Showing posts with label In-Memory. Show all posts
Showing posts with label In-Memory. Show all posts
Wednesday, February 11, 2015
12c: Cursor wait with In-Memory and Parallel queries running
I think it is interesting that the MOS article "Significant 'Cursor: Pin S Wait On X' Waits When Using In-memory Column Store and Parallel Query" via note ID 1910787.1 is saying there is a serious bug with 12.1.0.2 where you try to query data in parallel with In-Memory options enabled. I would say the whole point of 12c is the In-Memory store, but remember that In-Memory is an option that came after 12c was introduced so really it is not that surprising that we will start seeing some of these one-off bugs start rattling out of early adopter systems!
Saturday, January 31, 2015
Weekend Learning: 12c In-Memory basics and Data Warehousing interaction
Did you miss my coverage of the 12c Database In-Memory product launch? Well tonight you are in luck, because the article "Oracle Database In-Memory Option (DBIM) Basics and Interaction with Data Warehousing Features" from note ID 1903683.1 is a great review and it has some good information to know if you want to use the option in a Data Warehouse (DW) type of situation. Will it always be faster, or be perform better than having the option turned off? Check out the article to find out!
Tuesday, June 10, 2014
12c In-Memory Launch
If you were on Twitter today, you would have seen the hashtag #DB12c being used a lot as Larry Ellison lead a launch event for the 12c In-Memory option which was announced already at Oracle Open World last year and several high profile gurus like Tom Kyte and Maria Colgan were already going to conferences spreading the word about the option.
What did today bring us?
World records as the 12c In-Memory enabled database ran on a trillion rows of data.
Application reports going from 4+ hours to 12 SECONDS.
Testers were bored because it was so easy.
Several Enkitec employees were involved in the presentation!
Do not just take my word for it, follow the hashtag link above to find the official press release and YouTube link of the replay.
What did today bring us?
World records as the 12c In-Memory enabled database ran on a trillion rows of data.
Application reports going from 4+ hours to 12 SECONDS.
Testers were bored because it was so easy.
Several Enkitec employees were involved in the presentation!
Do not just take my word for it, follow the hashtag link above to find the official press release and YouTube link of the replay.
Sunday, September 22, 2013
BREAKING NEWS: OOW13 News
We interrupt your regularly scheduled blog --
Larry Ellison has just announced a new Database In-Memory Option which can make queries 100 times faster, but better than that is the faster transaction rates such as 3 to 4 times faster row insertion. How is this being done? Oracle 12c stores data in rows and columns simultaneously with this new option so the columns (introduction of "Columnar technology") are being kept "in-memory" and transactional integrity is kept due to the fact that there are no transactions being saved back on the columns, just the rows. Cores will be able to scan billions of rows a second per vector instruction, and report results will be returned sub-second because the current OLTP architecture is being slowed down by analytic indexes which the column views will actually take advantage of.
This option makes the database easier to tune, run faster, just as secure as today but once the switch is flipped you get better performance even though nothing structurally changes when you flip the switch. A live demo of a 3 billion row table (courtesy of Wikipedia data) showed how it was more than 1000 times faster when the In-Memory Option is turned on! This allows companies to Scale-Out and Scale-Up with different types of hardware. Speaking of hardware, Larry introduced several new components:
M6-32 Big Memory Machine - This has 32 Terabytes of DRAM memory, 32 SPARC M6 chips (twice as many cores as the M5), and as he casually mentioned is available right now. Today. Go buy one! Then he elaborated on the Interconnect it uses to process 3 TeraBYTES, not TeraBITS, per second and on top of everything else the machine costs $3M which is less than a third of the cost of IBMs biggest machine. He noted that their IBM partners in the room weren't clapping. Awesome.
M6-32 SuperCluster - Just like it's brother the Big Memory Machine, it is the fastest database machine out there but the plus for this is that it connects to Exadata as integrated storage to speed things up by a factor of 10 via the Infiniband I/O Interconnect.
Oracle Database Backup, Logging, Recovery Appliance - Quite a mouthful, but Larry says he named it so I guess that's that. This doesn't act like a file backup utility that loses data, slows down the business, and isn't scalable; instead this platform which is "like a Exadata machine" ships all of the logs so it results in the ability to restore as of any point in time. Going a step further, this can all be delivered as a Cloud Service!
He closed with a snippet of what is the "Datacenter of the Future", which I think is a sneak-peak of things to come, where he details out a data center that has specialized servers for the core, backup, analytic and database functions to increase what customers can get out of their data centers.
-- We now return you to your regularly scheduled blog.
Larry Ellison has just announced a new Database In-Memory Option which can make queries 100 times faster, but better than that is the faster transaction rates such as 3 to 4 times faster row insertion. How is this being done? Oracle 12c stores data in rows and columns simultaneously with this new option so the columns (introduction of "Columnar technology") are being kept "in-memory" and transactional integrity is kept due to the fact that there are no transactions being saved back on the columns, just the rows. Cores will be able to scan billions of rows a second per vector instruction, and report results will be returned sub-second because the current OLTP architecture is being slowed down by analytic indexes which the column views will actually take advantage of.
This option makes the database easier to tune, run faster, just as secure as today but once the switch is flipped you get better performance even though nothing structurally changes when you flip the switch. A live demo of a 3 billion row table (courtesy of Wikipedia data) showed how it was more than 1000 times faster when the In-Memory Option is turned on! This allows companies to Scale-Out and Scale-Up with different types of hardware. Speaking of hardware, Larry introduced several new components:
M6-32 Big Memory Machine - This has 32 Terabytes of DRAM memory, 32 SPARC M6 chips (twice as many cores as the M5), and as he casually mentioned is available right now. Today. Go buy one! Then he elaborated on the Interconnect it uses to process 3 TeraBYTES, not TeraBITS, per second and on top of everything else the machine costs $3M which is less than a third of the cost of IBMs biggest machine. He noted that their IBM partners in the room weren't clapping. Awesome.
M6-32 SuperCluster - Just like it's brother the Big Memory Machine, it is the fastest database machine out there but the plus for this is that it connects to Exadata as integrated storage to speed things up by a factor of 10 via the Infiniband I/O Interconnect.
Oracle Database Backup, Logging, Recovery Appliance - Quite a mouthful, but Larry says he named it so I guess that's that. This doesn't act like a file backup utility that loses data, slows down the business, and isn't scalable; instead this platform which is "like a Exadata machine" ships all of the logs so it results in the ability to restore as of any point in time. Going a step further, this can all be delivered as a Cloud Service!
He closed with a snippet of what is the "Datacenter of the Future", which I think is a sneak-peak of things to come, where he details out a data center that has specialized servers for the core, backup, analytic and database functions to increase what customers can get out of their data centers.
-- We now return you to your regularly scheduled blog.
Subscribe to:
Posts (Atom)