A data warehouse pricing complication: Software vs. appliances
Juan Loaiza of Oracle disagrees with a number of my opinions. We plan to talk about some of that when I visit on Thursday, after Teradata Partners. 🙂 But I’d like to throw one of his ideas out there right now. Juan contends that comparisons of Oracle Exadata pricing are apt to be misleading because — among other reasons — Oracle licenses can be reused on other hardware, in ways that appliance software can not. (The same reasoning would of course apply to almost everybody else except Teradata and Netezza.)
Juan has a point. According to my spreadsheet, it is approximately the case that:
- 12% of Exadata’s list price is for hardware
- 30% of Exadata’s list price is for software priced on a per-disk-drive basis
- 58% of Exadata’s list price is for software priced on a per-core basis
For the sake of argument, let’s suppose:
- You buy an Exadata installation. In a few years, you want to retire the hardware and reuse the software.
- You need no more disk drives on the new hardware as you did on the old, because disks have gotten larger in line with your data volume growth.
- You need 4X the cores as you did before, because Moore’s Law is instantiated primarily via increases in core count.
Then your savings vs. purchase of a totally new machine are 30% + .25 * 58%, or 45% overall. That’s not trivial.
On the other hand, let’s suppose you’re trying to make an argument like “Oracle superficially costs around 2X what the competition does, but that doesn’t really count because half my purchase is reusable.” That argument could be swamped by maintenance costs. For example, suppose the pricing is something like:
- Oracle: 100 clams, plus 20 clams/year, and you can get 45 clams back down the road
- Teratezza: 50 clams, plus 10 clams/year
Then Oracle doesn’t look anything like a low-cost provider.
Bottom line: If a software-license solution is similarly priced to an appliance, the flexibility of software licensing can create an effective cost advantage. But if the software-license solution costs significantly more than the appliance, that flexibility is unlikely to make up for the price difference.
Comments
2 Responses to “A data warehouse pricing complication: Software vs. appliances”
Leave a Reply
Pricing can be a complicated subject as your article correctly points out. However, the pricing decision is not divided solely along software v. appliance lines. Software only licenses can vary from vendor to vendor as well. New vendors not only innovate with respect to product offerings, but often innovate with respect to licensing and business models. Vertica offers several innovative pricing and delivery options. First, we offer complete flexibility. You can purchase a software-based license, an appliance with software included, or a cloud-based license. Second, the software license is based on the amount of data you are loading into the system, not the amount of hardware you are using. Whether you need to add more hardware or upgrade your hardware at some point has no impact on the cost of the license. In addition, your development and test systems are also included in the license fee. We believe this pricing model makes it easier to understand and plan for the costs associated with an analytical database.
Dave,
Quite right. The lack of charging for development/test is nice, and in line with industry trends (charge for run-time tool use rather than development seats, open source, etc.) I believe Oracle offers free development options that work well for small OLTP/general-purpose apps, to develop and test functionality, but not so well for DW performance.
As for charging per terabyte — Oracle’s contention is that one can use the same licenses, down the road, to manage more terabytes than one does now. My rejoinder is that this would be more compelling if they weren’t also trying to charge more money each year per unit of processing power ..
CAM