Oracle rewrites 'the brain' of its database to take on SAP Oracle's future is its rival's past, but the database giant isn't worried because the changes it has made to let its database store data in speedy consumer memory afford far more backwards compatibility than does SAP HANA. With the launch of the Oracle Database In-Memory tech on Tuesday, the world's most influential data- …
SAP ASE, or Sybase ASE as was, has had in memory databases for some years. They run standard SQL so no changes to the code and you can mix and match different tables or parts of tables into RAM as you wish. You can also use a template DB (with data if wanted) and choose your recoverability.
Hmmm, how about this?
Run a complicated sql update query through MySQL and it might do something entirely unrelated to what you were asking. That was my experience, with 5.0x. The funny thing, I even had a missing parenthesis, so not only did MySQL run in the weeds, it also ran a syntaxically invalid query. It just got confused, poor thing. Started moving the whole app to Postgres the next day.
Friend of mine calls me up for his ecommerce website. Guess what? Despite his best efforts, read locks & write locks, he kept on seeing the occasional duplicate order numbers coming out of an incremented ordernum column, when two people ordered at exactly the same time. His solution, after my advice also failed? Used PHP to lock a file on the file system when processing an order and releasing it to allow the next db read. Nice. Less than two years ago, with a current MySQL.
For someone who works with complicated sql day in, day out, not something I'd recommend. Loved their initial casual approach to ACID as well - "you know, ACID's just not that important, most of the time".
Say what you like about Oracle, Larry's boats, Java bugs, predatory pricing, Fusion shenanigans, lawyer-heaviness, bloated Java stacks.
But their database kit is top-notch and serious. MySQL is A-OK too, as long as you use for what it was intended to do at first, which is low-fuss, possibly high volume, simple, CRUD for websites.
For anything more serious, I'll take Postgres in OSS. And probably Oracle if money was NO object whatsoever.
Asides from monetary considerations, I'll tip my hat to Oracle on this one. Innovations in SQL should be transparent to applications, as much as possible. And if possible, it should minimize impact to the DBAs as well. We'll see if reality lives up to the hype.
So far it looks like a winner. In a month when we get our grubby hands on it and kick the tires we should know for sure.
If I can avoid building and maintaining indexes and stats on billion+ row tables, and even get the same performance (let alone quicker) - it's a clear winner.
Hana's huge pitfall is HA and the cost of nodes. Don't want to have to wait (a long long time) while a node reconstructs and loads the data for a failed node, you have to duplicate the entire array as a mirror - at a huge $$ cost. Oracle is very expensive, but Hana makes it look like bargain-bin software by comparison.
"...Hana's huge pitfall is HA and the cost of nodes..."
Hana is using a distributed cluster for their in RAM memory database, a scale out architecture. There will be problems as you have to spread everything across all nodes. Latency will suffer as each node has to talk to others across the network.
Oracle is the only vendor with a 32TB RAM server on the market, it is called the SPARC M6-32. That much RAM in a single server beats any cluster, especially if you compress the data. If you can not manufacture a huge server, you cheat and create a cluster instead. Which is what Hana has done. Oracle huge RAM servers will beat any cluster. Of course, nothing prevents you from creating a cluster with huge Oracle servers too. :)
BTW, Oracle regained no 1 spot in Unix server market:
https://blogs.oracle.com/partnertech/entry/oracle_regains_the_1_unix
IBM is falling behind.
That's the web syndrome for you, ever less revenue, so less money for sub editors, so lower quality, so less revenue...
ZDNet is getting hilarious, many of the "columnists" there seem to rely on tablets and smartphones autocompleting every word for them and the text makes no sense at all!
In comparison, it looks like El Reg is still forcing their staff to use real keyboards and actually write the text themselves, so fewer mistakes, but still the odd funny. :-P
Way back then it was "very large memory database" VLMDB on DEC Alpha with literally HUNDREDS of megabytes of memory, With VLMDB and Oracle Parallel Server OPS, they claimed to eclipse everyone, except somehow it was slower than and less reliable than (errm) a mainframe. Given that Oracle is the company to add 'REAL' to its cluster tech (bought from DEC) so customers would know that RAC was not the same as old OPS, you've got to wonder with it is Ellison guff.
So what have they done for all those "legacy" oracle apps that rely on ROWID?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
