February 2, 2011

Exadata notes

It’s been a while since I penetrated Oracle’s tight message control and actually talked with them about Exadata. But Doug Henschen wrote a good article about Exadata based on an Andy Mendelsohn webcast. I agree with almost all of it. At first I was a little surprised that Exadata’s emphasis shift from data warehousing to OLTP/generic consolidation hasn’t gone more quickly, but on the other hand:

Doug did overstate when he said that columnar architectures give 100X or more compression. That doesn’t happen. Yes, columnar compression can be >10X in a variety of use cases, while pre-Exadata Oracle index bloat can approach 10X at times; but even if you’re counting that way I doubt there are many instances in which it actually multiplies out to >100.

In 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 told me that Oracle won’t let them text Exadata except on Oracle premises. In one case, this is a deal-breaker keeping Exadata from being considered for a purchase, and Oracle still won’t budge.

Finally, I’m pretty sure that this “new” Softbank Teradata replacement Oracle has been touting since September as competitive evidence — which Doug’s article also references — isn’t quite what it sounds like. I believe Teradata’s version of the story, which somewhat edited goes like this: 

  • The Oracle Exadata decision at Softbank Mobile was driven by business management in spite of Teradata being recommended by the technical team.
  • To reiterate, the data warehouse project team recommended Teradata over Oracle.  The Teradata proposal was well received in terms of TCO, performance, ease of use and safety of transition, etc. against Oracle Exadata.  However, the technical team’s recommendation was overruled due to the business mandate to standardize on Oracle throughout the company.
  • SoftBank Mobile has over 800 Oracle specialists in IT departments and Software subsidiaries.
  • The Exadata performance is being compared to the existing production system.  Teradata was NOT invited to benchmark a current generation system.
  • Also, Softbank Mobile is a reseller of Oracle.

Teradata went on to clarify:

Here are some key points regarding the Teradata systems at SoftBank:

  • Two Teradata systems:  Production #1 – 32 nodes.  Production #2 – 12 nodes.
  • Production #1 had nodes ranging from ~3-7 years old.
  • Production #2 had nodes that were ~8 years old.
  • Teradata V2R5 was end of life at the time of replacement.
  • We did not get a chance to compete for this business.

Bottom line: Oracle’s big competitive replacement of Teradata systems was against 3-8 year old boxes that the customer’s technical staff recommended be replaced by more Teradata gear.

Comments

26 Responses to “Exadata notes”

  1. Chris Bird on February 2nd, 2011 9:40 am

    The telling statement, at least for me, in the excellent post above is “the technical team’s recommendation was overruled due to the business mandate to standardize on Oracle throughout the company.”

    Admittedly that was in your characterization of Oracle’s response, but I think this heightens the issue that at some level, a database is a database is a database. As far as many in leadership (especially IT operations leadership) you want to reduce the number of platforms and management complexity. It then doesn’t matter what evidence is used; that in-house intransigence makes penetration very difficult.
    I then hear arguments like “these niche players are all interesting (e.g. Vertica, Netezza, GreenPlum or whatever), but we know that our chosen Vendor will get there (they always have) so let’s put this off until they do.” Cost vs innovation, yet again. IT operations isn’t rewarded by being innovative, it is rewarded by keeping the lights on and managing to budgeted cost. Cost reduction is only useful if it is budgeted/forced cost reduction.

  2. Jim McManus on February 3rd, 2011 10:21 am

    Curt,
    In every competitive deal, there is always a winner and a loser inside a customer’s 4 walls. There is always someone that supports the loser. I’m sure that you could write the same thing for deals that TD won over Netezza, Oracle and others.
    I don’t understand why you would write about how the loser lost and not include more detail as to how SoftBank made the correct decision to transform their business, reduce their TCO, etc? Are you slamming SoftBank for making a bad decision or supporting TD? I don’t understand your point

    Also, Oracle is currently doing a lot of Exadata POC in the field which is now backfiring on Netezza and Teradata. You might want to investigate.

    Regards,

    Jim

  3. Curt Monash on February 3rd, 2011 4:53 pm

    Hi Jim,

    My point re Softbank is that Oracle claims the win is validation of Exadata technical superiority over Teradata, but the facts don’t support Oracle’s claim. Maybe Exadata is better than Teradata for some customers, and maybe Softbank is even one of them. But Softbank’s decision seems to have been based on generic trust in Oracle, not on its independent research.

    As for investigating Oracle Exadata POCs in the field, I’d love to investigate, but I’ve never heard of a single specific example of it happening, let alone one who would give me enough information to be interesting.

  4. Greg Rahn on February 3rd, 2011 8:52 pm

    Curt-
    Certainly there have been numerous database vendors, including Teradata, who have claimed competitive wins that involve:

    • old existing hardware
    • old existing software
    • a technical staff that is experienced with and wants the incumbent database
    • an incumbent database vendor not given an opportunity to defend

    One doesn’t have to look very far to find them. In fact, I’d say that most competitive wins involve the first three of those. Interesting that this one has drawn your criticism.  Is there a particular point that makes this one news worthy for you?
    I also think it is worth noting that SoftBank Mobile did do a POC with Exadata prior to their decision, and IIRC there also was another vendor in that POC as well, it just was not Teradata. In addition, SoftBank Mobile has gone on record speaking quite highly of their new Exadata system and IIRC one of the quoted is a Senior IT Manager there, so it seems at least one technical staff member didn’t recommend Teradata.

  5. David Jackson on February 3rd, 2011 10:26 pm

    Gee Jim, are you the Director – Exadata Solution Architecture at Oracle by any chance? I guess you’re opinion isn’t really unbiased in that case then, is it? LinkedIn is such a gem for that stuff…

  6. Curt Monash on February 4th, 2011 12:02 am

    Greg,

    The particular point making this one newsworthy is the emphasis Oracle placed on it when first announcing it.

    Also, your “didn’t recommend” doesn’t quite logically follow. ;)

    That said — is this “on record” just a quote you guys wrote for them, or is it a whole speech or something?

    Thanks,

    CAM

  7. Greg Rahn on February 4th, 2011 2:33 am

    More details of SoftBank Mobile’s decision to replace Teradata with Exadata, including quotes, can be found in this Oracle Magazine excerpt.

  8. Curt Monash on February 4th, 2011 6:51 am

    Thanks, Greg. Good crunch.

    Exadata replaced 3-7 year old Teradata systems, and got a 4-5X performance increase. Using Teradata over Exadata would have given better performance than that. And the only technical reason I see for the move in the article is performance.

    What, if anything, am I missing?

  9. RC on February 4th, 2011 10:44 am

    You can standarize a lot of companies on Oracle but you can’t standarize a lot of companies on Teradata. That is very technology related. It is a result of all the technological choices Oracle and Teradata have made the last decades.

    If that isn’t a technological reason than what is…?

    I don’t think that Softbank Mobile can become an only Teradata company. Or can Teradata do OLTP too?

  10. Greg Rahn on February 4th, 2011 11:11 am

    Curt-

    >”Using Teradata over Exadata would have given better performance than that.”

    Really?!?!?!? On what technical data did you base that gross assertion?

    I have no doubt that a modern Teradata system would have been faster than the old Teradata system, but I can assure you, based on several fist hand competitive data points (latest gen Exadata vs latest gen Teradata POCs) that I have information on, it takes Teradata much more hardware to get the same performance as Exadata. On one such project I worked on last year it took a 12 node Teradata EDW 5600 series system with over 1500 HDDs to match the performance of an 8 node V2 Exadata system with 168 HDDs. Given reasonable discounts on both, there is no way that Teradata would be cheaper than Exadata, and certainly the datacenter footprint and power requirements for Teradata were significantly higher than Exadata for equivalently performing systems.

    You probably won’t read about it in the headlines or press releases, but Exadata is very successful at defending Teradata in Oracle accounts. My definition of “successful” is much more heavily weighted toward the engineering and performance side of things (in addition to not losing the account), specifically knowing Exadata has a much better price/performance number than Teradata.

  11. Curt Monash on February 4th, 2011 6:33 pm

    @Greg,

    I’m basing it on the performance increases of Teradata over the years. In no way does that invalidate your general, detail-light claim that Exadata kicks Teradata butt at other unspecified accounts, perhaps large numbers of same. I’m just saying that at the account you DID choose to name, Exadata’s comparative-performance story vs. Teradata happens to be unpersuasive.

  12. Curt Monash on February 4th, 2011 6:34 pm

    @RC,

    I’m not a big fan of extreme database standardization. Also, that’s not really the claim Greg Rahn or Larry Ellison or whoever is making — they’re saying that Exadata is better apples-to-apples AND you get the benefits of vendor consolidation on top of that.

  13. Alexander Egorov on February 5th, 2011 7:46 am

    2 Greg
    >On one such project I worked on last year it took a 12 node Teradata EDW 5600 series system with over 1500 HDDs to match the performance of an 8 node V2 Exadata system with 168 HDDs.

    Don’t try to confuse people who doesn’t fully understand all technical details. You are comparing Oracle’s sequential reads performance with Teradata random reads performance. It’s not apple to apple comparison. And yes, on one-table-full-scan test Exadata will beat Teradata with less number of nodes, but in real world’s workloads, in real life (not in marketing papers prepared by Oracle) there are no only sequential reads.

  14. Greg Rahn on February 6th, 2011 4:28 pm

    @Alexander

    There is no intent to confuse on my part by citing technical details, but contrary to your comment, I am not comparing made-for-marketing benchmarks or numbers. I am comparing a customer defined workload of around 100 BI reports run on two different systems loaded with the same data, thus the metric being compared isn’t speed of table scans or random vs sequential reads, it’s query/report/workload elapsed time as well as cost of the solution.

    The interesting data point missing here is what Curt believes to be the hardware configuration it would take Teradata to be faster than the Exadata system purchased by SoftBank Mobile –so a modern Teradata system sized for ~5x the TPerf metric of the old Teradata systems. Simply claiming “A” would be faster then “B” w/o defining the hardware underpinnings is a very uninteresting comment.

  15. Alexander on February 7th, 2011 1:31 pm

    2 Greg

    >There is no intent to confuse on my part by citing technical details, but contrary to your comment, I am not comparing made-for-marketing benchmarks or numbers. I am comparing a customer defined workload of around 100 BI reports run on two different systems loaded with the same data

    Oh, my apologies, Greg.
    It sounds very promising for Oracle! Can you tell more details? What exactly was compared: how much Exadata cells, how much DB nodes, what data model was used, how much data was loaded, workload reports? Did Oracle publish this case study? If it didn’t, so why?
    Thank you.

  16. Alexander on February 7th, 2011 2:33 pm

    >8 node V2 Exadata system with 168 HDDs
    Eight DB Server nodes with 2 Quad-core CPU and 96 Gb RAM on each plus 14 Exadata nodes with 2 Quad-core CPU and 24 Gb RAM on each, so the whole system is 176 cores and 1104 Gb of RAM. It’s against 12 nodes of Teradata 5600 which contain 2 Quad-core CPU and 96 Gb RAM per node, total 96 cores and 1152 Gb RAM.

  17. Greg Rahn on February 7th, 2011 6:59 pm

    @Alexander

    There is no case study because it was not a current system. You see, every performance claim is a new system vs. an old system. There are a few reasons for that. First, one certainly gets a larger “times faster” number when the platform being compared to is old. Second, many vendors prohibit publications of POC numbers. The later is why there will never be a formal report on new vs. new from any vendor.

  18. Matt on February 8th, 2011 3:05 pm

    I would agree with a large portion of this thread. It is difficult to compare performance in a like for like construct.

    I have had the opportunity to work various technologies. I have witnessed better performance from Teradata and Neteazza over Exadata, only to have have Exadata win. For similar reasons, it easier to make business case for a solution that has the “Potential” to reach a larger portion of a company(s) IT Portfolio.

    I have also had the experience of testing other solutions that are more flexible thus have a greater “Potential” to reach a larger portion of a company(s) IT Portfolio. Using Commodity Gear and Flash Memory Arrays

  19. Matt on February 8th, 2011 3:09 pm

    Also – based on the link in the article here, it appears that Softbank acquired 3 Oracle Database machines to replace the Teradata solution.

    I am not aware of the Teradata configuration nor how all 3 Oracle DBMs were implemented. Though it does echo my statement above.

    Like for Like comparisons are quite difficult.

  20. Martin on February 17th, 2011 10:13 am

    Just to add:
    – Exadata requires 11g which seems to be not so obvious for mamy people as good as for customers which are most likely on 10g which is enough forged.
    – Oracle raise Infiniband feature in marketing papers while if requires two full racks have to be connected with 10Gb.
    – RAC in most cases is a bottleneck as bringing mamy problmes in line with applications (OLTP) or working on 8 nodes (typically RAC is working in 2 node enviroment)

  21. Jim McManus on February 17th, 2011 1:19 pm

    At the end of the day, the right solution for customer typically comes down to the workload the customer needs addressed, migration risk levels and the technology direction within the company. In the past month, I’ve ran two Exadata POCs against TD (my name is all over them so TD can confirm)and we beat them. They were DWs, but both had an OLTP component to it that put Exadata ahead; TD just doesn’t do well in that space. To be fair, there is one that I was involved with where TD won, however, a POC was not done at the customer’s request. Sometimes those things happen in favor of a vendor due to relationships and other factors. It is what it is.

    I guess what bothers me are generate statements that one technology can do everything when indeed, there isn’t one out there. It also bothers me that a bit when folks can’t admit that another technology could actually be better than another in certain situations.

    Oracle currently has over 15 POCs in North America under way(I’m running 2). Making blanket statement that they don’t do POCs is wrong. If Oracle isn’t doing POCs, how is everyone comparing Exadata performance to the competitors? I guess they are all lies when folks make those statements that TD is faster than Exadata. Hmm….

  22. Martin on February 22nd, 2011 8:50 am

    Jim,

    You are raising OLTP flavour, so based on my experience, Exadata is good fitted for DW solution, which are backoffice systems. But in OLTP space which is really in HA mode, I am sorry but Exadata is way behind IBM or HP regarding platform stability. Perhaps in next 2-3 years but not now. And we are talking about real customers not these which were bought for being references.

  23. Enterprise Headlines and Summaries, June 2011 on July 1st, 2011 6:19 pm

    [...] #Oracle Exadata notes A couple of existing Oracle users — one rather well-known — recently told me that Oracle won’t let them text Exadata except on Oracle premises. In one case, this is a deal-breaker keeping Exadata from being considered for a purchase, and Oracle still won’t budge. [...]

  24. Ryan on October 18th, 2011 3:02 pm

    Hi Curt,

    do you know of cost analysis of Exadata vs other alternatives to arrive at your conclusion that using Exadata could save customers money? Can you point me to that?

    Thanks

  25. Curt Monash on October 18th, 2011 4:47 pm

    Ryan,

    No, and it’s hard to imagine such a study that I’d believe in, because each specific situation is too different. But anecdotally I’ve had people tell me that Exadata seems like it would make their administrative burden much lighter, and their opinions seemed well-reasoned.

  26. The eternal bogosity of performance marketing | DBMS 2 : DataBase Management System Services on July 25th, 2012 2:32 am

    [...] example of Oracle exaggeration was around the Exadata replacement of Teradata at Softbank. But the bogosity flows both ways. Netezza used to make a flat claim of 50X better performance than [...]

Leave a Reply




Feed: DBMS (database management system), DW (data warehousing), BI (business intelligence), and analytics technology Subscribe to the Monash Research feed via RSS or email:

Login

Search our blogs and white papers

Monash Research blogs

User consulting

Building a short list? Refining your strategic plan? We can help.

Vendor advisory

We tell vendors what's happening -- and, more important, what they should do about it.

Monash Research highlights

Learn about white papers, webcasts, and blog highlights, by RSS or email.