Friday Nov 20, 2015

Questions You Asked: What happens to a table In-Memory if one of my RAC nodes goes down?

We’ve written a number of blog posts on how Database In-Memory behaves in a RAC environment but recently we’ve gotten a lot of questions regarding what happens if one of the RAC nodes should fail. So, I thought I would try tackle this question and point out a couple of other interesting aspects of running Database In-Memory on RAC in this post.

Quick Recap

If you recall from part 1 of our RAC series each RAC node has it’s own In-Memory column store (IM column store). When a table is populated into memory in a RAC environment it will be distributed across all of the IM column stores in the cluster. That is to say, a piece of the table will appear in each RAC node.

Let’s take a look at an example using the LINEORDER table, which has 5 million rows in it and is approximately 550MB in size on my 3 node RAC cluster.

[Read More]

Thursday Oct 29, 2015

Day 5 of Oracle OpenWorld 2015 October 29th

It's the last day of Oracle OpenWorld and we have saved the very best for last. So hopefully you are still awake and functioning after the party!

Today, we have a fantastic line up of In-Memory sessions, so grab your coffee and start your last day at our hands-on-lab Oracle Database In-Memory Bootcamp (Session HOL10483) at the slightly later time of 9:30am in the Golden Gate room (25th Floor) at the Hotel Nikko.

If you have already attended our hands on lab, check out our last Database In-Memory technical session Introducing Analytic Views—Bringing Star Queries into the Twenty-First Century (Session CON8742) at 9:30am in Moscone South - room 309. During this session Bud Endress will introduce Analytic Views and describe how they easily transform tables into the next-generation data source for BI applications and Excel Pivot Tables.

Before you head for the airport, don't forget to stop by the Engineered System Showcase to check out the Database In-Memory demo and pick up both an In-Memory bumper sticker and a SQL in Silicon sticker.

We hope you have a safe trip home and we will see you all again next year, when we will have even more great session on Oracle Database In-Memory!

Wednesday Oct 28, 2015

Day 4 of Oracle OpenWorld 2015 October 28th

Thanks to all those who stopped by the demogrounds to chat with the In-Memory developers over the last two days. Remember, today is the last day of the demogrounds, so if you haven't had a chance to stop by yet, do so today. The In-Memory developers will be there from 10:00am until 4pm. This may be your last chance to snag an In-Memory bumper sticker, so don't miss out!

But before you head to the demogrounds why not start your day at our hands-on-lab Oracle Database In-Memory Bootcamp (Session HOL10483) in the Golden Gate room (25th Floor) at the Hotel Nikko. The fun begins bright and early at 8:45 am!

There are also some great technical sessions on today, including your first opportunity to ask questions of production Oracle Database In-Memory customers and see if the technology lives up to the hype. I will be hosting our first Oracle Database In-Memory Customer Panel (Session CON8657) at 1:45pm in Moscone South - room 103. The session will begin with each customer briefly sharing how they are using Oracle Database In-Memory before we dive into an open discussion. I strongly encouraged you to actively participate in this session and pose any questions you may have about Database In-Memory and other alternative solutions considered by the panelists.

Then in the afternoon we have our last customer session on Database In-Memory given by General Mills, How Exadata & Database In-Memory Power Precision Marketing at General Mills (Session CON4292) at 3:00pm in Moscone South - room 254. Mark and Josh from General Mills explain how Oracle Exadata with Oracle Database In-Memory gave them a scalable and reliable platform from which to build a consumer-first data warehouse.

Finally before you rush off to put on your dancing shoes for the appreciation event, we have another great technical session, Oracle Database In-Memory: Under the Hood (Session CON8652) at 4:15pm in Moscone South - room 103 where Tirthankar Lahiri, the VP of Database In-Memory development, will provide an insiders view of the technology underlying Oracle Database In-Memory and give you a sneak peek at what you can expect from Oracle Database In-Memory in our next generation!

Tuesday Oct 27, 2015

Day 3 of Oracle OpenWorld 2015 October 27th

Hopefully you enjoyed yesterday, the first full day of technical sessions at Oracle OpenWorld, and are ready for more today!

First thing this morning at 8:45 am is your second opportunity to attend our hands-on-lab, Oracle Database In-Memory Bootcamp (Session HOL10483) in the Golden Gate room (25th Floor) at the Hotel Nikko. I know it's early but trust me it will be a great way to start your day as it will leave you inspired and energized!

Today we will also give our first two technical sessions. First up is  Best Practices for Getting Started with Oracle Database In-Memory (Session CON6775) at 11:00am, in Moscone South - room 104.

In this session, I'll provide you with a step-by-step guide on how to get started with Oracle Database In-Memory, including advice on when and where you should take advantage of Oracle Database In-Memory. I'll also outlines strategies to help ensure that you get the promised performance boost regardless of your database environment. If you do come to this session, be sure to keep an eye out for this little fellow, as he will be on top of all of the slides with the top tips!

Later in the day we have our 3rd customer session when Dan Huls from AT&T WiFi presents, Oracle Database In-Memory: A Practical Solution (Session CON3087) at 4pm in Moscone South - room 307. Dan will share with you how they implemented Oracle Database In-Memory on a large customer-facing data warehousing system in under two months with little or no changes to their application. He will also provide details on why they chose Oracle Database In-Memory versus a more expensive SAP HANA solution.

Finally Juan Loaiza, EVP of Database Development, will present Oracle Database In-Memory—What’s New and What’s Coming (Session CON8659) at 5:15pm in Moscone South— room 103. In this session Juan will explain in detail what motivated Oracle to develop this new technology and provides a clear understanding of how this unique “dual format” architecture works. He will also give you a sneak preview of what to expect next for Database In-Memory. A definite must see session!

Saturday Sep 12, 2015

Oracle Open World 2015 Time to plan your schedule!

There are only 6 weeks to go until Oracle Open World, the largest gathering of Oracle customers, partners, developers, and technology enthusiasts, which begins on October 25th in San Francisco.

Of course the In-Memory development group will be there and you will have multiple opportunities to meet up with us, in one of our technical sessions, our hands-on-labs or at the Oracle demogrounds.

This year the In-Memory team has 4 technical sessions and there are also 5 excellent customer sessions you shouldn't miss.

[Read More]

Monday Sep 07, 2015

Star Schema Challenge - Part 4

In my previous post Star Schema Challenge - Part 3, I reported the first of the in-memory performance results with 25 users querying the star schema in the In-Memory column store (IM column store). In Star Schema Challenge - Part 3.1 I did the same for the de-normalized fact table in the IM column store.  This serves as a comparison between the row format in-memory (buffer cache) and IM column store. For this comparison I stopped at 25 users because it became clear this was enough users for the row store table.  Now it is time to start working towards 100 concurrent users.

[Read More]

Friday Aug 21, 2015

Understanding new In-Memory notes in an execution plan

If you have started to play around with Database In-Memory, chances are you have been examining execution plans to see if the optimizer has chosen to use the In-Memory column store (IM column store) or not. In addition to the execution plan itself, you should also check out the NOTE section under the plan, which contains more information about how a query was executed, such as if dynamic sampling was used during the query optimization.

[Read More]

Friday Aug 14, 2015

Star Schema Challenge - Part 3.1

In my previous post (Star Schema Challenge – Part 3) I revealed the first of the Database In-Memory results, 25 users querying a 500 million row star schema with tables in the In-Memory column store(IM column store) on a commodity Intel-based server with 60 cores and 1TB of DRAM.

In this post I will reveal the first of the in-memory results for the de-normalized version of the fact table.

[Read More]

Friday Jul 17, 2015

Star Schema Challenge - Part 3

In my previous posts on the Star Schema Challenge, I established baseline results for query performance with row store tables (that is, without using Database In-Memory). A 9 dimensional, 500 million row star schema supported 25 users with a median query time of 16.9 seconds. (With a median of 16.9 seconds, there was not much point in increasing the number of users with the row store tables). The same data in a 500 million row de-normalized table supported 25 users with a median query time of 33.5 seconds per query.

Remember all of these tests are being conducted on a 60 core Intel server with commodity disk and 1 TB of DRAM.

Row Store Query

Median query performance (in seconds) of the query workload with 25 users on row store tables.

[Read More]

Friday Jul 10, 2015

Star Schema Challenge - Part 2.1

In Star Schema Challenge – Part 2 I revealed baseline results for my query workload running on a 500 million row star schema with all tables marked NO INMEMORY and In-Memory Aggregation prevented using the NO_VECTOR_TRANSFORM hint. With a median of 16.9 seconds per query (allowing for an average think time between queries of 5 seconds), I decided that 25 users were enough for a baseline.

After my first post in this series someone suggested that the best implementation might be a de-normalized table. I decided to satisfy this user’s curiosity and test a de-normalized table also, starting with NO INMEMORY test. (Vector Transform is not applicable to the de-normalized table because it does not join to another table.)

[Read More]

Friday Jun 19, 2015

Star Schema Performance Challenge - Part 2

In my previous post, Star Schema Performance Challenge – Part 1, I outlined a challenge to support 100 concurrent users querying a 9 dimensional, 500 million row star schema using a Sun X4-4 (with 60 cores, 1 TB DRAM and commodity storage).  The users, of course, expect great query performance.  The challenging part of this is 100 active users on a 60 core machine.  With far fewer or less active users this might not be so challenging, but 100 users on this size machine is a different story.

In this post I’ll talk more about the workload and share some baseline results.

[Read More]

Monday Jun 15, 2015

SQL Plan Management

Have you ever upgraded something to a new version, like your phone or your laptop, and found that while most of it worked great there were just a few things that were worse or even broken? You might have decided to restore back to the previous version because the problem(s) were just too annoying or catastrophic to ignore. Well this was one of the motivations behind the release of SQL Plan Management back in Oracle Database 11g. SQL Plan Management (SPM) provides a framework for completely transparent, controlled execution plan evolution. The main idea being that your application should see no performance regressions due to execution plan changes when you upgrade or change something in your database environment.

[Read More]

Wednesday Jun 03, 2015

Oracle Database In-Memory Bundle Patch 8 Released

As most of you know, we are releasing regular bundle patches for Oracle Database In-Memory. These bundle patches are available periodically through My Oracle Support.

The latest bundle patch to be released was bundle patch 8 or (DATABASE PATCH FOR ENGINEERED SYSTEMS AND DB IN-MEMORY More information on the latest bundle patch can be found in the MOS note, Doc ID. 20879108.8 or 1937782.1.

The bundle patches are cumulative, which means bundle patch 8 includes all fixes from bundle patch 1 to 7 in addition to new content. Don't let the name of the bundle patch fool you. You can apply this bundle patch on non-engineered systems running Oracle Database In-Memory.

Going forward, we will keep you posted when a new bundle patch is released, so you will always know what is the latest bundle patch. We recommend that you always run with the latest bundle patch.

Friday Apr 24, 2015

In-Memory Priority

I'd like to thank everyone who attended our two Hands On Lab sessions at Collaborate last week. We had a great time presenting them and we received some really good feedback. One of the questions that came up, and gets asked periodically, is how does In-Memory priority really work?

Many times people have the misconception that the In-Memory priority attribute affects more than just the order of population but that’s not the case. The only thing the priority affects is the order of population at database startup. We discussed population briefly some time ago, but I think it's worth a quick review.

[Read More]

Tuesday Mar 31, 2015

Push-Down: Making Queries Fast!

In our previous posts we discussed the basic architecture of the In-Memory column store (IM column store) and now we want to drill down into some of the unique performance enhancing features. Push-down is one of the optimizations that makes scanning columns in the IM column store very efficient. Oracle Database In-Memory has the ability to push predicates, aggregations and group-bys down into the scan of a column or columns.  This ability to push-down allows us to take advantage of other performance enhancing features of Database In-Memory like SIMD vector processing and storage indexes.

[Read More]

The Oracle Database In-Memory blog is written by the Oracle product management team and sheds light on all things In-Memory.


« November 2015