DocumentDB

I wonder what the search story is. One technology that really does deliver, and has totally impressed me, is Lucene&#x2F;ElasticSearch. I&#x27;m used to all sorts of hyperbolic claims, but holy shit, ElasticSearch just delivers. We tossed in about 40M documents from a SQL Server DB, and not only did it require less resources (a 30%? reduction in size), the queries are beyond anything that&#x27;d be approachable using SQL Server. And I&#x27;ve only touched the surface, using it as a pure plug-n-play setup.<p>With DocumentDB, not having a local version severely limits what I&#x27;d consider this for. Losing that flexibility is a big deal. Maybe this is just a limited preview and they haven&#x27;t build the management side for local installs.

<a href="http://azure.microsoft.com/en-us/documentation/articles/documentdb-introduction/" rel="nofollow">http:&#x2F;&#x2F;azure.microsoft.com&#x2F;en-us&#x2F;documentation&#x2F;articles&#x2F;docu...</a><p>&gt; Want to edit or suggest changes to this content? You can edit and submit changes to this article using GitHub.<p>Pretty remarkable given Microsoft&#x27;s approach to open source in the 1990s that they&#x27;re now using a service built around Linus&#x27;s bespoke open source version control system to allow people to suggest changes to their documentation.

Ad hoc queries using SQL like syntax. No need to define indexes.<p>Javascript execution within database. Stored procedures, triggers and functions can be written with Javascript. &quot;All JavaScript logic is executed within an ambient ACID transaction with snapshot isolation. During the course of its execution, if the JavaScript throws an exception, then the entire transaction is aborted.&quot;<p>Pricing is based on &quot;capacity units&quot;. Starts with $22.50 per month (this includes 50% preview period discount). One capacity unit (CU) gives 10GB of storage and can perform 2000 reads per second, 500 insert&#x2F;replace&#x2F;delete, 1000 simple queries returning one doc.<p>In order to see pricing details, change the region to &quot;US West&quot;: <a href="http://azure.microsoft.com/en-us/pricing/details/documentdb/" rel="nofollow">http:&#x2F;&#x2F;azure.microsoft.com&#x2F;en-us&#x2F;pricing&#x2F;details&#x2F;documentdb&#x2F;</a><p>Very interesting addition to Microsoft offering. I was actually just yesterday wondering if they have any plans for this kind of service. Table Storage is quite primitive and Azure SQL on the other hand gets expensive when you have lots of data.<p>One potential &quot;problem&quot; with this is the bundling of storage capacity and processing power. If I understand this correctly, I would need to buy 10 CUs per month to store 100GB of data even if I&#x27;m not very actively using that data.

<i>DocumentDB utilizes a highly concurrent, lock free, log structured indexing technology to automatically index all document content. This enables rich real-time queries without the need to specify schema hints, secondary indexes or views.</i><p>How does that work? Isn&#x27;t that going to incur a major performance hit? If not, why don&#x27;t other databases get rid of indexes?<p>Also, if anyone from MS is reading, <a href="http://azure.microsoft.com/en-us/documentation/articles/documentdb-introduction/" rel="nofollow">http:&#x2F;&#x2F;azure.microsoft.com&#x2F;en-us&#x2F;documentation&#x2F;articles&#x2F;docu...</a> links to <a href="http://azure.microsoft.com/en-us/documentation/articles/documentdb-concepts/" rel="nofollow">http:&#x2F;&#x2F;azure.microsoft.com&#x2F;en-us&#x2F;documentation&#x2F;articles&#x2F;docu...</a> which is a 404 error.

No local installation. No banana.<p>I wouldn&#x27;t tie a product to a single cloud vendor.

...and MS goes after MongoDB. It would be nice to see an on-premises version, if only to compare performance&#x2F;consistency with Mongo.

I liked everything about it until I saw the API for the Python client. What a catastrophe.<p>I pray Microsoft is looking for Python developers: <a href="https://gist.github.com/whalesalad/2142f0075c6896f4547c" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;whalesalad&#x2F;2142f0075c6896f4547c</a>

&quot;All JavaScript logic is executed within an ambient ACID transaction with snapshot isolation. During the course of its execution, if the JavaScript throws an exception, then the entire transaction is aborted.&quot;<p>Have I missed something, or have MS delivered a novel and valuable feature? I&#x27;m not aware of support for transactions across documents in other NoSQL platforms. I&#x27;d be grateful if someone has any experience or better information in that regard, thanks.

A quick comparison between DocumentDB vs MongoDB: <a href="http://daprlabs.com/blog/blog/2014/08/22/azure-documentdb/" rel="nofollow">http:&#x2F;&#x2F;daprlabs.com&#x2F;blog&#x2F;blog&#x2F;2014&#x2F;08&#x2F;22&#x2F;azure-documentdb&#x2F;</a>

If I understand correctly, their multi-document ACID transaction support is a big deal. The only other NoSQL&#x2F;NewSQL systems I&#x27;m aware of with that ability are FoundationDB and Google Spanner&#x2F;F1.

Sounds very similar to CouchDB. Server side Javascript written by the user, and an HTTP interface. The ability to adjust consistency is really neat.

What are the limits of DocumentDB? You know, like max size of database, max size of document, max number of documents per database, max. number of attributes per document, max. number of databases per DocumentDB account.<p>What&#x27;s the max. duration of database query, max size of query result.<p>What kind of performance can be expected, does it decrease as the size of database increases or it remains constant?<p>I&#x27;m going to wait a few days until hype settles.

And that is a creative product name. Well played MS.

What are the size limits on a collection? Docs mention transaction support is offered only within a collection. Is a collection essentially limited to a single physical machine in the background or does it span across machines? It looks like in Standard Preview, the max collection size is 10GB.

Interesting: <a href="https://github.com/Azure/azure-documentdb-python" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;Azure&#x2F;azure-documentdb-python</a> (it&#x27;s empty for the moment, but glad to see first-party support for Python)

The @DocumentDB twitter links to a tutorial on DocDB: <a href="http://www.documentdb.com/sql/tutorial" rel="nofollow">http:&#x2F;&#x2F;www.documentdb.com&#x2F;sql&#x2F;tutorial</a>

Does anyone know how this compares to AWS DynamoDB[1] ?<p>[1] <a href="https://aws.amazon.com/dynamodb/" rel="nofollow">https:&#x2F;&#x2F;aws.amazon.com&#x2F;dynamodb&#x2F;</a>

Am I alone in thinking that sql-like syntax is actually a step backwards from building query documents programmatically (MongoDB style)?

Spatial queries and indexing. Most data has some location component. I didn&#x27;t see anything with this. Is it in there, or planned?

So does it run on a cluster? If so which of Consistency, Availability and Partition tolerance does it NOT offer? (See CAP theorem)?

It would have been nice to see some actual details of how it works so that it can be compared to the competition.

I wonder if this is built on JetDB

Another case of Not Invented Here syndrome from Microsoft. One wonders why they couldn&#x27;t just take the open source and very well architected RavenDB <a href="http://ravendb.net" rel="nofollow">http:&#x2F;&#x2F;ravendb.net</a> .Net Document DB and provide first class support for that within Azure.

I find it surprising that DocumentDB wasn&#x27;t already a copyrighted name. ;)

Can I use DocumentDB out of Azure (hook my own)?

Leave it to Microsoft to give it the most generic sounding name possible.

Sick! Love it

&#x27;cause what the world needs is another proprietary NoSQL solution.

THIS THING IS A BEAST!! It is absolutely bad ass

A new &quot;cool&quot;, locked-in service served on a silver platter by Microsoft to the brainwashed.<p>Everybody else uses open source on premises or their cloud of choice.