Will there be a Cloud Bubble?

“The Network is the Computer” was the tagline used by Sun Microsystems for years. It captures the essence of what we take for granted today using our smart phones, wireless laptops and general World Wide Web usage.

Today, many customers are actively looking at hosted computing that has become ubiquitous with the word “cloud”. This “cloud” is a virtual pool of computing resources; servers, storage, network and applications. Many of these services have been around for 15 years such as Googles Gmail, Hotmail owned by Microsoft now moved to Outlook.com. Consumers would open a browser, login to send, receive and store emails. This was followed by retailers such as Amazon and ebay then social media such as MySpace, Facebook, Twitter and LinkedIn. Corporate customers cut their teeth with SalesForce.com as one of the early innovators of cloud services.

Now we have not just a plethora of web services from Dropbox, YouTube, Twitter to every major retailer having a web presence and countless more. Corporations though are looking at cloud providers: both major players, custom players and boutique players. The major players consist of offerings like Microsoft Azure, Amazon Web Services (AWS), Google Cloud, Oracle Cloud, IBM’s SoftLayer and VMware’s vCloud Air. Some of the custom cloud providers may have a mix of “CoLo” or Co-Location where customers place their own compute resources into someone else’s datacenter. They may retain their own administrative resources or use a managed service offering from that CoLo facility. Examples that come to mind are ATT, OVH, SuperNap, ScaleMatrix and TeraGo. Boutique clouds tend to be high touch, white glove and often specialize in specific services. Examples of Boutique cloud providers are SIS (www.thinksis.com) who I currently work for, Black Mesh and Global IT.

Regardless of what online services you use as a consumer or which corporate cloud provider they use, what are the threats to using these services? Clearly, one of the most insidious and ever increasing threats has been cyber terrorism. These threats, attacks, damage and thefts are ongoing. Some of the cyber terrorists are individuals and criminal syndicates; often from the former USSR countries and the Eastern Bloc countries. Other cyber terrorism is state sponsored meaning it is backed by the government of a country such as China, North Korea, Russia, Iran and others. Lastly, there are corporate terrorists. These are sometimes linked with state sponsored terrorists where the corporation is used as a front for their malicious activity.

Other threats to cloud users is the provider itself. They suffer outages due to infrastructure failures, failures outside of their control such as a construction crew putting in a new water line accidentally digs up a data line carrying network traffic out of the cloud providers. Most of the issues described by the latter issue are addressed as the better providers have redundant everything from Power, networking, cooling, facilities, etc.

The general concept for a cloud provider to make money is to deploy as inexpensive of infrastructure as they can, deploy as many customer workloads onto as few servers as possible driving up the system utilization just to the point of pain then dial it back 1 notch. This businesses model is at risk if they are not diversified or have other revenue streams as they only make money between the cost of goods provided and cost of goods sold. They are constantly trying to identify ways to reduce the cost of servers, disk drives, storage, adapters, network components as well as using lower cost resources to manage the operations desk or provide administration. Depending on the business, employee cost is at the top of expense cost. Finding resources with most of the needed skills for 25% less cost is probably good enough. The sum of these cost saving efforts can lead to a weakened environment due to increased operator error. Compatibility with software due to the use of white-box servers is less predictable leading to problems often not seen until specific features are needed. Reliability may be lessened with white box servers because they build the servers themselves such as Google does or obtain from a no-name provider who delivers a basic server consisting of cpu, memory and I/O with no bells and whistles that translates into lacking reliability & serviceability features. This may all be managed by resources that are junior, lacking professional certifications, experience and skills handling not just the day to day processes but also when there is a critical event. Do they do the proper triage, problem analysis and determination implementing a logical plan to resolution?

Of course, every provider is different and cloud providers are not the only ones at risk as many customers have computing operations far worse than most clouds. This may be why they are looking to cloud providers to improve their situation. In many, if not most cases, placing workloads often described as “Systems of Engagement” into the cloud makes sense while retaining “Systems of Record” at the customers premise. Cloud providers are especially vulnerable to cyber attacks because of the many holes coming into their facilities to support the customer diversity. Once inside the network an attacker will wait patiently to learn information about the environment to plan their next attack. With regard to the infrastructure, some shops invest in both architects and engineers to build, deploy and maintain their infrastructure. With discipline and skill they overcome the challenges that lesser providers struggle with; maintaining cash flow while keeping the lights on and payroll met.

We just have to look at the recent outages at AWS http://fortune.com/2015/09/20/amazon-cloud-snafu/ over the past week (Sept 26, 2015) to see how it can impact a business. we all have probably felt the impact of this whether it is our bank using internet banking, logging into a school computer down for maintenance or using a major internet service like NetFlix. How many customers like this will endure frequent outages before they decide to move back to an on-premise solution or at least a hybrid solution? With regard to cyber attacks, the list of companies attacked in 2015 would double the size of this blog to write them all out. How many of those will occur before businesses decide they cannot afford to not be in control of their network backbone, interview and hire their own people so they know the capabilities of the resources responsible for solution remaining safe and available.

I am not predicting a bubble but have to wonder why we wouldn’t expect one due to the security and availability threats.

Tell me what you think in the comments section.

Thank you

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.

Are you keeping score?

Are you keeping up with all of the changes taking place with IBM’s Power portfolio over the last 2 years?  How could you as there have been so many changes to just about every area on the platform.  Just in case, let’s review some of the features and changes.  Now, since I don’t work at IBM, don’t have a magic 8 ball or a direct line to Doug Balog, IBM’s Power General Manager there is a chance I will get something wrong and I’m sure I’ll leave a few things out.  What I will do though, is not talk about any upcoming products – which between you and me – They will be freakin awesome!

  • Power8 processors support both Big Endian and Little Endian
  • Little Endian offerings are with Linux: RedHat 7.1, SUSE 12 & Ubuntu 14 & 15
  • Big Endian offerings are AIX, IBM i as well as Linux (SUSE 11 & RedHat 6.5 & 7.1)
  • PowerVM supports both BE and LE Operating Systems (concurrently)
  • IBM introduced an alternative hypervisor to PowerVM; an open source alternative based on KVM
  • PowerKVM on Scale-Out models support both BE and LE Linux
  • Still supports the use of VIOS for PowerVM environments with HMC enhancements
  • Uses Kimchi to manage PowerKVM environments
  • SR-IOV adapters (finally) available
  • Nice set of quad port Ethernet (2×10+2x1Gb) adapters available for AIX & Linux workloads
  • 40 Gb RoCE adapters
  • PowerVC based on OpenStack replaces Systems Director (portions of it)
  • PowerVC integrates with VMware’s vRealize allowing it to manage & provision Power Systems
  • Gen 2 of a Converged solution called PurePower using S822(L) servers + V7000 + Mellanox switches
  • 2X performance increase over Power7. Greatest increase that I recall from 1 gen to the next
  • 3X greater memory bandwidth for Scale-Out over Power7/7+ Entry servers
  • Just under 2X greater memory bandwidth for Power8 Enterprise vs Power7/7+ Enterprise servers
  • 3X greater I/O bandwidth for all Power8 servers over all Power7/7+ servers
  • 2X more L2 cache
  • Addition of L4 cache
  • SMT8 that is dynamic per VM unlike Intel with their static 2 way Hyperthreading
  • Up to 1 TB Ram per socket
  • Centaur buffer supports DDR3 & DDR4 memory
  • Scale-Out servers (S & L models) use Enterprise Memory just like the Enterprise servers
  • Significantly more Fault Isolation Registers & Checkers
  • Significant reliability enhancements to processor, cache, memory & I/O subsystems
  • Gen3 I/O Drawers with more PCIe slots per drawer
  • Gen3 PCIe slots (x8 & x16) in all servers
  • Available split backplane in ALL Scale-out servers – YEAH!
  • Significant increase in the number of internal Scale-Out server disk slots: 8 – 18 slots
  • 6 or 8 x 1.8″ disk slots (model dependent) using IBM’s Award Winning (Is it?) Easy Tier
  • Hot swap PCIe slots in ALL models
  • Scale-Out internal disk RAID cache increased from 175 MB in Power7 to 7 GB
  • PowerVM, PowerKVM and RHEV hypervisor options
  • Bare-Metal server option (certain models)
  • Enhanced Voltage Regulator Modules (VRM)
  • Enhanced EnergyScale
  • Some models lowered their software tier from Large to Medium and Medium to Small
  • Some servers received improved warranties
  • Power Enterprise Pools with Mobile Cores & Mobile Memory
  • Integrated Facility for Linux (IFL’s) on E850, E870 & E880
  • L model servers that use PowerVM is called PowerVM Linux Edition equivalent to Enterprise Edition
  • Supported with SAP HANA
  • All non-L models supported with EPIC
  • All Enterprise servers supported with EPIC
  • Delivering roughly 2X performance increase over Intel Haswell EP/EX processors
  • Power8 has 3X greater memory bandwidth than Intel Haswell EP memory in Lock-Step mode
  • Power8 has 2.5(ish) greater memory bandwidth than Intel Haswell EX memory in Lock-Step mode
  • Power8 has 2X greater L1 cache
  • Power8 has 2X greater L2 cache
  • Power8 has 2.5X greater L3 cache
  • Intel has no L4 cache
  • Intel has significantly fewer Checkers
  • Intel does not use anything similar to Fault Isolation Registers
  • Intel 1 & 2 socket servers do not support Machine Check Architecture (MCA)
  • Only Intel 4 socket & greater servers support Machine Check Architecture (MCA)
  • Intel rates their memory capabilities in performance mode which has limited RAS capability
  • For Intel to increase memory resiliency, must use lock-step mode which decreases performance
  • For Intel to increase memory resiliency, can use memory mirroring reducing memory capacity by 1/2
  • All Power8 servers have at least 1 Coherent Accelerator Processor Interface (CAPI)
  • Several CAPI solutions are available today with more coming (very soon – that’s the only hint)
  • Using IBM’s Advanced ToolChain delivers an enhanced SDK & GCC delivering increased results
  • IBM optimized software stack for Power8 delivering greater results per core
  • Optimized ISV stack for Power8 delivering greater results per core
  • ISV’s like MariaDB, WebFocus Express, Redis Labs, ColdFusion, EnterpriseDB, Magento
  • ISV’s like McObject, Veristorm, HelpSystems, SugarCRM, Zato, OpenPro and many more
  • Guaranteed utilization levels from 65 – 80% depending on the model
  • Backbone of Watson
  • Supports Apache Hadoop & Apache Spark
  • Supports BigInsights
  • Supports internal disks for Big Data
  • Supports superior BD storage option using Elastic Storage Server: S822L & GPFS Native Raid
  • Run DB2 10.5 with BLU Acceleration 2X faster than Intel
  • WebSphere & Cognos faster
  • SPSS faster and many more
  • 150+ members in OpenPower Foundation
  • Blue badged Power based solutions
  • Non-blue-badged Power based solutions available
  • PowerVP is awesome
  • Virtual HMC is coming (was a IBM SOD)

I tried to create an exhaustive list to make the point that IBM is investing, engineering and building game changing technology to help customers solve real solutions. However, unlike Oracle and Intel, who in their own ways  are both becoming very rigid and proprietary.  IBM is going in the opposite direction by actually opening up their own technology as well as embracing the open source community.  The list of ISV’s, the use of OpenStack, the use of a open source hypervisor and the most recent announcement whereby VMware’s vRealize will be able to provision & manage Power Systems is a testimony to the change taking place within IBM and with customers.  Expect more of the above and more beyond this. Expect it to blow your mind.

I may come back over time and enhance this list if I get froggy.  Of course, if you are a customer you could always invite me to speak with you and I would be happy to discuss any and all of this.

What would you like to see in addition to the extensive list above?

IBM Power Systems & VMware; Why Not!

On August 31, 2015 at VMware’s VMworld conference in San Francisco both IBM and VMware made an announcement that many would never have anticipated.  In support of private and hybrid clouds, VMware’s vRealize Automation is now able to provision and manage virtual machines (LPAR’s in traditional parlance) on IBM’s Power Systems.

Details are sparse as I write this, as well as what I have found from IBM and other online sources.  Here is a IBM blog with most of the details.  I won’t repeat what it says as it is not the focus of this article.

The focus of this blog is on why wouldn’t VMware want to do this and why shouldn’t the traditional VMware shop embrace this first step?  First, it is further evidence by both companies as they weave their unique products with open source OpenStack to deliver greater features with expanded capabilities to customers.  I am not very versed in vRealize so I’ll take the position of speaking about it in generalities. However, I am very well versed in IBM’s PowerVC which is based on the OpenStack framework.  This site management tool for Power Systems provides for image management (fancy talk for OS backup & provisioning), placement policies (aka affinity), full stack configuration of VM’s (ability to configure networking & storage), VM replication, Remote Restart (excellent VM recovery feature) and much, much more.

Because it is based on OpenStack, not only can it be enhanced to suit a customers environment via industry standard API, it can also be integrated into a customers management solution.  Enter VMware’s vRealize Automation product.  This is where the vRealize people say “yep, I know how that works” and where the people unfamiliar with vRealize “google it”.

This creates an expanded market for VMware for platform management.  We can speculate if this is all that IBM & VMware have up their sleeves OR if this is just the first of many future announcements – I don’t know but I’m erring on the side of this being the first of many future announcements.  It also gives them a bump in terms of promoting their cloud solution to customers by claiming it can support not just x86 technology but also  Big Iron like Power and System Z (yes, they announced support for it today as well but you will have to read a Z blog for more details on that).

For customers, this delivers a significant endorsement on the viability of IBM Power Systems. The x86 vendors have for years labeled RISC systems as legacy and dying. Although true for HP and SPARC Unix platforms (If I used RISC because of HP’s Itanium I would surely get criticized so I jumped to UNIX to avoid it) it has not been true for IBM.  That change happened during the Power6 to Power7 model change.  Power6 was the last of the Big Iron systems with the Big Iron mentality.  Power7 began to shift towards the new battlefront consisting of x86 based technologies.  With Power8 the shift by IBM to not only be competitive but a leader in many categories was firmly in place. Platform openness with hypervisor choice (PowerVM, PowerKVM, RHEV & Bare Metal) and OS flexibility using traditional Big Endian AIX, IBM i and Linux (RedHat v6.5 & v7.1 and SUSE 11) as well as Little Endian Linux (RedHat v7.1, SUSE 12 and Ubuntu 14). Systems management using PowerVC based on OpenStack abandoning its own and long in tooth Systems Director product.  Even using Nagio’s in IBM’s newest secure, converged and integrated PurePower ready to deploy cloud solution.  Additional features such as reliability, security, virtualization efficiency, serviceability, significant performance increases and cost competitiveness.  Add in their technology sharing with the 150+ member strong OpenPower Foundation delivering a open platform for the community to develop on and develop to solutions that Intel specifically has continued to close out.

Now with POWER8 delivering roughly 2X the performance per core running the same Linux byte ordering as what is available on x86 systems customers have a choice. Is ‘Good Enough’ good enough when I can get better for the same price with all of the features mentioned above with greater performance that can now be managed by my preferred management platform from VMware.  This doesn’t mean to imply that x86 is going away anytime soon or that Power will overtake the data center overnight.  It does mean to imply though that customers now have choices and it doesn’t have to be just one x86 vendor against another but also with a superior architecture.

Yes, this was very smart by VMware as they win no matter what the customer decides to do.  It was also very smart for IBM as it should mean increased adoption of their Power platform into shops that would otherwise avoid it for lack of integration into the VMware stack.  Both companies will then get the chance to introduce customers to other products in their portfolio which means they are already in the door….that is half the battle!

Do some manufacturers purposefully mislead? PART 1

This is a multi-part blog on the topic of “what if you like their technology but can’t trust what the manufacturer says?”.

Have you been in the situation where you like certain products from a company but find dealing with them is less than desirable? Where you have to have your guard up when dealing with them as they always have an angle. You can’t trust them in other words.

This is my experience with Oracle. Some of their products are market leaders and highly regarded while others lag the competition. Yet to listen to Oracle marketing and their sellers they were #1 in every category, their sales are the highest and growing, their performance is beyond reach of anyone else and just because they set their mind on an industry they almost immediately become the market leader (Cloud comes to mind).

Oracle is a master marketing company. Their approach seems to be if we publish it such as this example of false system leadership http://www.enterprisemanagement360.com/blog/oracles-latest-attack-ad-fizzles/

then it must be true …. why else would a company say it; if it weren’t? IBM’s Elisabeth Stahl offers an excellent response to this example at https://benchmarkingblog.wordpress.com/tag/wall-street-journal/.

Why do I bring this up? Well, Oracle doesn’t learn positive moral lessons from their mistakes, they continually refine their tactics to bluff and mislead the consumer. Look at this tweet from an 18-year Oracle veteran who is part of the Oracle marketing team. I added the numbers and circle to reference during this multi-part blog.


For item #1, Oracle marketing claims IBM is making misleading claims about Power8 having the fastest cores as well as having the fastest CPU. Let’s start with definitions because Oracle will let you believe what you want whereas I think we need to work with the same understanding. For purposes of this discussion a processor and a core are synonymous unless otherwise stated. A chip, cpu and socket are synonymous as well unless otherwise stated.   A past trick Sun Microsystems would use and carried on by Oracle is to compare servers using the terms of each manufacturer. IBM would say they have 8 processors in a Power6 570 server, which was actually 4 chips with 2 cores each for a total of 8 cores. Sun/Oracle often refers to their chips/sockets as processors. They would compare an 8 processor SPARC server which is actually 8 sockets with possibly 8 cores in each for 64 total cores to a Power server and claim they beat it by 2X – really when in reality they had 8X more cores?!

Working backwards, I cannot think of an example where IBM claims to have the fastest CPU. I have tweeted though as seen here on my twitter account @PowerMan_SIS:


The data to support my claim is readily available from SAP at http://global.sap.com/solutions/benchmark/sd2tier.epx

Take your pick of a SPARC server to see their results as seen here:


Manufacturer Model # Processors* # Cores # Users # Users/chip # Users/core # SAPS # SAPS/chip # SAPS/core
Fujitsu M10-4S 40 640 153,000 3,825 239 844,420 21,110 1,319
Fujitsu M10-4S 32 512 153,050 4,782 299 836,550 26,142 1,634
Oracle M6-32 32 384 140,000 4,375 365 793,930 24,810 2,068

* This is one of those “otherwise stated” examples. Here a processor equals a socket because that is the definition used by the SAP benchmark result.

Compare this to Power servers.



Manufacturer Model # Processors* # Cores # Cores/Proc # Users # Users/chip # Users/core # SAPS # SAPS/chip # SAPS/core
IBM Power8 E870 8 80 10 79,750 9,969 997 436,100 54,513 5,451
IBM Power7 795 32 256 8 126,063 3,939 492 688,630 21,519 2,690
Fujitsu M10-4S 40 640 16 153,000 3,825 239 844,420 21,110 1,319
Fujitsu M10-4S 32 512 16 153,050 4,782 299 836,550 26,142 1,634
Oracle M6-32 32 384 12 140,000 4,375 365 793,930 24,810 2,068

What are some observations? The data isn’t mine; it is accepted and published by SAP. The rest is MATH. The Power7 server released in September 2010 has 8 cores per chip (i.e. Processor by SAP’s definition) compared to 12 and 16 for each SPARC entry. I’m not minimizing they have more Users and SAPS per chip but just pointing out they need 50% to 100% more cores to get up to 25% more results. The Power8 server released in November 2014 has results that are very interesting. This packaging of the E870 uses 10 cores per chip and delivers over 2X the results for Users/chip, Users/core, SAPS/chip and SAPS/core. This is with a 2-system node server and IBM scheduled to deliver their 4-system node server with up to 192 cores. I think the results speak for themselves.

Another observation is that in order for Oracle to achieve results, they have to have massively large server; high numbers of chips and cores. 8X more cores in the case of the Fujitsu M10-4S just to obtain 2X the results. The MATH not only demonstrates that but also proves it.

I will wrap up Part 1 of this blog to let you digest the data, the facts and the discussion on the first of several misleading statements by Oracle marketing.

Oracle blogger speaks with forked tongue!

Oracle blogger “kgee” wrote the following at https://blogs.oracle.com/hardware/entry/2_ways_ibm_has_over.

“Two Ways IBM Has Over-promised and Under-delivered with POWER8 to Date”. “Kgee” goes on to say the the following in which I provide the highlights.

  1. Power8 is More than a Year late.
  2. Where is AIX8?
  3. Fact: AIX 7 TL3 last November just released “WPAR alt_disk ….”
  4. Fact: Apparently per IBM’s roadmaps, AIX does not yet support SR-IOV
  5. Fact: Consider all of the advantages Oracle just released in Solaris 11.2

Just because you can say it, doesn’t make it true. This Oracle blog is full of wrong statements, mis-statements and “so what”.

1) IBM (imo) releases products when the market is ready for them.  Products may be pulled or pushed as the market and competition require it. Also, SPARC & Solaris are no longer considered significant competition much to the chagrin of Larry who thinks Oracle should be the most relevant.

2) Is that all you have to criticize Power8 for?  I’ll take it! Let me know when you want me to publish the side by side comparison of Power vs SPARC delivery dates. Oracle bought Sun so you own their dates as well!

3) Marketing drives model names for hardware and OS. Remember Solaris 2.4, 2.5, 2.6 then 2.7 – I mean Solaris 7? Then Solaris 8, 9, 10 and 11?  Constant major OS re-numbering require ISV’s to re-certify which is costly and often slow. I applaud IBM for sticking with the current OS strategy  that standardizes on the current two OSes in a effort to eliminate the disruption. Customers can now do minor updates for feature & bug enhancements to get to the next server generation without being required to do a major OS upgrade.

4) The author ‘kgee’ seems to have all of the AIX facts so he/she can tell us if AIX8 is late? I do not think it is and I am not expecting one but I wasn’t expecting one so not sure how it can be late.  That said, AIX delivers more concurrent, dynamic, scalable, and secure features than Solaris.  AIX is also integrated with the hypervisor and hardware for performance, security, serviceability and reliability. Can’t say that with Solaris, much of its RAS features are in the OS.  To do it in the hardware requires significant engineering effort which IBM has in spades.

5) For your WPAR Alt_disk…. comment – congrats! You picked a OS virtualization feature to give the impression AIX is lacking or playing catch-up. Solaris only had Zones, which is a OS virtualization feature as an option for years with no hardware level virtualization offering or capability. Power has delivered PowerVM years before Solaris’s OS only option. AIX added WPAR’s with Solaris 6 which was available in 2007 – just 2 years after Solaris.  As far as the “Alt_disk” vs Live Upgrade. AIX has had that feature since 2001. Guess that puts Solaris behind by 4 years.

6) Unified Archives – Oh great, another new feature from Oracle. Just like the T series, what is that 5 generations of servers in 6 or 7 years? Congrats on a new feature. I’ll take investment protection and stability.

7) No compromise virtualization with Solaris Zones – ha, really. Read your documentation. If you say “No compromise” that means none, nada, nothing yet you say in the definition “an even greater …” inferring an improvement. How can that be if it is already “No compromise”?

8) Power, PowerVM and AIX have delivered Quality of Service for cpu, ram, I/O for years. Congrats for catching up and using the latest buzzword “Software Defined”.

9) While Oracle works to lower the compliance effort with their offering, Power and PowerVM eliminate the effort of meeting compliance with IBM i and AIX through the use of PowerSC.

10) (Note: I forgot to add this in my original response to “Kgee’s” blog. Power servers have offered SR-IOV capabilities starting in October 2012 with Power7+ 770 & 780 servers. However, for all of the neat benefits of SR-IOV, those features are mutually exclusive to features  long available in the Virtual I/O Server (VIOS). So, while Solaris was needing SR-IOV to get these features, IBM has been delivering this kind of functionality since 2004 with Power5.

Other than these 9 (now 10) items I thought your article was pretty good.  Look forward to the next one.”

Solaris is a very good OS, just like HP-UX and other Unix OSes. AIX has enterprise features that in my opinion offer customers more features and benefits. It is a bit of Ford & Chevy. However, where there is no comparison is between Power servers vs SPARC servers. Even though Oracle delivers SPARC products, even new products they are years behind in functionality, capability, security, flexibility and performance compared to Power.  At the end of the day – actually at the start of the day, customers want their servers available, secure, using as much of the resources as needed for as many workloads as possible keeping the real costs under control which is with software like Oracle database. Power controls these products while SPARC and x86 for that matter are meant to deliver a software license delivery vehicle to increase licenses for Oracles profit. Nothing wrong with profit but let’s call a spade a spade.  TCO of Power will always beat the TCO of SPARC (and x86) for these kinds of workloads!


Why Oracle should sell “cars” – not servers! – Part 1

Yes, I must apologize to car salesman as they should feel slighted by comparing them to an Oracle sales person. However, when you watch a YouTube video at https://www.youtube.com/watch?v=2bvRO8ipXWc by Scott Lynn, Solaris Product Manager at Oracle you can’t help but wonder how businesses continue to fall for what appears to be constant misleading claims and statements.

Here are a few of the claims made by Scott that I will discuss in detail below.

  • Barely 40 seconds into the video Scott claims T5 is the Worlds Fastest Processor – Why yes, this car is the “Fastest” in its class!
  • Claims each generation of SPARC has seen a 2X performance increase (while his own chart doesn’t even show that) – The HP for this engine is up 2X from last years model!
  • Power7+ is only 10% faster than it’s predecessor over the last 3 years – My car is the only manufacturer to make real improvements model year after model year.
  • Compares the $/Perf of M9000 at $18 to T5 @ $4 which is a 78% decrease in price performance – Our latest car is better, faster and cheaper whereas the competition has had issues, slower and cost more for less – that’s what I am hearing at least.
  • Made claims for a European Telco moving from a x86 / Linux / VMware solution to T4 obtaining 10X greater transactions and 2X the performance for the same number of licenses – one customer was able to achieve 55 mpg using normal driving habits – I’m sure you are no different than he is!
  • Solaris is 85% less costly than a typical x86 solution – This car pays for itself!
  • Solaris is almost $1K less per VM than competition at $2543 vs $1591 – For what you get, this car is less expensive. Let’s go talk to the Finance Manager now.
  • VMware uses 10 – 30% overhead while Solaris has virtually 0% overhead – their car has emissions controls but this car is wide open!

I almost do not know where to begin. At the end of the video he threw in a set of floor mats, mud flaps and a oil change.  Below are my responses to Scott’s claims that I listed above.

  • T5 is the Worlds Fastest Processor so says “Oracle”. That’s right, based on their own internal testing the T3 outperforms T2 by 2X. T4 outperforms T3 by 2X. T5 outperforms T4 by 2X. First, by what standard? On a core vs core basis? Socket vs socket? Maybe actual benchmarks? None of us know because they do not publish anything. All we have to go by are Scott’s words – because he said so, it must be true!  Sun published very, very few benchmarks or any other public results for the T2 & T3 servers. They started publishing some with T4 after adding the S3 core – must’ve been feeling their oats :)

The image below could be interpreted with some literary license I guess but I’m taking a literal interpretation because they do that to not just imply x86 and Power do not perform from generation to generation like SPARC T series has but in the very picture in which they criticize the competition they show themselves to not scale from generation to generation.  Stunningly inept? Oversight? It will just take a bit more research to provide various data points to support what Oracle actually shows.


The picture below shows an example of performance, pricing and sizing manipulation. Oracle uses 2 x 64 core servers to get a 28.8K result vs 2 x 16 core Power7 servers to get a 10.9K result 4X the number of cores to get just 2.6X higher performance. A more likely server solution would be 2 x Power7+ 740 servers, each with 16 cores.  Those servers would have a sell price of approximately $115K each.  There is a similar result on the cousin server running Linux called the 7R2 which delivered 13.1K EJOPs.  Those 2 x 16 core servers with 1/4 the cores of the T5 are roughly 1/2 the results of the T5’s 128 cores at a lower price. With Oracle EE database cost of $47,500 per core and WebLogic at $35,000 per core, each carrying 22% software maintenance per year – starting with the first year.  Of course, the Power solution is using DB2 and WebSphere Application Server which are both less costly than the equivalent Oracle products but the licensing model favors Power in this case. Not because of the Oracle reasons which is to manipulate vendors servers like they do with x86 and SPARC but because the 740 and 7R2 both are 2 socket servers.  Like all 2 socket servers they have a PVU rating of 70 per core.

T5 with 64 cores vs Power7 780 with 16 cores. Oracle picks a entry level vs IBM enterprise class. Intentionally use 4X the cores to get higher results while different classes to show wider cost delta.

T5 with 64 cores vs Power7 780 with 16 cores. Oracle picks a entry level vs IBM enterprise class. Intentionally use 4X the cores to get higher results while different classes to show wider cost delta.

  • Even Scott’s own chart doesn’t show a 2X increase for each generation. If I was going to say it and had the bazillions of marketing dollars that Oracle has I would at least have the red line reflect 2X from generation to generation because now it just looks like he is lying.
  • X86 has only had a 20 – 50% increase. Again, he doesn’t provide any data so we can only speculate. I will write some future blogs on the overstatement of x86 performance but in general, they do get more performance (per socket) with each generation because they tend to double the number of cores per socket: 2 => 4 => 8 => 15.  I can’t find any statements by x86 where they claim a 50% increase on a core vs core basis.  Where they make statements is usually on a socket basis and a lesser degree on a core basis – again, because their per core increase isn’t typically all that spectacular. They increase performance by doubling or adding cores.
  • In comparing Power7+ servers over 3 years one might think you were comparing it to Power6 or Power6+ but I think your marketing department is trying to be sneakier than that. However, Power7+ was first introduced in October 2012 then February 2013. Power7 wasn’t first available until March 2010 which means that Power6+ was the generation of server available 3 years prior. First, P7+ was ~30% per core better over P7 but remember that isn’t who they said they were comparing it to since it was in 3 years. Power7+ is ~43% better over Power6+. By the way, I used a P7+ 750 with 8 cores @4.0 GHz vs a P7 with 8 cores @ 3.0 GHz. In comparing to Power6, I had to pick a server that existed in each generation. I chose the P7+ 780 with 16 cores @ 4.42 GHz vs a P6+ 570 with 16 cores @ 5.0 GHz. Unlike Oracle which likes to compare a IBM Power 795 to their entry level T5 server which is like comparing a SPARC M9000 server to a SPARC T2000 – the servers simply are not in the same class.  The servers I selected are either the same in it’s class in the case of the 750 with one being first generation power7 from March 2010 vs a Power7+ 750 available in February 2013.  Also, the Power7+ 780 which replaces the Power6+ 570.  Unlike some mysterious Oracle internal testing numbers I used IBM’s rPerf numbers which are specific to Power servers (what Power Architects use to size Power vs Power) and used to compare one model to another. The data is available to the public online in their Systems Performance Report.
  • Scott compares the M9000 to a T5. That is like comparing a Cadillac to a Chevrolet Malibu. The M9000 is a true enterprise class RISC server. The T5 is not in the same league as the M9000…For that matter, nor is the M5 or M6 servers (in the same league as the M9000). This is a perfect example of the egregious marketing behavior of Oracle. They compare the cost of the top end enterprise server designed to scale I/O, cpu cores and memory based on the technology of the day with very high RAS features – which comes at a premium. T5’s are entry level servers using current technology for the processors and memory. They glue them together to scale from 1 to 8 sockets. They don’t even say which T5 server? -2? -4? -8?
  • For the European Telco – How many x86 servers? What generation of processors? How many T4 servers? What model? Number of cores, etc? Easy to make claims – He said “This car will go 0 – 6 mph in under 5 sec and get 35 mpg in the city!”.
  • Solaris is 85% less costly than a typical x86 / VMware / Red Hat solution. Here is another Oracle marketing tactic which is to shift topics around. The video is on the economics of Solaris so they mix cost efficiencies of Solaris for SPARC and Solaris x86 leaving the viewer the option to interpret what they say for one (x86) must be the same for the other (SPARC). What they lead you to believe is that Solaris virtualization is comparable by “features” to VMware but nowhere close in cost with OracleVM being considerably less expensive. I am no VMware specialist but what I do know is that Oracle is using OS based virtualization called Solaris Zones, which is included in the OS. That is how they get the cost to $0. On a feature comparison though, VMware is much more robust and feature rich than LDOM’s and Solaris Zones which they generically label both as OracleVM.
  • I don’t dispute the VMware overhead. VMware and its users seem to dispute it but that isn’t the purpose of the comment. We can discuss this in a future blog. They claim the Solaris virtualization is virtually 0%. For which product? They mention Solaris Zones but they also have LDOM’s now called Oracle VM. That has a different overhead amount that requires VM’s for Control and I/O so leaving the reader to believe virtualization is the same between platforms is disingenuous. This snapshot from the SPEC website for SPECjEnterprise2010 shows quite a difference in performance results for a 128 core T5-8 server with one using Solaris Zones and the other OracleVM, presumably LDOM’s. Of course, Solaris has Zones for both x86 and SPARC but they are apples to oranges when compared to VMware for features and functionality. Using the Oracle standard for financial disclosure I will have to do a future blog comparing a Power8 server with AIX using Workload Partitions, which are similar to Solaris Zones. Unlike the x86 server there is still an underlying hypervisor that would allow for separate VM’s with their own OS instance all without having to pay for a virtualization suite
Looks like overhead when using Solaris Zones

Looks like overhead when using Solaris Zones

I’ll pick up part 2 of this “car buying” journey.  You’ve already watched the video and hopefully with my comments pointing out how Oracle massages, manipulates, mis-states and generally exaggerates results or capabilities you are hopefully becoming a skeptic of everything they say.

I’m putting this together a bit faster than I intended but I have a tweet to my new found buddy Phil Dunn of Oracle that I want to hit “send” on so I’m sure I’ll make a few tweaks, updates and corrections.  My intention is to accurately capture their mis-statements without making my own. I’ll correct the record though.  Will they?