Java – ideas for an alternative to ORM RDBMS?

I am currently developing proof of concept for alternative data storage The reason why I need to enhance a webapp that mainly reads clusters is also because I want to get rid of the pain of sometimes over complex ORM + RDBMS solution

Overall, this idea is very similar to the distributed cache with persistence (making the cluster SOR), but:

>I want to be able to retrieve any object and its sub object ID (provide class & ID) [only at the beginning, as the main query part, Lucene has been used in my application]. > A mapping with type mapping (table world in relational table) is required, in which maps of 'dehydrated' storage objects are distributed (flattening object graph by reflection depth cloning) > bin logs (e.g. previyler)

>If the entire cluster fails, it will eventually recover > development (and the ability to refactor code / change structure) > possibly asynchronous processing for other purposes (reporting, etc.)

>Finally, try to integrate static query mechanisms, such as Jaqu of LINQ, Jaque or H2 / see odbs / Lucene (?) > It must be transaction aware (but not "JTA type")

I'm going to implement this idea with hazelcast (I like its super simple API) or terracotta (I've never used it - but I know their 'sweet spot', interim data) If you like, my goal is to do more or less what Jonas once blogged about Using one of them, the stored data must roughly fit the sum of the JVM heap of the cluster

This should be very simple to avoid relationship impedance mismatch (i.e. save as ODBC) and JDBC I / O overhead

Are you aware of other tools / frameworks or their combinations that already provide similar functionality, which I ignored? Can you suggest other ways to solve the problem of "getting rid of database"? What flaws have you seen in this idea? Does concurrency make sense to consider Scala rather than Java?

How about non relational data storage such as couch dB, neo4j, hypertable and HBase?

I asked the similar question a month ago – but there was no specific solution

By the way, I stumbled upon the concept of enterprise data fabric. To my surprise, it describes many of these ideas

Solution

Be sure to try terracotta It's free (unless you go to an enterprise with SLA and support) It is a JVM level cluster, so to speak, so you won't encounter problems related to sessions on multiple boxes behind different JK workers (assuming you use it for J2EE applications)

I'm just walking, so look here: http://en.wikipedia.org/wiki/Terracotta_Cluster

Update a lot of information about terracotta on the Internet, such as http://blog.terracottatech.com/2007/12/fud_of_the_week_terracotta_doe.html

Background: I work for a company with a large audience We have an enterprise MySQL running the master server and about 5 slave servers (considering that we have 2 channels and 4 application servers per channel), and use the JDBC replication driver of MySQL (we have submitted various patches) We use spring's declarative JTA transaction management to use spring 2 5 / Hibernate 3, so read onlies will go to slave With the advent of many Ajax enhancements in future versions of our website, the load on our database server has increased – we create pricing summaries for thousands of products in all countries, taking into account the tariff / tax rules (and promotions) and real-time auctions in all these countries, and then the Ajax service has the latest prices in the blink of an eye Terracotta solves the burden of database and application servers by providing these prices to all application servers on the JVM layer, and all JVMs are connected together Therefore, server a can update the price every few minutes. If Ajax clicks server B, the price is available immediately I know that some people / companies have similar businesses and may have better ideas and implementation, so I am always open to discussion, but this is my two cents

I also get inspiration from people on Facebook, such as this very rich article: http://www.facebook.com/note.php?note_id=23844338919

They talk about memcached, you should also check it