May 3, 2011

Oracle and Exadata: Business and technical notes

Last Friday I stopped by Oracle for my first conversation since January, 2010, in this case for a chat with Andy Mendelsohn, Mark Townsend, Tim Shetler, and George Lumpkin, covering Exadata and the Oracle DBMS. Key points included: 

*Oracle asked me to delete a point on pricing they went out of their way to make, because they are in quiet period — even though nobody said it was confidential at the time, we weren’t under NDA, and it looks like public information to me anyway. Frankly, I’m not sure I was right to comply.

Oracle also told me quite a bit about Exadata onsite POCs (Proofs of Concept) and Exadata references, but I’ll save those subjects for future posts. The same goes for workload management.

Oracle’s version names and numbers can get confusing, but it turns out that:

I didn’t really understand the discussion as to why certain workloads and/or workload consolidations go better on the SMP boxes of Exadata X2-8 than the blades of Exadata X2-2, but Oracle assures me that some do. I also suspect that some Oracle customers prefer large SMP boxes for no good reason other than familiarity.

As for recent-release adoption:

Exadata obviously enjoys huge performance gains over existing Oracle installations for certain analytic queries, and therefore for some whole analytic workloads. Oracle has happily trumpeted these. But it turns out that Exadata’s OLTP (OnLine Transaction Processing) performance gains are less dramatic. This makes all kinds of sense, given that Oracle’s analytic query performance was in pretty bad shape pre-Exadata, while OLTP has been just fine. The range Oracle used was 2-3X OLTP performance gains vs. existing Oracle installations on several-year-old hardware. Oracle says somewhere over 50% of Exadata physical I/O* goes against flash cache in uses cases such as running Oracle’s application suite.

*Note that physical I/O may be only a small fraction of logical; e.g., SAP long ago said that >99% of SAP transactions never hit disk.

Finally, we talked about a variety of options or other related products. Highlights included:


9 Responses to “Oracle and Exadata: Business and technical notes”

  1. Niall Litchfield on May 3rd, 2011 10:56 am


    *Database* version numbers are *pronounced* as you write them, but delimited differently. The versions you refer to would be and This actually helps make sense since there is a logic to them 11 is the major version number, 2 the minor version number, 0 reserved for something I forget and haven’t seen used in a while and 2,3,4 etc for the patchset release – though numbering there starts at different for the base releases of the minor versions. There’s also a 5th decimal place now in use for what are confusingly cslled patchset updates or PSUs and are the quarterly security plus recommended patches interim releases. so would refer to 10g release two with patchset 4 and psu 2. So there is a logic, honest, even if not a great one in practice.

    And advanced security has been a licensable option for ages, but most new security options go into that pricing structure, it isn’t new with the features you describe above – rather they are new to it.

  2. Kevin Closson on May 3rd, 2011 12:15 pm

    I’m relieved to hear the message regarding OLTP on Exadata was metered. I’d like to point out, however, that this news of 50% of OLTP physical I/O being serviced in Exadata Smart Flash Cache deserves some analysis.

    1. The flash cache is read only.
    2. The Exadata full-rack aggregate application write IOPS rating from spinning media is 25,000 This budgets for ASM redundant writes but does not does budget for transaction (redo) logging.

    We (well, I do actually) have no idea what read/write ratio lurks behind this nebulous 50% estimate for flash cache hit ratio, but OLTP very seldom gets above 60R:40W. That would, indeed, be a brutal R:W ratio. If we start by saturating the write-side (25,000 WIOPS) of a full rack then these applications only demand 37,500 RIOPS from flash?

    The data sheet value for flash IOPS capacity of a full rack X2 configuration, is 1.5 million IOPS–100% read of course. Not many OLTP applications exhibit a 60:1 R:W ratio and even less have no transaction logging overhead.

  3. Curt Monash on May 3rd, 2011 9:04 pm

    Thanks, Niall! Fixed.

  4. Oracle on active-active replication : DBMS 2 : DataBase Management System Services on May 3rd, 2011 9:08 pm

    […] of the reasons that Oracle likes to review analyst publications before they go out. Notwithstanding what an Oracle executive told me Friday, I received an email from Irem Radzik of Oracle which said in part: I am the product marketing […]

  5. Hemant K Chitale on May 4th, 2011 2:22 am

    My experience with EBusiness Suite and Peoplesoft is that a large portion of the ‘performance’ issues is CPU-bound logical I/Os (in the SGA). Physical I/O is an issue only on occassion. So, I, too, am sceptical about the gains from the Database Machine. Performance gains will come from the moving to 11.2 optimizer and faster CPUs rather than from the storage and flash cache.
    Pure speculation on my part without having actually seen a Database Machine.

  6. Marco Ullasci on May 5th, 2011 8:24 am

    Even if the X2-8 nodes are big they’re not really an alternative to traditional (Sparc, Power, Itanium) Unix boxes; when I worked in HP the big point with my customers was: Intel server => “not for real production”.

    I’d not be surprised if some analytical workloads that have continuous updates/inserts can get way better results when the inter-node communication is reduced as it happens by getting a smaller number of bigger nodes.

    Why solaris? At the UKOUG in december I heard a person hinting that Solaris for intel could make a better use of the huge number of cores in the X2-8

  7. Robin Moffatt on May 6th, 2011 9:46 am

    “40-50% of customers have Oracle 11g running somewhere in their shops” – any clarity on whether that “somewhere” is Production usage, or just a DBA with a dev instance on a server under their desk?

    I’m interested that you mention POCs as onsite, as I’d heard that Oracle would only do POCs at their facilities and not at customer sites. Is this not the case?

  8. Curt Monash on May 6th, 2011 10:04 am

    No clarity on the 40-50% figure. Sorry.

    The point on the POCs is that Oracle now does SOME Exadata POCs onsite — a lot less than 50%, but also not 0%. More on that in a future post.

  9. Patrick McFadin on May 10th, 2011 12:21 am

    I did a Exadata POC and I left fairly impressed with the technology. It’s clear that the hardware and software are working tightly together. The SQL optimizer is aware that it is using Exadata hardware and makes some interesting decisions. It’s easy to see how you could potentially consolidate several workloads into a single Exadata system.

    We made the decision to not buy because the performance benefit didn’t justify the cost, but we almost bought it for the management. When you purchase the system you get it all. Network, storage, power distribution and even the cabinet. That means that one less call to EMC or NetApp for disk. (Which is ok in my book) If Oracle sells a lot of Exadata, and I think that’s probably going to happen, how is that going to effect the enterprise storage vendors bottom line?

    Just wait until they release a virtualization solution in the same hardware-software-in-a-box next year. It will strangely start looking like mainframes all over again except we’re replacing the IBM logo with Oracle.

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:


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.