Oracle says they do onsite Exadata POCs after all
When I first asked Oracle about Netezza’s claim that Oracle doesn’t do onsite Exadata POCs, they blew off the question. Then I showed Oracle an article draft saying they don’t do onsite Exadata proofs-of-concept. At that point, Oracle denied Netezza’s claim, and told me there indeed have been onsite Exadata POCs. Oracle has not yet been able to provide me with any actual examples of same, but perhaps that will change soon. In the mean time, I continue with the assumption that Oracle is, at best, reluctant to do Exadata POCs at customer sites.
I do understand multiple reasons for vendors to prefer POCs be done on their own sites, both innocent (cost) and nefarious (excessive degrees of control). And vendor-site POCs can be done quite respectably. For example, the guy who first focused me on baseball-bat testing is Teradata’s chief of in-house benchmarking.
Even so, there are advantages to having a POC under your own control (or under the control of a consultant you trust). You can get a better sense of the work involved in getting the POC results, because for the most part you can physically see all the work. That helps you assess the product’s simplicity in the actual use-case being tested. More generally, you can play with the product as you want, without asking anybody’s permission. Indeed, you can play with it when the vendor representatives aren’t even around.
Bottom line: On-site — or in some cases cloud — POCs are simply more reliable as to their conclusions than ones fully under vendor control.
Edit: Chris Kanaracus has a story on a recent Netezza win (eHarmony) that mentions on onsite POC that took 24 hours to get up and running.
Comments
10 Responses to “Oracle says they do onsite Exadata POCs after all”
Leave a Reply
While Netezza marketing has been aggressively targeting Oracle’s Exadata (and rightly so – it’s a big thread to them), they are being a bit hypocritical IMO. Even though Netezza boasts they do on-site POCs, which I am sure they do, my sources tell me the Netezza POCs they have done were all hosted at Netezza. Now, I will be up front in that each of these sources engaged at least two other vendors as well, whom also hosted their POCs. In my mind this makes perfect sense for a number of reasons. First, customers (and potential customers) are not in the business of running POCs and benchmarks, they have a day job to do, let alone run POCs on three platforms, of which two they are likely not familiar with at all. This, of course, assumes the incumbent is one of the three vendors. Second, if one is involving three vendors there needs to be data center floor space, power, networking, etc for all three sets of hardware. Again, this takes extra resources that most customers are not willing to commit to (and rightly so) for only a few weeks engagement. Third, from a vendor’s point of view, it is very time consuming and costly to ship hardware to people instead of having people come to hardware. Not only are there the transportation logistics, there is dealing with damaged or failed hardware and field service as well. I’m sure if you have talked with any data center operations person, they will tell you that on almost every hardware delivery, there is some issue to deal with. Fourth, in every case, the customer was involved in assisting the vendors in validations, etc and also was on site to observe their tests being run, so there was no “smoke and mirrors”. They had monitors with live performance metrics they were able to view while the tests were executing. They were also able to do hands-on testing as well.
Bottom line, I think any vendor would ship hardware to a customer if the deal warranted it, but it is certainly not preferred by vendors, and usually not preferred by potential customers.
Dominika,
I agree that every appliance vendor would rather do POCs at their own sites. I further agree that many users are sufficiently strapped for floor space or whatever to be happy to let them do that. Still, it wouldn’t be hard to come up with exceptions where Netezza did onsite POCs. In fact, one is in the news within the past few days. I’ll edit that link into the blog post.
Best,
CAM
Question is not really whether POC is onsite or offsite. The key question is “who” does actual work of the POC. Most of the times even if it is Netezza or Oracle SEs with a big backing of engineering advise are doing POCs. Customers should do the POC themselves even if they have to send one of their DBA to Vendor’s data center. After all these are appliances, cannot be that difficult to load the data and query it. Problem with Vendor driving their own car and saying we are the best is that its not a real scenario. Load and run the query, no fragmentation, no incrementals, its not real.
A fair point. Running it yourself at the vendor site is better than having the vendor ship you the box, and then you not being involved in running it at your own site.
Even so, there are still advantages to having it on your site, all else being equal. It lets you know there’s no funny stuff going on. It lets more people from your shop be involved in throwing random queries at the system. It just lets you be surer that you did a rigorous test.
CAM
So far the most fair POCs are done by ParAccel. I have been involved in POCs from DatAllegro, Teradata, Netezza etc. ParAccel folks do the most professional and unbiased POCs. They will even let you drive the car and are very helpful in doing deep dive in their architecture during the POC. The product is also much better in doing quick POCs, its pure load and go. Believed it when I saw it in action in a customer environment.
I find it interesting that you say that. ParAccel seems very insistent on tightly controlling their POCs in-house. And while it’s great if they make Barry Zane available for discussion, I’d be concerned about any vendor that insists its CTO be deeply involved in every POC.
I frankly think users should prefer Vertica’s “Sure, download it yourself and play with it, or run it yourself in the cloud” attitude to ParAccel’s “You mustn’t go near the product without our close supervision” approach. And based on the two vendors’ widely differing sales results, it seems that’s what users DO prefer.
But if you’re simply saying that you find software-only and/or columnar analytic DBMS easier to deal with than appliances and/or row-based systems, and ParAccel is the main example of such you’re familiar with, I’m not going to try to dispute your preference … at least not in this particular comment thread. 😉
I also should note for those just tuning in that you not long ago posted a false, negative comment about Vertica, and that you post anonymously …
[…] a lot more work than one might at first guess. One such reason is that it spawned four related blog posts over the past few […]
I am not sure where do you get the information regarding CTO involved in POC. We as a DW consultants are not biased towards any company. ParAccel has allowed customers to play with their software all the time and have done head to head POCs. Vertica does the same too and maybe do it more openly. What can I say from your comment is that you are being biased towards Vertica as you were with DatAllegro even though DatAllegro really didnot have any meat. Anyway, I just posted my comments with what we experienced.
I got my information straight from ParAccel. If you don’t think that’s a reliable source, so be it.
As for your anonymous claims that you’re not biased while I am — I yawn.
[…] other Exadata news, the long-standing observation that Oracle doesn’t like to do on-site Exadata POCs still holds true. A couple of existing Oracle users — one rather well-known — recently […]