Monthly Archives: October 2012

Oracle 12c new features

In Andy Mendelsohn’s openworld 2012 keynote presentation, he mentioned 3 key new features of the oracle database 12c. For those of you who were unable to attend the keynote and do not have the 50 minutes to watch the replay, here is the reader’s digest version of the features.

Pluggable Databases

In a pluggable database environment, you create a single database container, and plug multiple databases into this container. They key design feature here is that, all these databases then share the exact same oracle server processes (aka background processes) and memory (Unlike in the previous versions where each database got its own set of background processes and shared memory allocation).

In oracle versions upto 11gr2, when you used database resource management, you had to setup resource plans per database, and each of the database did not know about the resource utilization of other databases on the same server. So you have to use Instance Caging in order to ensure that database’s used only their allocated amount of cpu resources. In Oracle 12c, since all the databases use the same container, the container will know about the resource utilization of all the databases and hence can do the database resource management efficiently.

This lends itself well to consolidating into larger databases.

Database Heatmaps

In 12c the oracle database keep’s track of which data in your tables are being selected/updated/deleted/inserted frequently. Then the database can decide what type of compression to apply to data that has different transaction profiles. Oracle 12c will also have the ability to compress the data as per the above tracking and analysis.

Database consolidated replay

When you are consolidating multiple databases into a single database (Maybe in the oracle database machine), you can now capture workloads from multiple databases and replay them on a single target database.

This helps with consolidating databases into pluggable databases in 12c.

Andy did not forget to mention that, there are around 500 new features in 12c.

The details on how these features work, will become available, closer to when the database 12c is actually released.

Exadata Deployment Assistant

Previously, for an Oracle database machine installation, customers had to fill out the configuration worksheet, with information regarding the hostnames, ip addresses, how they want the machine configured etc. The file generated from the configuration worksheet served as the input to OneCommand.

Now there is a new utility called the “Exadata Deployment Assistant”. This is a java based, wizard driven configuration file generator, which replaces the configuration worksheet. You can get the utility by downloading the latest OneCommand Patch. You can find the latest OneCommand patch by referring to the Onecommand section of Mos Note 888828.1.

You can also find the latest Oracle Exadata Deployment Assistant at http://www.oracle.com/technetwork/database/exadata/oeda-download-2076737.html

Download and unzip the patch.

cd linux-x64

./config.sh

The details of the command, and the inputs it looks for are in the chapter titled “Using Oracle Exadata Deployment Assistant”, in the latest “Exadata database machine Owner’s guide”.

Oracle database machine x3-2

The Oracle database machine, gets a major makeover. As Larry Ellison phrased it in his Openworld 2012 Keynote, “Thought that the x2-2 was fast ? You Aint seen nothin Yet”.

If you go to http://www.oracle.com/technetwork/server-storage/engineered-systems/exadata/index.html, at the middle of the page, in the section titled “What’s New”, you can see a in depth technical discussion of the changes incorporated in the x3-2.

So without further Ado, let me explain what the changes are, in the x3-2 compared to the x2-2

Hardware Improvements

Faster CPU’s/More Cores.

– The Oracle Database Machine x3-2, uses the Intel Xeon E5-2690 Processors (2.9Ghz). 2 Sockets, 8 cores each, total 16 cores in each database node (The x2-2 had 12 cores per node). These are the Sandy bridge processors (x2-2 had the Intel Xeon westmere processors), which have a new micro architecture, and are extremely fast (Comparable in speed to the IBM Power7 cpu’s).

So now in the full Rack of x3-2, the database machine has 128 CPU Cores (The x2-2 had 96 Cores).

– The CPU’s on the exadata cells have been upgraded to use the Intel Xeon E5-2630L (2.0Ghz) Sandybridge processors. The Cpu’s are 6 cores each.

More Physical Memory (DRAM)

– The Oracle Database Machine x3-2 has 128Gb of DRAM memory per database server. This is expandable to 256Gb of Memory. So in the Full Rack you can have upto 2048Gb (2Tb) of physical memory.

– The physical memory on the x3-2 exadata cells, has been upgraded to have 64Gbytes of Ram.

More 10GigE networking ports

– The 4 Networking ports on the database server, mother board are now 1/10Gbe. They are autosensing,and are copper only. The remaining 2 Network ports are 10Gbe and can be connected via fiber.

More Flash Cache.

– The x3-2 exadata storage servers now use the Sun F40 Flash cards instead of the Sun F20 Flash cards used in the x2-2. Each Card is 400Gb. There are 4 PCI-E Flash cards in each cell. So you have 1600Gbytes of Flash cache in each cell. In a full rack x3-2, you get 22.4Tb of Flash cache (The x2-2 had 5Tb of Flash cache in a full rack).

So what does this increased amount of Flash mean in terms of performance ?

On an x3-2 full rack, you can get
– 1.5 Million datatase read iops from the flash cache.
– 1 Million database write iops from flash cache
– 100Gbytes/sec Flash Cache, scan throughput

New 1/8th Rack

A new configuration (In addition to the Full, Half & Quarter configurations) of a 1/8th Rack has been announced. So customers can now buy a configuration smaller than the quarter rack. It is really a 1/4th rack with half the cpu’s, half the flash cards and half the disks turned off. So the hardware price is lower and the software licensing costs are lower.

The other improvements include lower power consumption and improved cabling and airflow.

One notable change is that, the x3-2 now, does not have a KVM. This leaves 2U at the top of the Rack, where customers can deploy their in home switches, for network connectivity.

The number of disks, the type of disks, the disk capacities and speeds, in the exadata x3-2 cells,remain the same as it was in the x2-2 cells.

Software Improvements

Exadata Smart Flash Cache Write-Back

With the improved write speeds of the new PCI-E flash cards, the flash cache can now used as a write-back cache. This means that as soon as the data is written to flash cache, oracle database considers the write complete (ie it does not have to wait till the data is written to the physical magnetic disk). This helps improve the performance of applications that are currently bottlenecked on database writes.

On the x2-2, the random writes were written to the flash cache too, however it had to be written to disk (Or strictly speaking, to the disk controller cache) before the write was acknowledged by the database as completed. With the write-back cache functionality in x3-2 as soon as the write is persisted in the flash cache the database considers the write as complete. The writes to disk only get done when the ESS software detects that new blocks need to be read from disk to the flash cache and there is no free space in the flash cache. At such times, least frequently used data from the flash cache gets written to physical disk.

The smart flash cache algorithm makes sure that things like backups do not overwrite the entire cache.

The Full Rack x2-2 can do 1 million write iops to flash cache using this new functionality.

Reduced database brownout time during cell failure/removal.

In previous versions of the ESS software there could be upto 8 seconds of brown out time, when a cell failed, which has been now reduced to sub second.

Unbreakable Enterprise Kernel

– The database servers and Exadata storage servers on the x3-2 now use Oracle Unbreakable Enterprise Kernel 1.

The UEK1 was the operating system on the x2-8’s for a while now. With the x3-2’s we now use the UEK Kernel on the x3-2 database and storage server.

DBFS

– DBFS now supported on Solaris and Sparc Super Cluster.

The above list of hardware and software changes are just the highlights, not a complete list.