Tag Archives: Database

Blended Data Warehouse SW/HW Solutions Phased Into the Cloud

Relational Database Solutions “In a Box”

Several of the relational database software vendors, such as IBM, Oracle, and Teradata have developed proprietary data warehouse software to be tightly coupled with server hardware to maximize performance.  These solutions have been developed and refined as “on-prem” solutions for many years.

We’ve seen the rise of “Database (DW)  as a Service” from companies like Amazon, who sell Redshift services.

Amazon Redshift is a fast, fully managed data warehouse that makes it simple and cost-effective to analyze all your data using standard SQL and your existing Business Intelligence (BI) tools.  It allows you to run complex analytic queries against petabytes of structured data, using sophisticated query optimization, columnar storage on high-performance local disks, and massively parallel query execution. Most results come back in seconds.

RDB Complex Software/Hardware Maintenance

In recent times, the traditional relational database software vendors shifted gears to become service providers offering maximum performance from a solution hosted by them, the vendor, in the Cloud.    On the positive side, the added complexity of configuring and tuning a blended software/hardware data warehouse has been shifted from the client’s team resources such as Database Administrators (DBAs), Network Administrators,  Unix/Windows Server Admins,… to the database software service provider.  The complexity of tuning for scalability, and other maintenance challenges shifts to the software vendor’s expertise, if that’s the abstraction you select.  There is some ambiguity in the delineation of responsibilities with the RDBMS vendor’s cloud offerings.

Total Cost of Ownership

Quantifying the total cost of ownership of a solution may be a bit tricky, especially if you’re trying to quantify the RDBMS hybrid software/hardware “on-prem” solution versus the same or similar capabilities brought to the client via “Database (DW) as a Service”.

“On-Prem”, RDB Client Hosted Solution

Several factors need to be considered when selecting ANY software and/or Hardware to be hosted at the client site.

  • Infrastructure “when in Rome”
    • Organizations have a quantifiable cost related to hosting physical or virtual servers in the client’s data center and may be boiled down to a number that may include things like HVAC, or new rack space.
    • Resources used to maintain/monitor DC usage, there may be an abstracted/blended figure.
  • Database Administrators maintain and monitor RDB solutions.
    • Activities may range from RDB patches/upgrades to resizing/scaling the DB storage “containers”.
    • Application Database Admins/Developers may be required to maintain the data warehouse architecture, such as new requirements, e.g. creating aggregate tables for BI analysis.
  • Network Administrators
    • Firewalls, VPN
    • Port Scanning
  • Windows/Unix Server Administrators
    • Antivirus
    • OS Patches

Trying to correlate these costs in some type of “Apples to Apples” comparison to the “Data Warehouse as a Service” may require accountants and technical folks to do extensive financial modeling to make the comparison.   Vendors, such as Oracle, offer fully managed services to the opposite end of the spectrum, the “Bare Metal”, essentially the “Infra as a Service.”  The Oracle Exadata solution can be a significant investment depending on the investment in redundancy and scalability leveraging Oracle Real Application Clusters (RAC). 

Support and Staffing Models for DW Cloud Vendors

In order for the traditional RDB software vendors to accommodate a “Data Warehouse as a Service” model, they may need to significantly increase staff for a variety of technical disciplines, as outlined above with the Client “On-Prem” model.  A significant ramp-up of staff and the organizational challenges of developing and implementing a support model based on a variety of factors may have relational database vendors ask: Should they leverage a top tier consulting agency such as Accenture, or Deloitte to define, implement, and refine a managed service?  It’s certainly a tall order to go from a software vendor to offering large scale services.  With corporate footprints globally and positive track records implementing managed services of all types, it’s an attractive proposition for both the RDB vendor and the consulting agency who wins the bid.  Looking at the DW Service billing models don’t seem sensical on some level.  Any consulting agency who implements a DW managed service would be responsible to ensure ROI both for the RDS vendor and their clients.  It may be opaque to the end client leveraging the Data Warehouse as a Service, but certainly, the quality of service provided should be nothing less than if implemented by the RDB vendor itself.  If the end game for the RDB vendor is for the consulting agency to implement, and mature the service then at some point bring the service in-house, it could help to keep costs down while maturing the managed service.

Oracle Exadata

Here are URLs for reference to understand the capabilities that are realized through Oracle’s managed services.










IBM Mainframe
IBM Mainframe

Note: The opinions shared here are my own.

WordPress Shortcode API to Cloud Storage to Sell Any Digital Intellectual Property.

So, I was a browsing, going through bills, and thinking, hey relating to my other article on Google Docs and their new API where you could use them as a data warehouse, it occurred to me.   Why can’t we have a public API for all the Cloud Storage systems like Amazon Web Services (AWS) S3 (or Box.com), create a plugin to WordPress, add E-Commerce, and you now have your own place to sell digital music, or any Digital intellectual, property store, or host your own database OLTP or OLAP.

And my bro, Fat Panda, might have been thinking the same thing.  He’s one step behind, but he will catch on.  I will try to update for ‘the cheap seats’ in a bit.

For the cheap seats, even those static files stored up in the cloud, you can use a similar model to Google Docs <-> Google Fusion where you add tabular data to storage, read,over-write, or update using home made table locking mechanism, and essentially use the cloud as a data warehouse, or even a database.  Microsoft seems to have a lead on transitional and analytical storage with Microsoft Azure, relational in nature in the cloud, but it is so much simpler than that with cloud storage, although if not implemented with ‘row’ locking,there is an issue with OLTP (On Line Transaction Processing) row level, high volume, but with OLAP, On Line Analytic Processing, not so much, analyzing the way your business does business, and profit more from your consumer data.  There are easy ways to implement row level locking for row level locking of tabular data stored in cloud storage like AWS or Box.Net,  The methods to implement row level locking for OLTP systems using storage in the cloud are easy to implement, and will remind you of old school type alternatives to supplement the AutoNumber columns in MS Access or Identity columns in SQL Server. At the end of the day to either sell digital intellectual property from a WordPress implementation, or run your entire business with a robust cloud database solution for OLTP or OLAP systems using flat file storage!  Why go through all this when the Amazons AWS and Microsoft Azure have or will yearn to start building these solutions in parallel?  Cost effective solutions, and the entire database arena monopolized by Oracle, IBM, Microsoft, and MySQL, just got extended to a whole lot of database vendors.  It may take a while, but we already know the big Gorilla in the room Google is the first to strike in this game, as a non-traditional database vendor, cloud storage provider with their updated Google Docs API, and optionally usage of their Fusion application.

PostgreSQL / local database and SOA, mid tier for cloud solutions to improve performance

In an article I read from the NY Times, Salesforce.com may be making a play to banish Oracle as a supported platform. However, the system which might be interesting would be a PostgreSQL, or in memory database, acts as a local cache for the transaction based system, clears the local database records/cache after it uploads the ‘staged’ data from the local database to a cloud database where the data is ultimately stored. The activities on the local database should be fast, and the cloud database is a) data that may be transformed to any cloud based solution vendor(s), if necessary, if an SOA is built on top of the local database which communicates with the cloud via APIs. b) enables a local data-mart, if not transferred in real time, i.e. use a nightly transformation and have access to “day of” BI on a limited set of local data, c) again transaction performance and data segregation of the warehouse. This architecture is already in use at many firms, but I wanted to call it out. Another option is to use two cloud database solutions, one ‘local’ to your region, and one globally dispersed for performance and redundancy using an ETL, although I am not convinced this would be a great architecture.  The second cloud tier can be a transformation from the first for regulatory archiving, if required by law either for finance or DR (Disaster Recovery) policy.