Tuesday, May 19, 2009

Oracle Licensing Update & Intel Conformance

Oracle Processor Licensing Change: Intel 5500 Series

In other blog postings (this, that), "Oracle Processor Core Factor Table" topped out on the 5400 series for the processor factor licensing discount of "0.5" with "All other multi-core chips" being tagged with a full price charge factor of "1.0".

With the release of the Intel 5500 series processors, the licensing had this new CPU family (with aggressively increased CPU throughput per core) had the factor falling under 1 processor charge per 1 core.

This had recently, and unceremoniously, changed. Oracle had provided this new quad-core 5500 series processors with the same "0.5" discount as the older multi-chip modules that Intel had been selling, for Oracle Enterprise Licensing.

See the bold-faced update below to the "Oracle Processor Core Factor Table"

Factor Vendor/Processor
0.50 Intel Xeon Series 74XX, Series 55XX or earlier Multicore chips


This is good news to the businesses which use Oracle Enterprise licenses, especially those businesses which hold a lot of data in databases, like Network Management businesses, where new data is acquired on 5, 10, 15, and 20 minute intervals on thousands of devices and stores for months.

Some Things Remain The Same: Intel Gets With The Program

This is the first real quad-core processor that Intel has produced, not using a multi-chip module - catching up to other vendors who have been in this space for years like Sun and more recently AMD.

While this does not sound like a significant change, in terms of existing Oracle licensing...
When licensing Oracle programs with Standard Edition One or Standard Edition in the product name, a processor is counted equivalent to a socket; however, in the case of multi-chip modules, each chip in the multi-chip module is counted as one occupied socket.
If applications in your Network Management Center are using Oracle Standard Edition licensing - this means businesses can upgrade hardware from older style Intel Multi-Chip Modules to the newer Single-Chip processors and get a significant increase in CPU capacity at the same license terms.

For example, comparing the licensing requirements in Oracle Standard Licensing for the newer single-chip Intel 5500 series against the multi-chip Intel 7400 series:

Series Cores Chips
7400 2 1
7400 4 2
7400 6 3
5500 4 1

In short, a (relatively new) single socket hex-core socket Intel 7400 platform (with 6 slow cores, being equivalent to an Oracle Standard Licensing 3 socket) can be replaced with a (brand new) quad-socket quad-core Intel 5500 platform (with 16 fast cores, being equivalent to an Oracle Standard Licensing 4 sockets.)

This may be a lot to absorb, but if you are using Oracle Standard Licensing on Intel, stay away from the old processors except in the smallest of configurations.

2 comments:

  1. Very interesting indeed. Has Oracle confirmed the cores versus chips definition as outlined above? It seems as if most people are interpreting the "multi-chip module" listed in the Oracle verbage to equiate to cores. Also, where can I find documentation on the 55xx series processer that confirms it is not using MCM technology?

    ReplyDelete
  2. If Oracle has placed the term "Multi-Chip Module" in a legal document, that is EXACTLY what they mean - not cores. Wikipedia is a reasonable place (although not exhaustive) to learn about this technology.
    http://en.wikipedia.org/wiki/Multichip_module

    Typically, vendors like Intel will mention when a CPU is a multi-chip module. For example, this forward looking article was describing the (future) Nahalem family members, some of which were not single-die CPU's.

    http://www.techradar.com/news/computing-components/graphics-cards/idf-not-all-intel-nehalem-cpus-single-die-147908
    "Following Intel's eight-core Nehalem CPU announcement at IDF, Tech.co.uk can exclusively reveal that not all members of the new 45nm processor family will be based on a single-die design."
    ...
    "Nehalem processors with integrated graphics will actually be built using two separate chips or dies packed into a single processor package. One die will contain the processing cores, the other the memory controller and integrated graphics core."
    ...
    "For now, it's not known what impact the multi-chip approach might have on performance. But it will give Intel more manufacturing flexibility and perhaps allow it to support new memory technologies more rapidly. Instead of requiring a respin of a monolithic single-die CPU"

    Another example announcement, note the "dual-die" vs "native" language:
    http://brightsideofnews.com/news/2009/4/16/intel-says-no-to-28nm2c-focuses-on-22nm-ivy-bridgehaswell--larrabee.aspx
    "Nehalem's die-shrink family is called Westmere and is consisted out of at least four chips: dual-core CPU [plus dual-core CPU+GPU], native quad-core, native octal-core and possibly a dual-die hex-core"

    The question about whether the Intel 5500 series is a MCM or a single die comes down to what Intel states (since we can't really open up the CPU package, nondestructively, and see.) The 5500 appears to be a single die:
    https://www.lemcomputers.com/store23/catalog/article_info.php?articles_id=1
    "MONOLITHIC QUAD-CORE.
    The Xeon 5500-series represents Intel’s first quad-core processor built on a single die. As a result, each core gets a smaller L2 cache and all four cores now share a large L3."

    ReplyDelete