It's a tuple store

[Animats (122034)]

It's a tuple store, like a Python "dict" or a C++ "map", but scaled up. This has its uses. One could implement RSS via a tuple store, for example. But it's not clear this is viable as a commercial service.

Re:

[TheRaven64 (641858)]

A better example might be LINDA [wikipedia.org].

Re:

[QuickFox (311231)]

A decade? You must be new here. More like a couple of days.

It looks like PICK lite

[Larry Lightbulb (781175)]

A quick scan of the details and it looks like they've based it on the concepts behind Pick.

Technology-wise, the interesting thing...

[frank_adrian314159 (469671)]

... is that a good deal of it appears to be coded in Erlang.

I'm having trouble with the article tags.

[The Ultimate Fartkno (756456)]

Does that one tag parse as "Cool as hell!" or "Cool! A shell!?" Either way, I'm troubled, because either someone is *way* too excited over a database or there's some sort of Mario Kart aspect to the design and I'm completely overlooking it in the article.

CouchDb is a better alternative

[RedDirt (3122)]

If you're looking for a buzzword-enabled web database, you'd do well to take a look at CouchDb [couchdb.org]. It uses an HTTP-based transport that's decently RESTful, JSON-formatted records and Javascript-based views as opposed to a query language like SQL. There's an active mailing-list [google.com] and a wiki [couchdbwiki.com] where you can learn more.

On the mailing list, I found this blog post comparing CouchDb to SimpleDB [automatthew.com]. It's kinda spooky how similar they are, though it looks like SimpleDB more closely matches up to the pre-0.7 release of CouchDb (XML results rather than JSON, less RESTful communications, etc).

Re:Vaild concept...

[CaptainPatent (1087643)]

...but who on earth is this targeted for? I fail to see the application of this service.

It's for people who don't have the time to learn the complexities of a database manager like SQL:

SELECT INSERT DELETE

and instead want to learn the much simpler SimpleDB:

QUERY PUT... DELETE

Now do you see!?

Re:

[mochan_s (536939)]

I highly doubt it's for SQL challenged people.

I think it's for people who really want a performance guarantee on their queries and absolute reliability for their data, Like expected query time is so-and-so microseconds etc and the data is never lost.

Yes, you can configure LAMP but you are hosting all your data in your site and if there's a fire, everything is gone. Plus, you have to worry about internet uptime and so on. I think this really is a data deployment service without having to worry about reli

Re:I do not understand Amazon's strategy

[AuMatar (183847)]

They're building blocks. The value is in having someone else write a highly scalable, highly available building block which you then program on top of. While writing a program to do this on a local machine isn't hard (its called a BDB), writing a webservice to do so to a 3 9s or higher level takes a lot of time, effort, and ongoing maintenance/support staff. Its easier and quicker to buy it.

Some are used internally too. But there are a lot of outsiders who use them, especially S3.

Re:"Small amounts of data"

[ZachPruckowski (918562)]

The advantage here is pure scaling. A lot of blogs have very sporadic load. They're virtually unvisited 75% of the time, but 24% of the time, they have a new post people are reading, and the other 1% of the time, they're being dugg/slashdotted/reddited/whatever. They need to build themselves around this major load in a LAMP server. You have 90% of your capacity to handle that 1% of the time. With SimpleDB and S3 you don't need to worry about that. You pay for whatever bandwidth you use, and whatever compute power you use (that's actually the catch on those 3TB/month hosts - more than a few dozen visitors and you're over your CPU allotment).

These start-up webapps like it because of the scaling. They don't need to go from a rented server to a dedicated one to 5 dedicated to 25, and there's no scaling issue with the DB or the CPU or anything.

I suspect that within 6 months, Amazon will have what you suggest - some sort of easy way to build a site using their tools. Even if it's just Wordpress with the default settings rigged to work with their tools, it'll be the poorman's website. Buy a URL, enter it in the dialog box, pick a theme, and you have your own commercial or personal site. Way easier than getting a dreamhost account, and way more options than a blogger.com account.

Re:

[vidarh (309115)]

Lets be realistic here - with EC2 you pay for Amazon to have the extra capacity sitting idle to be able to handle sudden large surges. EC2 is hugely expensive compared to a large number of hosting providers I've looked. So much so that when looking to move a large'ish setup to another hosting provider, the monthly cost for EC2 would've been $35k/month, and the preferred alternative would be around $20k-$22k with far better capabilities (i.e. no losing local data, as well as separate 1Gbps switched private n

Re:

[xant (99438)]

without a frontend web program like phpMyAdmin to manage creating and configuring domains, I can't see who they are actually targetting this at.

They're targetting this at application developers, like all their other web services. They want people to build, and resell, directory services; they're not trying to sell a directory service itself. It isn't that. It's just a web database.

The point of all the Amazon services is to encourage third-party vendors to charge in and develop the applications, and take

Re:

[Planesdragon (210349)]

A simple LAMP setup costs around $100 (one hundred) per year: you go to a Linux hosting company, and they give you everything, including provisioning, backup, etc., from a single source. And you get a standard database with local connections.

So, you push a button, and never touch it again?

Or is it just that you value your own time at $0 an hour?

Re:you're off by a factor of 1000

[Cylix (55374)]

I was kinda wondering when slave labor would migrate to IT.

I can see it now...

2 columns of 12 nerds each, all dressed in tattered white shirts and black pants, chained to computer desks on a boat. At the lead is their management cracking the whip overhead. "Code ye bastards, code!," he shouts! Behind him sits a man dressed in a Mountain Dew can outfit banging the drums... except his pace is erratic and he just giggles.

The slave ship LAMP, an interesting beast where even the best are broke. If you wondered where all those coders went after the bubble burst you might want to check the passenger log.

Re:

[m2943 (1140797)]

So, you push a button, and never touch it again?

Precisely. Which part of "hosting company" do you not understand?

The only difference between Amazon and LAMP hosting is that Amazon is proprietary and non-standard, and that if I go with Amazon's service, I'm locked in.

Or is it just that you value your own time at $0 an hour?

I do value my time, which is why I prefer industry-standard solutions (LAMP) with widely available tools and library support over Amazon's proprietary stuff.

Re:

[BitGeek (19506)]

"Amazon's infrastructure is orders of magnitude better than what your average chump can manage themselves."

This is a common assumption, based in part by Amazon claiming that these services ran the same software that ran their store. Well, the store's uptime is terrible, maybe %98, and these services are written from scratch anyway.

I think they have their place, but the assumption that Amazon means good engineering is a very erroneous one. I know, I used to work there.