HPE Memory RAS; Excels at being Average

A recent HPE blog stating memory errors are not the end of the world was meant to reassure clients to accept regular & unplanned platform disruptions. In reality what HPE ends up saying is there is little difference with the other commercial Intel server vendors and their own as they all range from below average to average at best.  Just so happens, this specific blog was written by the HPE Server Memory Product Manager who might be forgiven for painting this dire picture only to then present the best alternative; Yes, HPE SmartMemory. *shock*

To HPE’s credit, they have quite a bit of documentation discussing server Reliability, Availability & Serviceability (RAS) features, specifically about their memory subsystem. They are fairly forthright about their strengths and weaknesses of the entry, mid-range and high-end servers. Sadly though, at every level there message is full of qualifiers, limitations and restrictions which require the consumer to wade through and understand all of the requirements.

An HPE whitepaper from February 2016 titled “How memory RAS technologies can enhance the uptime of HPE ProLiant servers” paints a starkly different picture than the blog. The whitepaper states on page 2 in the 2nd paragraph of the introductory summary section “It might surprise you to know that memory device failures are far and away the most frequent type of failure for scale-up servers.“, up to 2X the rate of the next closest part when the memory is configured with a memory protection configuration not better than SDDC+1.  There is another graph that immediately follows this one showing when memory is configured using a protection scheme of DDDC+1 it decreases memory failures by 85%. That is pretty good, yet the value of 85% used in the whitepaper does not jive with the blog which states when using HPE SmartMemory, memory errors are reduced 99.9998% (yes, that is 5 x 9’s).  I call out this discrepancy because right after claiming 5×9’s they point the reader to the very whitepaper I am citing here.

This blog is not meant to define all of the different terms used, you will have to do some of that work. However, it is worth noting that all of the wonderful features touted in the HPE blog, in the HPE whitepaper and may other sources, the consumer will find there are many qualifiers, limitations and restrictions.  Such as.

  1. E5 chips do not support DDDC or DDDC+1
  2. E5 chips only support SDDC or SDDC + rank sparing
  3. Memory sparing consumes (wastes) either 25% or 12.5% of installed capacity
  4. EX chips support SDDC, SDDC + rank sparing, SDDC+1 and DDDC+1
  5. But, DDDC+1 is ONLY using x4 DIMMs and not x8 DIMMs
  6. DDDC+1 requires x4 DIMMs
  7. Advanced ECC is an option used across 2 DIMMs but can only fill 2 of 3 DIMM slots per channel
  8. Memory Mirroring is the most expensive in terms of cost & performance
  9. Memory Mirroring wastes 1/2 of the DIMM slots for the mirror – not usable
  10. Memory Mirroring only allows you to fill 2 of 3 DIMM slots per channel
  11. Memory Mirroring has a potential performance impact for WRITES

Let’s be clear, consumers have 3 primary options to configure memory on any of the Intel servers.

  1. Performance mode which delivers the highest bandwidth with the lowest reliability features. Not an ideal option for in-memory workloads despite the appeal to maximize the bandwidth.
  2. Lockstep Mode meant to strike a balance of slightly decreased bandwidth (can be up to 50%¹) while increasing reliability over performance mode.  Probably the most common option selected.
  3. Memory Mirroring Mode delivers the highest reliability at the expense of wasting 1/2 the memory capacity as well has a slight performance decrease (remember, this mode can only use 2 of the 3 DIMM slots per channel so you already lose 1/3 of the memory capacity).

What is HPE’s response to clients who want increased memory RAS; especially for those in-memory workloads such as SAP HANA?  Buy more expensive E7 based servers to receive slightly higher memory RAS capability OR install more memory on the already RAS-deficient E5 based servers to increase its capacity to utilize memory spare ranks.

Net-net is that HPE is pushing proprietary memory that is far more expensive than the industry standard memory traditionally used with Intel servers that has earned it the reputation as a low-cost leader relative to traditional Enterprise-class systems like IBM POWER or SPARC. That is evident in the SAP HANA space as the systems required to support these in-memory workloads tend to require more capacity; more cores to achieve the core to memory ratio’s and more sockets to achieve more memory capacity with its associated bandwidth.  Yet, HPE remains true to form as regardless of the path taken, it comes with increased cost, limitations, restrictions and qualifications.

Contrast the never-ending “Compromise” Intel options, IBM’s POWER8 servers use Enterprise memory that is “No Compromise”.  This buffered memory offers spare  capacity, spare lanes, memory instruction replay, chipkill and an incredible DDDC +1+1 allowing for multiple DRAM failures before experiencing a system event.  The design point for POWER8 memory is simple: Not to fail!

AS you consider platforms to host in-memory workloads such as SAP HANA, DB2 BLU, consider which basket you want to place all of your eggs into.  A platform with a memory subsystem designed not to fail or a platform with unending limitations as listed above. The choice should be easy – Choose POWER!

 

SAP HANA – could I have extra complexity please?

Just returned from IBM’s Systems Technical University conference held in Orlando having delivered presentations on 4 different topics.

  1. Benefits of SAP HANA on POWER vs Intel
  2. Why IBM POWER systems are datacenter leaders
  3. Only platform that controls Software Licensing
  4. Why DB2 beats Oracle on POWER (implied that it beats Intel).

With the SAP Sapphire conference last week in Orlando, there was a slew of announcements.  Quick reminder for the uninitiated with SAP HANA, that it is ONLY supported on Intel and POWER based systems running one OS; SUSE or RedHat Linux. With that, IBM POWER continues to deliver the best value.

What is the value offered with the POWER stack? Flexibility! It really is that simple.  If I had a mic on the plane as I write this, I would drop it. Conversely, what is the value offered going with an Intel stack? Compromise!

Some of the flexibility offered thru IBM POWER systems are: Scale-up, scale-out, complete virtualization, grow, shrink, move, perform concurrent maintenance, mix workloads: existing ECC workloads on AIX or IBM i with new HANA running Linux all on the same server.  All of this runs using the most resilient HANA platform available.

Why do I label Intel systems as “Compromise” solutions? It isn’t a competitive shot nor FUD.  Listen, as an Client Executive and Executive Architect for an Channel Reseller, I am able to offer my clients solutions from multiple vendors that include IBM POWER and Intel based systems manufacturers.  I’ve made the conscious decision though to promote IBM POWER over Intel.  Why? Because I not only believe in the capabilities of the platform but also having worked with some of the largest companies in the world, I regularly hear and see the impact running Enterprise workloads on Intel based servers has on the business.

If you read my previous blog, I mention a client who just recently moved their Oracle workloads from POWER to Intel.  Within months, they’ve had to buy over $5M in new licenses going from a simple standalone and a few 2-node clusters (all on the same servers) to an 8-node VMware based Oracle RAC cluster.  This environment is having daily stability issues significantly impacting their business.  Yes, their decision to standardize on a single platform has introduced complexity to the business costing them money, resources (exhausted & not having the proper skills to manage the complexity) that impacts their end-users.

The “Compromise” I mention to host SAP HANA on Intel is that everything has to be an asterisk by it – in other words a limitation or restriction – everything requires follow-up questions and research to ensure what the business wants to do, can be done. Here are some examples.
1) VMware vSphere 5.5 initially supported 1 VM per system which has now been increased to 4 VM’s, but with many qualifications.
a) Restricted to 2 & 4 socket Intel servers
1) VM’s are limited to a socket
2) 2 socket server ONLY supports 2 VM’s, 4 socket would be 4 x 1 sockets each
b) Only E5_v2, E5_v3, E7_v2 and E7_v3 chips are supported – NO Broadwell
c) Want to redeploy capacity for other? Appliances certified only for SoH or S4H
uses cannot be used for other purposes such as BW
d) Did I mention, those VM’s are also limited to 64 vCPU and 1 TB of memory each
e) If a VM needs more memory than what is attached to that socket? No problem, you have to add an additional socket and all of its memory – no sharing!
2) VMware vSphere 6.0 just recently went from 1 to 16 VM’s per system.
a) VM’s are still limited to a socket or 1/2 socket.
b) 1/2 socket isn’t as amazing as it sounds.  Since vSphere supports 2, 4 & 8 socket servers, there can be 16 x 1/2 socket VM’s.
c) What there cannot be, is any combination of VM’s >1 socket with 1/2 socket assigned. In other words, a VM cannot have 1.5 or 3.5 sockets. Any VM resource requirement above 1 socket requires the addition of an entire socket.  1.5 sockets would be 2 sockets.
d) Multi-node setups are NOT permitted …. at all!
e) VM’s larger than 2 sockets cannot use Ivy Bridge based systems, only Haswell or Broadwell chips – but ONLY on 4-socket servers.  Oh my gosh, this is making my head hurt!
f) If using an 8-socket system, it only supports a single production VM using Haswell ONLY processors.  NOT Ivy Bridge and NOT Broadwell!
g) VM’s are limited to 128 vCPU and 4 TB of memory
3) VMware vSphere 6.5 with SAP HANA SPS 12 only supports Intel Broadwell based systems. What if your HANA Appliance is based on Ivy-Bridge or Haswell processor technology? “Where is that Intel rep’s business card? Guess I’ll have to buy another one since I can’t upgrade these”
a) VM’s using >4 sockets are currently NOT supported with these Broadwell chips
b) Now, it gets better. I hope you are writing this down – For 2 OR 8 socket systems, the maximum VM size is 2 sockets.  Only a 4 socket system supports 1 VM with 4 sockets.
c) Same 1/2 socket restrictions as vSphere 6.0.
d) Servers with >8 sockets do NOT permit the use of VMware
e) If your VM requirements exceed 128 vCPU and 4 TB of memory, you must move it to a bare-metal system ….. Call me – I’ll put you on a POWER system where you can scale-up, scale-out without of this mess

Contrast all of these VMware + Intel limitations, restrictions, liabilities, qualification or simply said “Compromise” systems to the IBM Power System.

POWER8 servers run the POWER Hypervisor called PowerVM.  This Hypervisor and its suite of features deliver flexibility allowing all physical, all virtual and a combination of physical & virtual resource usage on each system. Even where there are VM limits such as 4 on the low-end system, that 4 could really be 423 VM’s.  I’m making a theoretical statement here to prove the point. Let’s use a 2 socket 24 core S824 server.  3 VM’s, each with 1 core (yes, I said core) for production usage and the 4th VM’s is really a Shared Processor Pool with 21 cores.  Those 21 cores support up to 20 VM’s per core or 420 VM’s. Any non-production use is permitted.

Each PowerVM VM supports up to 16 TB of memory and 144 cores.  VM size above 108 cores requires the use of SMT4 whereas <=108 cores permit SMT8.  Thus, 144 cores with SMT4 is 576 vCPU’s or 4.5X what Intel can do with 4X the memory footprint.  By the way, that 108 core VM would support 864 vCPU’s – just saying!  Note: I need to verify as the largest SMT8 VM may be 96 cores with only 768 vCPU.

Not only can we allocate physical cores to VM’s and NOT limited to 1/2 or full socket increments like Intel, but POWER systems granularity allows for adjustments at the vCPU level.

PowerVM supports scale-out and scale-up.  Then again, if you have heard or read about the Pfizer story for scale-out BW, you might rethink a literal scale-out approach. Read IBM’s Alfred Freudenberger’s blog on this subject at https://saponpower.wordpress.com/2016/05/26/update-sap-hana-support-for-vmware-ibm-power-systems-and-new-customer-testimonials/

While on the subject of BWoH/B4H, PowerVM supports 6 TB per VM whereas the vSphere 6.0 supports is 3 TB and the limitations increase from here.

Do you see why I choose to promote IBM Power vs Intel? When I walk into a client, the most valuable item I bring with me is my credibility.  HANA on Intel is a constant train wreck with constant changes & gotcha’s. Clients currently with HANA on Intel solutions or better yet, running ECC on Intel have options.  That option is to move to a HANA 2.0 environment using SUSE 12 or RedHat v7 Linux on POWER servers. Each server will host multiple VM’s with greater resiliency providing the business the flexibility desired from the critical business system that likely touches every part of the business.

HPE; there you go again! Part 1

Updated Sept 05, 2016: Split the blog into 2 parts (Part 2). Fixed several typo’s and sentence structure problems. Updated the description of the Superdome X blades to indicate they are 2 socket blades while using Intel E7 chips.

It must be the season as I find myself focused a bit on HPE.  Maybe it’s because they seem to be looking for their identity as they now consider selling their software business.  This time though, it is self-inflicted as there has been a series of conflicting marketing actions. From what they say in their recent HPE RAS whitepaper about the poor Intel server memory reliability stating in the introductory section that memory is far and away the highest source of component failures in a system.  Shortly after that RAS paper is released, they post a blog written by the HPE Server Memory Product Manager stating “Memory Errors aren’t the end of the World”.  Tell that to the SAP HANA and Oracle Database customers, the latter which I will be discussing in this blog.

HPE dares to step into the lion’s den on a topic with which it has little standing to imply it is an authority how Oracle Enterprise software products are licensing in IBM Power servers.  As a matter of fact, thanks to the President of VCE, Chad Sakac for acknowledging that VMware has a Oracle problem.  On August 17th, Chad penned what amounts to an open letter to Larry & Oracle begging them …. No, demanding that Larry leave his people alone.  And, by “his people”, I mean customers who run Oracle Enterprise Software Products licensed by the core on Intel servers using VMware.

Enter HPE with a recent blog by Jeff Kyle, Director of Mission Critical Solutions.  He doesn’t distinguish if he is in a product development, marketing or sales role.  I would bet he it is the latter two as I do not think a product developer would put themselves out like Jeff just did.  What he did is what all Intel marketing teams and sellers have done from the beginning of compute time when the first customer thought of running Oracle on a server that wasn’t “Big Iron”.

Jeff sets up a straw man stating “software licensing and support being one of the top cost items in any data center” followed by the obligatory claim that moving it to an “advanced” yet “industry-standard x86 servers” will deliver the ROI to achieve the goals of every customer while coming damn close to solving world hunger.

Next is where he enters the world of FUD while also stepping into the land of make-believe.  Yes, Jeff is talking about IBM Power technology as if it is treated by Oracle for licensing purposes the same as an Intel server, which it is not.  You will have to judge if he did this on purpose or simply out of ignorance.  He does throw the UNIX platforms a bone by saying they have “excellent stability and performance” but stops there as only to claim they cost more than their Industry standard x86 server counterparts.

He goes on to state UNIX servers <Hold Please> Attention: For purposes of this discussion, let’s go with the definition that future UNIX references = AIX and RISC references = IBM POWER unless otherwise stated.  As I was saying, Jeff next claims AIX & POWER are not well positioned for forward-looking Cloud deployments continuing his diminutive descriptors suggesting proper clients wouldn’t want to work with “proprietary RISC chips like IBM Power”. But, the granddaddy of all of his statements and the one that is complete disingenuous is:  <low monotone voice> “The Oracle license charge per CPU core for IBM Power is twice (2X) the amount charged for Intel x86 servers” </low monotone voice>.

In his next paragraph, he uses some sleight of hand by altering the presentation of the traditional full List Price cost for Oracle RAC that is associated with Oracle Enterprise Edition Database.  Oracle EE DB is $47,500 per license + 22% maintenance per year, starting with year 1.  Oracle RAC for Oracle EE EB is $23,000 per license + 22% maintenance per year, starting with year 1.  If you have Oracle RAC then you would by definition also have a corresponding Oracle EE DB Licenses.  The author uses a price of $11,500 per x86 CPU core and although by doing he isn’t wrong per se, I just do not like that he does not disclose the full license cost of #23,000 up front as it looks like he is trying to minimize the cost of Oracle on x86.

A quick licensing review. Oracle has an Oracle License Factor Table for different platforms to determine how to license its products that are licensed by core. Most modern Intel servers are 0.5 per License.  IBM Power is 1.0 per License.  HP Itanium 95XX chip based servers, so you know also has a license factor of 1.0.  Oracle, since they own the table and the software in question can manipulate it to favor their own platforms as they do, especially with the SPARC servers.  It ranges from 0.25 to 0.75 while Oracle’s Intel servers are consistent with the other Intel servers at 0.5.  Let’s exclude the Oracle Intel servers for purposes of what I am talking about here for reason I said, which is they manipulate the situation to favor themselves. All other Intel servers “MUST” license ALL cores in the server with very, very limited exceptions “times” the licensing factor which is 0.5.  Thus, a 2 x 18 core socket would have 36 cores. Ex: 2s x 18c = 36c x 0.5 License Factor = 18 Licenses.  That would equal 18 Oracle Licenses for whatever the product being used.

What Jeff does next was a bit surprising to me.  He suggests customers not bother with 1 & 2 socket Intel “Scale-out” servers which generally rely on Intel E5 aka EP chipsets.  By the way, Oracle with their Exadata & Oracle Database Appliances now ONLY use 2 socket servers with the E5 processors; let that sink in as to why.  The EP chips tend to have features that on paper have less performance such as less memory bandwidth & fewer cores while other features such as clock frequency are higher, a feature that is good for Oracle DB.   These chips also have lower RAS capabilities, such as missing the MCA (Machine Check Architecture) feature only found in the E7 chips.  He instead suggests clients look at “scale-up” servers which commonly classified as 4 sockets and larger systems.  This is where I need to clarify a few things.  The HP Superdome X system, although it scales to 16 sockets, does so using 2 socket blades.  Each socket uses the Intel E7 processor, which given this is a 2 socket blade is counter to what I described at the beginning of this paragraph where 1 & 2 socket servers used E5 processors.  The design of the HP SD-X is meant to scale from 1 blade to 8 blades or 2 to 16 sockets which requires the E7 processor.

With the latest Intel Broadwell EX or E7 chipsets, the number of cores available for the HD SD-X range from 4 to 24 cores per socket.  Configuring a blades with the 24 core E7_v4 (v4 indicates Broadwell) equals 48 cores or 24 Oracle Licenses.  Reference the discussion two paragraphs above.  His assertion is by moving to a larger server you get a larger memory capacity for those “in-memory compute models” and it is this combination that will dramatically improve your database performance while lowering your overall Total Cost of Ownership (TCO).

He uses a customer success story for Pella (windows) who avoided $200,000 in Oracle licensing fees after moving off a UNIX (not AIX in this case) platform to 2 x HPE Superdome X servers running Linux.  This HPE customer case study says the UNIX platform which Pella moved off 9 years ago was actually a HP Superdome with Intel Itanium processors server running HP-UX.  Did you get this? HP migrated off their own 9-year-old server while implying it might be from a competitor – maybe even AIX on Power since it was referenced earlier in the story.  That circa 2006 era Itanium may have used a Montecito class processor. All of the early models before Tukwila were pigs, in my estimation.  A lot of bluff and hyperbole but rarely delivering on the claims.  That era of SD would have also used an Oracle license factor of 0.5 as Oracle didn’t change it until 2010 and only on the newer 95xx series chips.  Older systems were grandfathered and as I recall as long as they didn’t add new licenses they would remain under the 0.5 license model.  I would expect a 2014/2015 era Intel processor would outperform a 2006 era chip, although if it would have been against a POWER5 1.9 or 2.2 GHz chip I might call it 50-50 J .

We have to spend some time discussing HP server technology as Jeff is doing some major league sleight of hand as the Superdome X server supports a special hardware partitioning capability (more details below) that DOES allow for reduced licensing that IS NOT available on non-Superdome x86 servers or from most other Intel vendors unless they also have an 8 socket or larger system like SGI – oh wait, HP just bought them.  Huh, wonder why they did this if the HPE Superdome X is so good.

Jeff then mentions an IDC research study; big deal, here is a note from my Pastor that says the HPE Superdome is not very good; who are you going to believe?

Moving the rest of the blog to Part 2.

 

 

Power investments continue to pay off!

This article is to highlight the announcements IBM is making in their Fourth Quarter 2015 related to the Power platform.  This is one of the largest announcements in years that I can recall touching Linux, IBM i, AIX, virtualization, management, ISV’s and the platform itself.  Since I am not an IBMer with access to schedules there may be a few things that differ.

First,  you will want to register.  This is a virtual event which is convenient as you can not only register at anytime but also watch it at anytime online.  Register at https://engage.vevent.com/index.jsp?eid=556&seid=80414&code=Social_Tiles.

As a Business Partner I am glad to see that IBM is delivering on what they told us over the past several years.  They are taking their $B investment delivering useful and leading technologies  with Linux on Power as is needed but also with AIX and IBM i.  These latter two Power pillars are far more mature and do not require the technology enhancements nor the ISV adoption like Linux on Power (LoP) requires.  Stands to reason there will be more activity around the LoP space, not because that is the future and the others will diminish but  for what I mentioned, it is less mature relative to the enterprise AIX & IBM i markets.

This is the extensive list of features being announced this quarter.  I will add a reference section after the announcement(s) to allow you to get more information on each of the features.

AIX

  • AIX 7.2 – some really good features!
    • “Live Update” or apply AIX updates concurrently without requiring a reboot
    • RDSv3 over RoCE optimizes Oracle RAC performance using Oracle RDSv3 protocol with Mellanox Connect RoCE adapters (up to 40 Gb)
    • Workload optimization with Flash
    • Dynamic System Optimizer
    • BigFix Lifecycle for automated and simplified patching
  • New AIX Enterprise Edition packaging
  • AIX 6.1 Withdrawal from Marketing April 2016

IBM i

  • New IBM i v7.1 TR11
  • New IBM i v7.2 TR3
  • S822 expanded capabilities – supports IBM i
    • Requires VIOS for I/O

Virtualization – PowerVM

  • New VIOS release – v2.2.4 based on AIX 7
  • NovaLink architecture provides scalability features for OpenStack deployments
  • New SRIOV capabilities
  • Introducing vNIC Adapter – increases performance with SRIOV
  • Shared Storage Pool enhancements

HMC

  • New HMC model – CR9
  • New HMC version – 8.8.4
  • New virtual HMC offering – Run 8.8.4 in a VMware or RHEV VM (x86)

Power platform

  • New Power8 firmware release – 840 or 8.4 level
  • New PCIe adapters
  • PurePower enhancements
    •  IBM i support
    • vHMC support
    • PurePower Integrated Manager improvements
    • Order  both S822 & S822L with initial order

Management

  • New PowerVC 1.3 version – more management & OpenStack integration features
    • Advanced policy-based management
    • Supports MSPP
    • Expanded vSCSI & NPIV support for certain storage models
  • Manage Power servers using PowerVC with OpenStack with VMware’s vRealize

Security

  • PowerSC NERC Profile compliments existing PCI, DOD STIG, HIPPA, SOX-COBIT

High Availability

  • New PowerHA 7.2 version
    • Integrates Power Enterprise Pools as part of a PowerHA failover operation
    • Improved integration with LPM
  • Non-disruptive upgrade
  • Integrates with new AIX Live Update feature
  • New wizard to use GLVM for low cost mirroring option
  • Enhanced EMC SRDF support
  • Supported on AIX 6.1 TL9 and later
  • Supports Power6 and new servers

Linux

  • New Power Linux server models  – true price parity with Intel servers. Built on OpenPOWER
    • S822LC – up to 20 cores, 1 TB, 2 SFF HDD & 5 PCIe slot 2U server
    • S812LC – up to 10 cores, 1 TB, 14 LFF HDD & 4 PCIe slot 2U server
  • PowerKVM features
    • Dynamically add/remove cpu & memory resources from VM’s
    • Live Migration
  • IFL enhancements – IFL’s run IBM software in a Linux VM on 4 socket & larger Power servers with a 70 PVU vs 100 or 120

Performance

  • New CAPI offerings
  • New SSD offerings – Gen 4 drives, higher performance & capacities
  • 36 port EDR 100 Gb/s Infiniband Switch delivering latency as low as 130 ns

ISV & Software

  • New Linux ISV partnerships – More & more ISV’s are coming to IBM asking to be a part of the Power market revolution taking place
  • SAP HANA announcements
  • New BigInsights 4.1 features with Hadoop & Spark
  • PureApp now available with Power8 servers (announced July 2015)

Cloud

  • SoftLayer announcements – Linux on Power bare metal offerings
  • Power Enterprise Pool enhancements

The above list is fairly complete although lacking a lot of detail which is available online in the announcement letters or better yet contact your IBM Power Sales Specialist or Business Partner.  If your Business Partner is not proactively offering to keep you updated on these types of announcements you may want to reevaluate what value your Value Added Re-seller is providing and look for another.  Don’t settle for an order taker but a technology enabler.

IBM continues to deliver innovation, value, solutions and options to the “Good Enough” alternative with Intel where it has become obvious over their last 2 chip releases they are taking  customers for granted.  Hear how the performance of the Power8 processor gives  the equivalent of a 35 PVU vs Intel’s 70 PVU (this is a PowerMan example and not IBM itself). With IBM Software this is an immediate 50% reduction in licensing and maintenance costs.  Factor in the hypervisor efficiency and that should increase significantly.

Who doesn’t want more performance, more reliability for the same price as the competition? You can have it  your way with IBM Power8 & OpenPOWER.

 

Shiny objects & Distractions

Yet another blog on the non-stop marketing tactics by Oracle where they attempt to deflect attention on their many product weaknesses and try to create differentiation where there is none.

This latest attempt by Oracle has them promoting the performance of Oracle 12c over SAP HANA for the SAP Business Warehouse Enhanced Mixed Workload benchmark also known as BW-EML.  Oracle promotes this claim at https://www.oracle.com/corporate/features/oracle-powers-sap.html with a whitepaper posted at http://www.oracle.com/technetwork/database/in-memory/overview/benefits-of-dbim-for-sap-apps-2672504.html available for download as a pdf.

Oracle is known for making wild claims only supported by marketing claims and “Oracle internal tests”.  This is important to understand as these claims may entice customers to consider products that have not undergone any critical analysis.  One example is Oracle’s Exadata product. The Exadata name has become synonymous with Oracle’s family of integrated appliances that include discrete solutions: database, application  and data warehouse. The Exadata database solution  has zero published benchmarks yet the web is riddled with claims by Oracle on its superior performance over competitive offerings. Oracle is now claiming they have submitted their Oracle 12c in-memory database results to SAP for review and publishing for the BW-EML benchmark which SAP has yet to do.

It appears since SAP has chosen to not publish Oracle’s 12c result that Oracle is taking matters into its own hand to publish their BM-EML result since SAP is not. I have no knowledge if SAP is choosing to sit on the results but I do know this; the reason you see very few industry benchmarks on non-Oracle systems using Oracle database (enterprise edition) is for the same reason Oracle is accusing SAP of doing.  As part of Oracle’s end user license agreement they require any user who publishes performance results to submit the results to Oracle for review and approval.  If Oracle does not approve the result that user / vendor cannot publish it.  A good example where Oracle has limited their competition from publishing benchmark results using Oracle database is with the SAP Tier-2 Sales & Distribution benchmark.  Benchmark results are available at http://global.sap.com/solutions/benchmark/sd2tier.epx.  I checked yesterday (Sept 19th) and could not find any current results using Oracle Enterprise Edition database on any non-Oracle or non-SPARC servers (ie Fujitsu has results on their SPARC servers) since a HP result from around 2008.

Oracle is trying to convince SAP customers their 12c database product is relevant for in-memory Business Warehouse workloads. They further tout superior performance with a 2 socket X5-2 server using Intel’s Haswell E5_v3 chipset totaling 36 cores.  Oracle typically achieves higher results by throwing significantly more cores and memory than required by competitors and definitely not by innovation; whether that competition is IBM POWER or Intel.  One of Oracle’s “Go to” tactics is to mask and manipulate the details  stating things like “Our (Oracle) 8 processor beats IBM’s 8 processor by 2X”.  We saw that when they compared their SPARC T5-8 to a Power7+ 780 server. Of course, the devil is in the details and those details are this.  Oracle historically refers to their chips as sockets and  processors (ie the full chip that plugs into the motherboard socket) and this is whats used in published results.  They use these names interchangeably.  IBM tends to use socket when referring to a model of server such as “The S824 is a 2 socket server” or “The E850 is a 4 socket server”.   In almost all cases they refer to performance results using cores or processors which are used interchangeably. IBM tends to use chip or socket synonymously and cores or processors as the component that makes up that socket/chip.  Using Oracle terminology, the T5-8 system is configured with 8 processors of 16 cores each totaling totaling 128 cores.   This specific  IBM Power7+ 780 server only has 32 cores though yet Oracle chose to compare their 128 core server to it.  Why are you asking? This model of Power7+ server uses a 4 core per socket configuration.  Each server chassis of which it can scale from 1 to 4 chassis scales from 4 to 16 sockets or 16 to 64 cores.  This is how Oracle marketing claims 8 socket vs 8 socket which they publish their results.  As you can see though, they do not divulge to the reader that it is really 128 SPARC cores vs 32 Power7+ cores.  They leave it to the reader and consumer to figure this out taking no responsibility that they are intentionally trying to deceive and distort the facts.

With this Oracle 12c result for the SAP BW-EML benchmark you will note several areas of omission and possible deception. They do not publish any pricing data for using the in-memory feature. Using list price will easily run close to $200K per core and of course depends on whether a few features are chosen or not.  Oracle claims their 1 server with 36 cores beats all others by 2X.  They state in the whitepaper this one server is actually one of 8 database servers in the Exadata X5-2 appliance and not a single 2 socket 36 core standalone server. This is very important to understand as it drives up the software by a factor of 8.  There are a few ways to reduce their Oracle licensing but this is not disclosed and I would argue not likely used.  Since the server used is part of a 8 node Exadata, it would require software be licensed for 8 servers times 36 cores times the Intel licensing factor of 0.5.  This equals 144 Oracle licenses which is multiplied by the licensing cost (let’s just use $200K list price for easy math – it is what it is and is not entirely fixed) $200K which comes to a grand total of $28,800,000. Yes, that is $28.8M USD. Of course, Oracle charges an annual maintenance fee that is 22% times the license price. For this example the customer would pay $6,336,000 per year and every year.

Next, Oracle claims to have used just 1 x 36 core X5-2 server for this workload yet it also has the storage that comes with the 8 server solution. They could have just as easily used their standalone X5-2 server in their attempt to achieve these results. Furthermore, why did they not use the 2 node Oracle Database Appliance (ODA)?  It seems obvious they need the full Exadata infrastructure which heavily relies on SSD based PCIe adapters to achieve the desired performance.  The SSD heavy architecture has become the default configuration on Exadata from previous solutions which relied on high capacity but slower 10K rpm HDD as it delivers higher performance and higher margins.  Don’t forget Oracle shifts some of the database processing and subsequent cost from the DB servers to the storage servers then charge $20k per disk…yes, I said per “disk”.  Suckers line up to the left and those who have done their homework are already running Oracle workloads on IBM’s Power servers.

SAP has been clear in their roadmap.  They are moving toward an architecture developed around HANA.  I can’t blame Oracle entirely for touting their product as a viable database alternative.  IBM’s DB2 with BLU technology is a superior product to Oracle Database Enterprise Edition that runs even faster on Power8 (over SPARC & Intel) and is even less expensive.  Yet, IBM has posted an SAP HANA result running Linux on POWER8 servers for the BW-EML benchmark.  If customers wants to see how DB2 on Power8 performs, they are welcome to view the SAP Tier-2 S&D Benchmark mentioned above for results that are 2X+ greater than Intel and anywhere from 3-4X greater than SPARC per core. Customers are absolutely free to choose Oracle or DB2 for their BW workloads but if they plan to stay with SAP for the long term they are probably investigating, evaluating if not implementing HANA technologies already.

The way Oracle could impress SAP shops would be to publish a HANA result on their infrastructure solutions.  They could always use that opportunity to co-sell the benefits of their own software solutions as better alternatives but as usual, they bust-out on stage holding up their shiny object making  wild claims in their non-stop attempt to distract customers.

I’ll close with this. As part of their performance claims they state they achieve 2X more navigation steps using a single 36 core Oracle X5-2 servers (remember it is really 1 of 8 DB servers + all the storage servers). Glad to see Oracle trying to compare per core performance.  I’ll be on the look out for other examples of their newly found realization that performance and cost is largely dependent on per core performance and not just the sum of excessive cores like the T5-8 or M6-32.  Expect the SPARC M7  with 32 sockets of 32 cores per chip to be released in 2016.  It’s really 8 clusters of 4 core chiplets or essentially two of the old Sun ROCK chips IMO.  The latency due to traffic across the interconnects for coherency and data will be unbelievable (horrible). Would not be surprised that for the few benchmarks they do publish that they M7 models with fewer chip to minimize the cross chip penalty.  Then again, they may go with the 1024 core model hoping it can best a 192 core Power8 E880 servers.  I’m guessing it will be close.