Oracle Switches CPU from Intel to AMD for Exadata Cloud Services

Working on Oracle Exadata Cloud Infrastructure X9M, I noticed the upper OCPU went up pretty high compared to X8M. This made me curious and I discovered the switch from Intel to AMD for the CPU:

Exadata X9M (Exadata Cloud Services)

[root@v1exacs01c1db01 ~]# more /proc/cpuinfo | grep "vendor_id|model name" | sort | uniq -c
8 model name : AMD EPYC 7J13 64-Core Processor
8 vendor_id : AuthenticAMD
[root@v1exacs01c1db01 ~]#

Compare this to on-premises Exadata and they are still Intel processors:

Exadata X9M (on-premises)

[root@v1exadb01 ~]# more /proc/cpuinfo | grep "vendor_id|model name" | sort | uniq -c
76 model name      : Intel(R) Xeon(R) Platinum 8358 CPU @ 2.60GHz
76 vendor_id       : GenuineIntel
[root@v1exadb01 ~]#

Comparing with previous Exadata Cloud @ Customer and this is also still Intel processors:

Exadata X8M (Exadata Cloud @ Customer)

[root@v1exacc01c1db01 ~]# more /proc/cpuinfo | grep "vendor_id|model name" | sort | uniq -c
6 model name      : Intel(R) Xeon(R) Platinum 8270CL CPU @ 2.70GHz
6 vendor_id       : GenuineIntel
[root@v1exacc01c1db01 ~]#

This change make sense as the latest Intel processor is the “Intel® Xeon® Platinum 8358 Processor”, which has 32 cores and 64 threads.

Compare this with latest AMD processor “AMD EPYC™ 7713”, which has 64 cores and 128 threads. I believe the AMD EPYC 7J13 in the Exadata Cloud Infrastructure X9M is of the same family.

This allows Oracle to offer 252 OCPU for Quarter Rack, 2 sockets x 64 cores x 2 database servers less 2 cores per database servers for KVM. This is double what could be offered if Intel processors were used. And seems as Oracle charge per a OCPU, it’s a smart move that was quietly done 😉

If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this oracle blog via WordPress/e-mail.


Zed DBA (Zahid Anwar)

Oracle and Microsoft to Interconnect Oracle Cloud and Microsoft Azure

On the 5th of June 2019, both Oracle and Microsoft made a joint announcement on the interconnection between Oracle Cloud and Microsoft Azure:

Oracle’s Press Release
Microsoft Press Release

The key aspects of the announcement are:

  1. A direct interconnect between Oracle Cloud and Azure Cloud, starting in Oracle’s Ashburn (North America) region and Azure Washington DC (US East) region, with plans to expand additional regions in the future*.
  2. Unified identity and access management, via a unified single sign-on experience to manage resources across both Oracle Cloud and Azure.
  3. Supported deployment of custom applications and packaged Oracle applications (such as JD Edwards, PeopleSoft and E-Business Suite typically referred to as Oracle Applications Unlimited) in Azure Cloud with the Oracle databases (such as RAC, Exadata, Autonomous Database) deployed in the Oracle Cloud.
  4. A collaborative support model for customers leveraging these new capabilities.

*With only one region available in both Oracle and Azure, this will allow for failure in a Availability Domain in Oracle Cloud and/or Availability Zone in Azure Cloud but not a failure of a whole region in either.  So until more regions are added, Disaster Recover will be limited to Availability Domains/Availability Zones:


Figure 4 from Oracle’s blog “Overview of the Interconnect Between Oracle and Microsoft“.

What does this mean?

In a nutshell, for those customers who have Microsoft Azure as their Cloud platform of choice, can now migrate their application tier to Microsoft Azure, whilst migrating the database tier i.e. Oracle database (mandatory for Application Unlimited) without having to worry about the all important latency (high-throughput, low-latency as stated in Oracle’s blog post).  It is however unclear if there will be any charge for outbound/inbound traffic between the clouds, but it does seem from the documentation and blog post that both Oracle’s dedicated FastConnect and Azure’s dedicated ExpressRoute are required, which are both fix rate products.  It also helps those customers who require the favourable database licensing on the Oracle Cloud, more info can be found in my blog post here.

This is certainly a step towards the trend of multi-cloud/hybird-cloud platforms.

More info can be found regarding this announcement from our Version 1’s blog post here.

Another interesting read from SearchCloud Computing in regards to this announcement.


If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this oracle blog via WordPress/e-mail.


Zed DBA (Zahid Anwar)

Oracle’s Autonomous Database (Cloud)

So yesterday I attended the “Autonomous Database GTM Roadmap Sales Workshop” at Oracle’s London office.  This training is for Oracle partners such as Version 1, which is one of Oracle’s strategic partners.

A lot of what is in this blog post is subject to Oracle’s Safe Harbour statement.

My Key Takeaways

1 . Maturity

The Autonomous Database is still very new!  It’s like back in 2008 when the first Exadata Machine (V1) was launched, it was great, it was game changer for large Data Warehouses.  But it wasn’t suited for OLTP and as with anything new it had its fair share of “teething issues”.  However, now passing its 10 year anniversary last year and on its 8th iteration the X7, it’s now a very mature product.  It’s suited for mixed workloads (since the 2nd iteration) and has had so many new features over the years that makes it now a very compelling offering if it suits your business needs.

This is the same for the Autonomous Database, at launch it was only suited for Data Warehouse just as the first Exadata Machine (however soon after another offering was available for OLTP, see further on), it’s not perfect and it has it’s fair share of “teething issues”.  However, come its 10 years anniversary and all the features that are in the road map are implemented, it will be a different story and it will be another very compelling offering from Oracle, again if it suits your business needs.

2. Makeup

The make up of the Autonomous Database in the Oracle Cloud is:

  1. Oracle’s Extreme performance platform, Exadata part of the Oracle Engineered Systems
  2. A streamlined version of 18c database soon to be 19c
  3. Oracle Cloud Automated Data Centre Operations

This is the not so “secret sauce” 🙂

3. Infrastructure Offerings

So the Oracle Autonomous Database comes in 2 offerings:

  1. Serverless Exadata Cloud Infrastructure, which just means it’s shared.  This is for non-mission critical workloads and is non-deterministic performance.  The minimum is 1 TB storage and 1 OCPU and it’s the low cost entry point.  Please Note: This is the ONLY offering at present (Jan 2019).
  2. Dedicated Exadata Cloud Infrastructure, which is as the name suggests dedicated.  This is for mission critical workloads and is deterministic performance.  To be confirmed, but envisioned to be offered like Exadata conventional sizes, i.e. quarter, half and full rack.  The minimum is 1 TB storage and 1 OCPU to all OCPU in the rack size provisioned.  It will have private networks unlike the above offering which is public.  Expecting “soon”, so could be Q2 or Q3 of 2019.

4. Workload Offerings

Once you’ve selected between shared or dedicated, then you need to decide what type of workload as there are two products that apply the autonomous optimisations:

  1. Autonomous Data Warehouse (ADW), which optimises complex SQL, stores in columnar format and creates data summaries.  This was the only offering at launch.
  2. Autonomous Transaction Processing (ATP), which optimises response time, stores in row format and creates indexes autonomously.  Now also available.

The current offering doesn’t let you change between the two, however it is on the road map to be able to converted from one to another, for example to want to test which works best for you or if you have in hindsight made the wrong selection.

5. Automatic Indexing

This one is probably a contentious yet interesting topic!  Us DBAs are used to the world of indexes and us “knowing” what’s right, however the world moved on and AI and Machine Learning is taking away laborious task from us.  The Autonomous Database in the ATP can analyse the workload and use AI and ML to see what indexes are needed over a period of time and eventually have the same elapse time of a workload, however the most interesting aspect is that it will only have indexes that are needed and have a net reduction in indexes, which can often get left behind and have little to no benefit.  There’s no denying we can know better and have a set of indexes with some redundant indexes too but how often is this reviewed to remove unused indexes, add new ones as queries change?  This Automatic Indexing takes away that headache with some volatility as it works out what is required.  I can really see the benefits here and see this being the norm just as Automatic Undo Management is, who in this day and age manages undo segments?

6. Autonomous

The Autonomous Database is:

  1. Self-Driving, performs database maintenance tasks such as tablespace space management, etc.  Automate upgrades and release updates.
  2. Self-Securing, automatically apply secure patch online.  Out the box, all data and network traffic is encrypted.
  3. Self-Repairing, can automatically detect and apply fixes data issues, i.e. resolve block corruption using Active Data Guard, ensure high availability using Real Application Clusters (RAC) and in the event of disaster, use Data Guard physical standby.

7. Is it for you?

Just talking Oracle platforms, there’s a spectrum of platforms, from most Manual to most Autonomous:

  1. Database on commodity hardware on premise
  2. Database on Engineered Systems (Exadata) on premise
  3. Database on Oracle Cloud Infrastructure (OCI)
  4. Exadata Cloud Services / Exadata Cloud @ Customer
  5. Autonomous Database Cloud Services

The more autonomous you go, the more you can focus on your business.

Anyone who’s interested in Autonomous Database, come talk to us 🙂

If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this oracle blog via WordPress/e-mail.


Zed DBA (Zahid Anwar)

How to easily delete files in the Oracle Cloud using CloudBerry Explorer

So you have some Oracle Cloud storage, which was probably thrown in as a freebie initially by Oracle 🙂  Now your freebie is expiring and you decide you want to retain the service but now as metered, so you want to delete all the unnecessary files, which in my case are old Oracle database backups.

As you can see here I have 50k plus files using 897Gb:

Oracle Cloud Web Console Details

Here are the files listed in the Web Console:

Oracle Cloud Web Console List Objects

To delete them one by one isn’t feasible.

So the solution is to use a 3rd party file explorer, in my case CloudBerry Explorer from CloudBerry Labs:

The freeware version is perfectly fine, no need to purchase Pro or use the trial.  Just click ‘Cancel’ on the Register Product dialogue and then the application will load.

Once installed, to connect to your Oracle Cloud storage, you can follow this link:

However, the key to connecting is to have the username in the format of:

<Your Oracle Cloud Service Instance Name>-<Your Oracle Cloud Identity Domain>:<Your Oracle Cloud User Name>


Also select the correct ‘Account location’ which will fill the ‘Authentication Service’:

CloudBerry Login

Keystone, set to ‘Do not use’.

When you finally manage to get connected, you’ll see something like this:

CloudBerry Explorer

Now you have the freedom, to browse your files and delete at leisure 🙂

If you found this blog post useful, please like as well as follow me through my various Social Media avenues available on the sidebar and/or subscribe to this oracle blog via WordPress/e-mail.


Zed DBA (Zahid Anwar)