Amazon DynamoDB – a Fast and Scalable NoSQL Database Service from AWS

http://www.allthingsdistributed.com/2012/01/amazon-dynamodb.html

by werner • 15 years ago

296 126 15 years ago

dannielo22• 15 years ago

"Amazon DynamoDB stores data on Solid State Drives (SSDs)" This is big.

ceejayoz2• 15 years ago

I'd imagine we'll soon see a SSD option for EBS volumes. Sounds like it'll be pricier (as you'd expect) - $1/GB for DynamoDB storage (but replicated to three volumes, sounds like).

onk• 15 years ago

When Amazon notes "SSD," from a client's perspective, it is only marketing. The storage media matter when you are managing your own hardware. The storage media do not matter in SaaS. For example, the storage media could be floppys and you would get satisfactory performance if there was a memory cache. Similarly you could get poor performance with SSD media if the networking layer(s) were slow. Similarly if the storage media were fault-likely CD's that wouldn't matter either, to us, because of the data replication performed in "the service." What matters in this case is the reported and actual latency.

zerosanity• 15 years ago

From the limits page(1) I see you can only have 256 tables per account.

(1) http://docs.amazonwebservices.com/amazondynamodb/latest/deve...

mikehuffman2• 15 years ago

The fact that you can make composite keys should make that irrelevant.

Edit: to clarify, I think that composite keys would make your problem limited to 256 distinct applications, not, for example, 256 customers, users, etc.

werner7• 15 years ago

As with all the other AWS services you can have your limits lifted upon request.

yahelc• 15 years ago

Does that apply for the 50-per-account cap on S3 buckets as well?

justinsb8• 15 years ago

Can the 10GB per domain limit be raised on SimpleDB? Or do I have to promise to refer to it as DynamoDB to do that? :-)

mikehuffman2• 15 years ago

While putting my data completely in the hands of another company makes me nervous, I have used s3 since its release and have had no problems. Amazons really seems to "get" developer needs. I do wish they would ease down a smidge on thier pricing in general, but otherwise I really feel about them the way I feel about google at this point.

sehugg• 15 years ago

Awesome. Anybody want to sublease a 3 yr RDS reserved instance? :P

rojabuck4• 15 years ago

Fast, Consistent, Secure & Replicated.

Game changed.

Maro• 15 years ago

Which game?

nosequel3• 15 years ago

Not really if you pay attention to the set of NoSQL DB's available already. They are just hosting their own now.

jarcoal• 15 years ago

Quality hosting/management is a game changer in my book.

rojabuck4• 15 years ago

Would be interested in details about any with Strong Consistency, decent replication across fault areas, which get the same level of performance (single digit ms) at a cost equivalent to dynamodb.

Quite a lot can be done with smart use of hugely scalable simple key-value maps.

gfodor• 15 years ago

"They are just hosting their own now."

^^ There's your game changer.

ytNumbers• 15 years ago

Here's a brief overview comparison of DynamoDB vs. BigTable:

http://vschart.com/compare/dynamo-db/vs/bigtable

The site is also able to compare many other different databases.

dhimes• 15 years ago

It seems as if this site is a user-edited wiki, and there are a lot of things that need to be filled in (in case anybody is up to speed on DynamoDB and wants to help). For instance, the Map/Reduce entry was still '?' when I wrote this.

lyddonb• 15 years ago

And Big Table (or at least GAE) does support transactions.

ridruejo3• 15 years ago

I simply do not see how competing cloud vendors can keep up with this. Most of them are still struggling to provide anything beyond a simple API to start/stop machines.

balakk5• 15 years ago

Well, Azure has been offering a similar NoSQL service for a few years now.

http://sigops.org/sosp/sosp11/current/2011-Cascais/printable...

ridruejo3• 15 years ago

No. They offer the equivalent of Amazon S3, SimpleDB and SQS, but nothing comparable to this

vyrotek3• 15 years ago

I've been using Azure Table Storage since the beginning and this doesn't seem to be the same. Like others have mentioned TS is more similar to SimpleDB. Now, I would love for someone to give me a tl;dr on the the feature set of DynamoDB so I can make an accurate comparison.

Table Storage does not allow any other indexes other than the main primary ones (Row Key and Partition Key). You also cannot store complex object within fields and use them in a query. You basically just serialize the data and stuff it into the field.

The dynamic schema is very nice if you can leverage it but the actually query support is TERRIBLE. (Sorry Microsoft, I'm a fanboy but you blew it here). There is no Order By or event Count support which makes a lot of things very difficult. Want to know how many "color=green" records there are? Guess what, you're going to retrieve all those rows and then count them yourself. They're starting to listen to the community and have just recently introduced upserts and projection (select). I would love to see them adopt something like MongoDB instead :)

For more issues check out: http://www.mygreatwindowsazureidea.com/forums/34192-windows-...

Edit - For what its worth. We've moved more things to SQL Azure now that it has Federation support. Scalability with the power of SQL. http://blogs.msdn.com/b/windowsazure/archive/2011/12/13/buil...

toddmorey• 15 years ago

Open source really helps here. Amazon are innovative, but they are not the only place innovation is happening. In fact, here's a pretty good writeup (if a wee biased) on how the new offering compares to to the open source Cassandra project: http://www.datastax.com/dev/blog/amazon-dynamodb

ridruejo3• 15 years ago

Cassandra is a great project, as it is Hadoop, MySQL, etc. The issue I am raising is that it is not so much which project is better on a feature basis, but the fact that Amazon is able to offer it as a service, in a scalable way that no other vendor is able to do (with the exception of Google and, on a good day, Microsoft). Most other "traditional" cloud vendors, such as Rackspace, do not have anything remotely comparable to this, EBS, SQS, RDS, etc.

mark_l_watson2• 15 years ago

For about $15/month for minimal "reserved capacity" charges and $1/gigabyte per month replicated disk space this service looks like it will cure a lot of deployment headaches for a reasonable cost.

Interesting that when I signed up for the service that they verified my identity with a cellphone call, like Google sometimes does.

bconway• 15 years ago

What's the concern for lock-in here - how difficult with DynamoDB be to migrate away from? Amazon seems to be increasingly catching the App Engine Syndrome, though to be fair, it's been mostly network traffic-style functionality in the past.

pacala• 15 years ago

In Simple Hash Key mode, it is just a key-value store. All the logic is in your app and/or supporting libraries, not in your DBMS.

In Composite Hash Key mode you also get limited range queries. These are supported by any DBMS that indexes its primary key via BTrees.

DynamoDB is cutting the features to the bone. This makes it easy to migrate out of DynamoDB and, at a first glance, hard to migrate into DynamoDB. Particularly hard to migrate from a RDBMS with complex schema/index/trigger support.

The beauty of it is that complex features can be built as libraries as the needs arise. The next five years look very exciting. Hats off to Amazon.

latch4• 15 years ago

Lock-in? You'll need to get all your data out and transform it into whatever alternative you pick. I'd say as far as AWS lock-in, this is on the low side of things.

justinsb8• 15 years ago

The lock-in comes not from the data format, but rather from the quirks of the particular system du jour: API lock-in.

dannielo22• 15 years ago

There are some bugs in signing up for the service. In console, if I go to dynamoDB tab I am asked to Sign Up first. I click the sign up button and I am told "You already have access to Amazon DynamoDB". Repeat.

werner7• 15 years ago

Can do me a favor and drop that in the AWS DynamoDB Forum so folks can look at it? https://forums.aws.amazon.com/forum.jspa?forumID=131

swatermasysk• 15 years ago

I would love the option to add N index (at cost).

My guess is they will add options for additional indexes in the future...everyone needs start somewhere. Even at Amazon's size and scale.

jsolson2• 15 years ago

> Even at Amazon's size and scale.

At Amazon's size and scale, it's all the more important that you start with something simple with well-understood performance characteristics from day one. AWS doesn't really get a grace period during which they get to fix scalability problems.

adpowers• 15 years ago

Exactly, compare this to SimpleDB. SimpleDB started out with an advanced query language that let query and filter your results in all sorts of ways. And guess what? SimpleDB is still limited to 10 GB per domain (aka, database). Want to horizontally scale? The official suggestion is to shard your data across domains. This is a really messy solution because you have to preshard based on estimated database size and resharding is nearly impossible (you'd have to rewrite your entire DB).

AppEngine went the other route and provided a very simple database API at first and all queries had to be range scans over an index. Any query you wanted to perform had to be precalculated by defining a composite index and some things (like inequalities on multiple fields) weren't supported. Over time they've built upon their basic database and added features such as a zigzag merge join algorithm which lets you perform queries that were otherwise impossible with a given set of indexes.[1]

I bet DynamoDB will be going the AppEngine route by starting with a simple, scalable base which can be used to build more advanced query engines and features.

1. http://code.google.com/appengine/articles/indexselection.htm...

Smrchy• 15 years ago

Why is there a 1MB reply limit on BatchGets and obviously no limit on Gets? Did anybody find limits on a single item?

PhrosTT• 15 years ago

The items are limited to 64kb so its implicit.

Johnyma22• 15 years ago

I'm too scared of the lock in. Afaik this isn't like EC3 where you can move a VM off the EC3 stack. The data is stored in a proprietary format so once your in, you're in.

Game changer? I need to see some evidence first on how well it performs and integrates.

weavejester2• 15 years ago

It's a database, so you can always get your data out again.

justinsb8• 15 years ago

It's a non-standard (NoSQL) database. Even SQL databases, despite years of standardization and efforts at poaching each other's customers, still have rough edges that make moving between different SQL database products non-trivial. Just because you can get your data out doesn't mean you're not locked in to all of SimpleDB/DynamoDB's quirks, of which there will be a lot more because it's not following a standard approach. Your code will have to go through contortions to work around / adapt to the SimpleDB limitations; effort which might well be wasted or counter-productive on a different system. That's the lock-in with AWS in general: API lock-in, not data format lock-in.

latch4• 15 years ago

I think this is huge. My first wish, before secondary indexes, is that they don't round up to the nearest 1KB for pricing.

werner7• 15 years ago

noted. (both requests that is).

mark_l_watson2• 15 years ago

A feature request (I see that Werner is reading this message thread): for development it woud be very nice to have the Java AWS SDK have a local emulation mode for SimpleDB and DynamoDB. This would allow development while flying or otherwise off the Internet. Similar functionality as AppEngine's local dev mode.

icey• 15 years ago

Since HN doesn't show vote totals I wanted to chime in and say this would be really useful to me as well.

viggity• 15 years ago

HN doesn't show vote totals, but it does sort posts based on the number of votes (descending), so the more popular posts rise to the top, just like the front page.

werner7• 15 years ago

It is on the requested features list

floodfx• 15 years ago

one of our developers built a local emulated simpledb clone for development/testing that is open sourced at:

https://github.com/stephenh/fakesdb

encoderer• 15 years ago

I work for a popular startup that has been privately testing dynamo for the last few months.

It's a fantastic product and even while in private beta has been stable and well supported.

lemming2• 15 years ago

This looks like a great service - there are some really interesting ideas here. The provisioned throughput is a great idea. It's interesting to see it runs on SSDs as well, I wonder if its storage is log based like LevelDB. The composite hashed indexes are really interesting as well - I guess they partition the index across nodes, and each node maintains a range of the sorted index. It'll be interesting to see how usable that is in practise.

I read with interest Steve Yegge's mistakenly-leaked email about the oppressive conditions for engineers at Amazon. It's hard to reconcile with the sort of innovation they consistently show.

jsolson2• 15 years ago

> I read with interest Steve Yegge's mistakenly-leaked email about the oppressive conditions for engineers at Amazon. It's hard to reconcile with the sort of innovation they consistently show.

My take on Steve's rant (as an engineer at Amazon) is that a lot of the issues he pointed out are legitimate, but at an entirely different scale than he was pitching them at.

Day to day I work on a product with another couple dozen or so engineers. We build what makes sense for our product, and for the most part we build it in a way that makes sense for us. Sometimes we are under pressure to leverage other parts of the platform, and sometimes that does entail a lot more work. Most of the time, though, it ends up reducing our operational load (because the systems we depend on support products much larger than ours :) and giving us someone we can page when things go pear-shaped.

Amazon isn't the perfect place to work, but it's generally not bad (other than the frugality thing; that sucks as an employee no matter which way you slice it).

lemming2• 15 years ago

Interesting, thanks for the perspective - he definitely made it sound like a sweatshop, and that's not the sort of environment normally associated with this kind of innovation.

justinsb8• 15 years ago

Very interesting that there's no mention of the CAP theorem here, despite Amazon's heavy reliance on it in the past when marketing their non-relational stuff.

chubot• 15 years ago

I'm very curious about this too. I posted in another thread:

"Doesn't that just sweep the latency tradeoff under the rug, or is flash making up the difference? What about the availability tradeoff? (I like the formulation here, consistency vs availability and consistency vs latency, as opposed to CAP which never made sense as a 3-way tradeoff: http://dbmsmusings.blogspot.com/2010/04/problems-with-cap-an...)

http://news.ycombinator.com/item?id=3480805

leef• 15 years ago

The CAP theorem is boiled down to you choosing whether or not you want consistent reads.

taylorbuley• 15 years ago

Pricing: $0.01 per hour for every 10 units of Write Capacity and $0.01 per hour for every 50 units of Read Capacity.

justinsb8• 15 years ago

It is amusing that was positioned to try to address SimpleDB's problem of "pricing complexity". These aren't those complicated "Machine Hours", they're "Capacity Units"!

As someone who has grappled in the past and is grappling again with the issues of pricing a database-as-a-service though, this is very much a non-trivial issue. If you have nice scaling characteristics and you want to charge the minimal price, your price isn't going to be predictable. Essentially, what Amazon have done here is to set a cap and then charge you the cap irrespective of actual usage, which seems to be the model that's winning on the cloud.

snewman4• 15 years ago

Capacity Units strike me as a big improvement (for the user) over Machine Hours, because it's very clear how a given usage pattern will translate into Capacity Units. I can predict how many Capacity Units I'll need. I've gotten badly burned over seemingly simple queries using unexpectedly high Machine Hours in SimpleDB.

nirvana2• 15 years ago

Trying to read thru all the hype, this looks like Riak hosted on machines with SSDs, with less features, and a nice billing system in front of it.

Of course for people who want a hosted solution, the fact that it is hosted, is what gives it a lot of value. There haven't been a lot of hosted NoSQL databases, at least on this scale and availability, out there.

But technologically, what's new here? Is there anything here that is really innovative? (not a sarcastic question)

PanMan3• 15 years ago

For one, it responds quicker than Riak: Riak has (cold) response times of about 300ms, while this service claims single-digit ms response times. Also setting up Riak is not exactly trivial, and using this service outsources that hassle.

samstokes3• 15 years ago

Could you clarify what you mean by "cold response time of 300ms"? Cold as in requesting data that hasn't yet been cached in RAM? How good does it get once the cache is warm?

nosequel3• 15 years ago

You can use Riak Smartmachines on Joyent's cloud that would get similar performance for an order of magnitude cheaper than what amazon is charging. If you are seeing 300ms response times, you are not using SSD's and you are not using a similar number of nodes that Amazon is charging you for.

I'm not hating on Amazon, it is a good move for them and they are doing some things that Riak cannot do, but cost and response is not one of them.

aphyr• 15 years ago

Riak has many shortcomings, but I wouldn't describe latency or installation as primary concerns. Our cold response times have a 99% bound of 8ms, and median of 5ms on commodity SSDs. Installation is handled by apt and is trivial to automate.

champion• 15 years ago

Also it is spread across AZs which is handy.

snewman4• 15 years ago

As you and others have pointed out, hosting + SSDs + synchronous replication across availability zones counts for a lot. If DynamoDB lives up to the hype, it could be a huge step forward in the world of "don't have to think about it" data storage.

DynamoDB does have at least one significant feature not provided by Riak -- range scans. This makes many common access patterns much easier to implement efficiently. Still, as you suggest, there don't appear to be any fundamental technical advances here. The advances are in the service model and operation.

And there are, of course, many limitations. (Just to name a few: items -- i.e. rows -- can't exceed 64K; queries don't use consistent reads; seemingly no atomic update of multiple items.) It's miles ahead of SimpleDB, but still not nearly as flexible as many of the existing NoSQL databases. If Amazon lives up to past performance, they'll make steady improvements, but slowly.

nirvana2• 15 years ago

>DynamoDB does have at least one significant feature not provided by Riak -- range scans.

Riak has the ability to select keys for processing via various queries, including range of the key.

Riak also has secondary indexes and full text search.

If there's something significant about the DynamoDB method of doing range scans I'm interested in hearing it. My purpose here isn't so much to bash DynamoDB (in fact, I don't want to do that at all) but to try and spread a little more awareness of Riak.

Riak really came into its own in 1.0.

werner7• 15 years ago

you can choose whether queries are consistent or eventually consistent.

jondot• 15 years ago

If anything, from my experience with Riak, Basho guys should be having an emergency meeting.

Not mentioned here - DynamoDB also has built in monitoring and management.

StavrosK• 15 years ago

What has been your experience with Riak?

count• 15 years ago

Riak is based on Dynamo, and the original paper by Werner (if I recall correctly?). This is just offering up an easy to use cloud service version of what's backing S3 already?

ed209• 15 years ago

With regard to reliability, how safe would a service like this be to store your data? They mention "back up to s3" but this sounds more like archiving. I'm wondering about backups in the case of a problem, or is the data so spread across all of their physical locations that data loss impossible?

ww5202• 15 years ago

SSD are pretty safe in general. Amazon also makes replicate of your data. It's the same if not safer than other NoSQL solution. Definitely safer than those RAM-based NoSQL approaches.

The "back up to s3" part is for archiving data, like periodic snapshot backup of your data so that you can get back the old data in case they are deleted by app/human.

Ataraxy2• 15 years ago

This still seems a bit expensive to me for an application that would require thousands of writes per second? ie. 5k writes per second is ~$120/day. Using this for performance based analytics for example would seem out of the realm of reason for the moment.

peripitea• 15 years ago

Can you explain your use case a bit more? I'm having a hard time imagining something that does ~430M DB writes/day but can't easily afford to pay $120 for those writes.

Yrlec2• 15 years ago

Remember that the throughput is per item, not per query. For instance we have an indexed query that returns ~1500 rows each time. Just doing that query a couple of times per second would create that kind of throughput requirement.

Ataraxy2• 15 years ago

A small mobile marketing company jumping into the wild wild west of real time bidding. It would be used more so for logging impression requests to be used later for further analysis. Our bidder would need to be able to handle upwards of 5000 bid requests per second. Though these requests can be throttled down, naturally the more data we can collect the better. This also doesn't include the associated costs with querying the data which would end up adding up quickly.

Now I'm not sure this would be the ideal solution for such a thing (in fact it probably is not), but it's just the first thing that came to mind. In the grand scheme of things sure that may seem like a trivial amount due to the use case, but we're still more in the realm of a startup where dropping ~$3k/month on the data store alone makes me cringe a little when we have other expenses to account for also. :)

jbellis• 15 years ago

I posted a comparison to Cassandra here: http://news.ycombinator.com/item?id=3480480

simonw• 15 years ago

I hope this means an SSD option for RDS is coming soon.

nosequel3• 15 years ago

https://img.skitch.com/20120118-etwh2gwu52isnn47fw2jpcir5p.p...

You could buy a lot of Riak or Cassie for that.

werner7• 15 years ago

5000 reads per sec of 64Kb items, would make you stream 2.5 Gbits/sec using consistent reads and 1Gbits/sec writes, moving close to 1.5TB each hour. At the end of the month you have read well over 800 TB and updates 160 TB... That is a substantial application you have in mind... :-)

tedivm• 15 years ago

That may be true for an application with a constant load, but applications with a less balanced load have to provision for their peaks. My company (Malwarebytes) has very irregular traffic (at the hour mark we get very big spikes, but only for a couple of minutes) and it seems like we would have to provision (for this specific app) that peak for the entire hour. I might be misunderstanding the billing for this service though- if we ask for more units for 15 minutes, would the billing be prorated?

This actually hits on my only real issue with AWS in general, which is the hourly billing. We've used the mapreduce service a bit, and having a cluster roll up and fail on the first job is heartbreaking when it has a hundred machine hours associated with it. Obviously that is far, far cheaper than us building out our own cluster (especially with spot instances, which I can't even describe how much I love), but for some of the services smaller billing times would be useful.

Loic• 15 years ago

1$/GB/month on SSD and replicated. So basically, 0.25$/rawGB/month if they replicate 4 times.

They are making money on the read/write and are selling the capacity at current cost. Which — knowing the tendency for AWS to decrease the prices very slowly combined with the huge decrease of the prices of the SSD drives in the past months/years — is not a bad strategy to convince us to switch.

meta• 15 years ago

I read through a number of the docs and can't quite find the answer to this question, hopefully someone here can help me out quick.

I already have a bunch of (large-ish, deeply nested) JSON objects defined for my application. I don't really want to go about redefining these since they work great between my various node processes and the front end. I am saving them in a nosql database already, I am curious about switching (to save on devops costs). I only request based on 1 Hash Key (int) and 1 Range Key (int) for all my current get operations.

Looking through the docs/examples I see a lot of this type of thing:

    {"TableName":"comp5",
    	"Item":
    		{"time":{"N":"300"},
    		"feeling":{"S":"not surprised"},
    		"user":{"S":"Riley"}
    		},
    	"Expected":
    		{"feeling":{"Value":{"S":"surprised"},"Exists":true}}
    	"ReturnValues":"ALL_OLD"
    }

The JSON item has a kinda-of 'type syntax' on it. I really don't want to redefine my deep objects, but would be willing to redefine the Hash key and Range key, while leaving the rest of the nested types alone.

Ok, my question: Do my JSON objects need to conform to this 'type syntax' JSON notation in the examples? Or can I save just any JSON object into this database and only annotate the Hash Key and Range Key using this special notation?

n_are_q2• 15 years ago

You will have to create an attribute for your json, where you'll store the json utf-8 encoded. If you want to index on parts of that json blob you'll have to pull them out into their own separate attributes and the recombine them into a single json object on read.

ww5202• 15 years ago

Their usage of JSON is just incidental to your usage of JSON. They use JSON as a REST transfer format. You can pretty much ignore their JSON if you use one of the high level libraries in the SDK.

You can define a table with 3 fields: yourKey, yourRange, and yourJson. Put your entire JSON data as string in the yourJson field.

hiroprot• 15 years ago

Is there a Ruby SDK?

ConstantineXVI2• 15 years ago

Yes (which I find mildly surprising given how long the Ruby library took to support batch ops on SQS)

https://github.com/amazonwebservices/aws-sdk-for-ruby

n_are_q2• 15 years ago

Any word on when boto will support this?

ConstantineXVI2• 15 years ago

Sooner than you think

https://github.com/boto/boto/commit/8ef550ec028b379bcd85e550...

joelreymont• 15 years ago

I wrote an Erlang API for DDB while it was still private. I'm hoping my client will let me opensource it in the near future.

andrew311• 15 years ago

Question about Composite Hash Keys that someone might have the answer to (or be able to relate to other known implementations):

The composite key has two attributes, a “hash attribute” and a “range attribute.” You can do a range query within records that have the same hash attribute.

It would obviously be untenable if they spread records with the same hash attribute across many servers. You'd have a scatter-gather issue. Range queries would need to query all servers and pop the min from each until it's done, and that significantly taxes network IO.

This implies that they try to keep the number of servers that host records for the same hash attribute to a minimum. Consequently, if you store too many documents with the same hash attribute, wouldn't you overburden that subset of servers, and thus see performance degradation?

Azure has similar functionality for their table service, requiring a partition key, and they explicitly say that there is a throughput limit for records with the same partition key. I haven't seen similar language from Amazon.

Whether you scatter-gather or try to cluster values to a small set of servers, you'll eventually degrade in performance. Does anyone have insight into Amazon's implementation?

Amazon DynamoDB – a Fast and Scalable NoSQL Database Service from AWS

Related Stories