The origin of Oracle NoSQL database

Previously NoSQL databases were indifferent to new and emerging projects around the world, and we often heard those with ambitious trumpeter patting their chests to ensure that their new nosql applications would break all the old frames and bring new things. In reality, however, it is overblown that NoSQL is still unable to get into Wall Street, even if the trendy developers are only willing to use it for trivial matters in their lives. Because of the limitations of the table structure is too large, if you can discard these, the speed of the database can be improved quality.

At this point, Oracle, the expert who has consistently developed the top-tier bulletproof SQL database, has brought its NoSQL database server, rugged, practical, and standard Oracle style. Crazy Dreamers will continue to write nosql new projects, but serious people will pay close attention to Oracle's products, because it brings not only the exciting features of NoSQL, but also the huge scale of rigorous large-scale projects.

The announcement of Oracle NoSQL database products may surprise NoSQL fans, but Oracle has been quietly on the road for some time. Five years ago, Oracle bought Sleepycat Software, the start-up of Open source Berkeley db, Berkeley DB was famous for its flexible key-value storage model, Berkeley DB technology is said to be the core of Oracle NoSQL database, although it seems to be completely rewritten.

Oracle NoSQL: Complete acid

The most interesting place in Oracle NoSQL is the key-value structure. You don't have to define an outline or lock yourself in a table schema, just create keywords and then associate the data with them. You can connect a string to the keyword, or you can connect to the previous image file, anything, the database accepts bytecode, regardless of what the content is.

Oracle divides the keywords into primary and secondary two parts, you can think the main part is the object's pointer, the secondary part is the various fields of the record. For example, you can put your name and Social Security card number in the main section, put your address and zip code, and other strings in the secondary section, which is different from the way some NoSQL tools use multiple fields of one object.

An important part of Oracle NoSQL is an approximate project for acid compliance, which allows Oracle NoSQL to meet the stringent standards that SQL databases can provide. Acid means "atomic,consistent,isolated,durable transactions", which refers to a tight, persistent, segregated and durable deal. Most NoSQL systems go the other way: base, "basically Available, Soft state, and eventually consistent", which is to ensure basic usability, flexible transaction processing, and ensure the consistency of final results. In other words, the emphasis on results does not emphasize the process.

There is a lot of debate about whether Oracle NoSQL really provides acid compliance, but Oracle NoSQL can make that commitment.

Final result consistency: controversy

Here's a question to ask, said Daniel Abadi, a professor of computer science at Yale University, who questioned in some cases that Oracle's NoSQL to the primary server would be lost, for example, if the primary server was down and the replication server was not ready. Soon Margo Seltzer, a professor of computer science at Harvard University (who we know is now an Oracle employee, who was involved in the creation of Sleepycat), commented that this is not a question of Oracle NoSQL, if it is to achieve "final consistency" in real sense, Data centers need to start writing data only if they are ready to replicate the server. It is clear that the result of this debate is likely to be more difficult than the annual football match between the two universities.

Of course, these arguments are not my interest, in order to test the speed of Oracel NoSQL, I deliberately found a low-end Mac configuration. I started a single point NoSQL server, then put 358,400 keywords into it, all of which are about 30 of the length of the string. On this old Mac, Oracel NoSQL for 119 seconds.

To contrast, I insert the same record into the latest version of the Voldermort database, which is 180 seconds on the open source Java NoSQL database that the LinkedIn symptom uses.

I'm very happy with this simple test result, Oracel NoSQL seems to be leading a lot. Creating a keyword requires an array of strings, and the instantiation of the object often becomes a Java bottleneck, and in this test Oracel NoSQL does not seem to have encountered this problem.

In general, Oracle NoSQL is worth a try. Because it provides so many rigorous functions, but also from such a rigorous data management vendors.