Kirk and Friend’s Rants on Mission Critical Computing

How Thick is Your OS?

Kirksblog — Thu, 18 Jun 2009 21:55:00 GMT

HP had a product announcement last month and it's one that I think was pretty exciting. BladeSystem Matrix is a new offering from HP that pulls together compute, storage, networking integrated with a co-designed management, automation, and orchestration software stack. From dramatically simplified ordering to installation on the order of minutes, every aspect of delivery has been aligned with the goal of allowing customers to purchase solution ready infrastructure instead of piece parts.

In looking at the combination of software and hardware that HP has pulled together to create BladeSystem Matrix, I was struck by some familar problems of providing an agile infrastructure. I look at BladeSystem Matrix and see many of the features and capabilties that HP has been delivering in our HP-UX, OpenVMS, and NonStop environments for a long time. But those environments were created in an era of highly vertically integrated vendors, and the world has more options that that today. To my mind, this exposes a tension between how we've done things and how we might see them evolve.

How thick is your OS?

When I look at the co-designed software stack of BladeSystem Matrix, I see functions for security, resource management, virtualization, availability and workload flow control. In the OSes running on the hardware I see all those same functions along with libraries and APIs. In fact, if I look at what's becoming one of the most important workloads running under that OS, a hypervisor, I'm seeing many of those same functions replicated yet again. At this point, there's an 'embarrasment of riches' argument, that all those cycles replicating functions would have been wasted anyways, so why not use some of them to allow you to get a fraction more efficient?

Now don't get me wrong. A ubiquitos API and a rich library of routines is critical to making sure that we have a steady stream of software developers who can ramp quickly and stay productive for a long career. That's not going to change and there are some great examples of "the next big thing" that failed to live up to expectations because they didn't allow a world's worth of mortal programmers to stay productive.

But what about those other functions? It's not just a problem of replication, it's also a problem of lack of perspective. Should we expect an OS inside of a virtual machine inside of a blade inside of an enclosure inside of a rack inside of a data center inside of an enterprise to make the right power versus performance tradeoffs? The same OSes that we used to expect to discover and manage a whole system are now being utilized unchanged to manage a small part of a much bigger whole and that causes problems.

What this has me wondering about is the potential to offer substantially thinned out versions of today's operating systems where we retain the APIs and libraries that allow programmers to remain productive and seperate those from the resource management. There are existing examples here from the world of high performance technical computing with thinned Linux distros, but in this case we'd need to replicate the same levels of reliability and availability as traditional OS models.

It could be that as we continue to innovate at the intelligent infrastructure level that for the OS of the future, thin is in.

HP-UX 25th Anniversary

Martin Fink — Thu, 18 Dec 2008 17:17:00 GMT

As we come to the close of 2008, I wanted to send out a special thanks to our HP-UX customers. To do this, I want to start by looking backwards to 1983. 1983 was an amazing year:

Cabbage Patch Kids were released
Gandhi wins Oscar for Best Picture
Gas was $1.24/gallon, Stamps $.20
The last episode of MASH aired
Margaret Thatcher wins landslide victory in General Elections in the UK
Compact disks are introduced

And most importantly – HP-UX was released. For 25 years, HP-UX has been driving IT efficiency. HP has strived to provide superior performance, availability, security, manageability, quality, buying experience and lower total cost of ownership. Did you know that: • HP-UX has shipped on over 1,000,000 servers• HP-UX continues to be a leading Unix for revenue in the fast growing economies of Latin America and Asia• HP-UX outpaces the market in high-end revenue market share− HP ships more Superdomes with HP-UX to China than any other country in the world• HP-UX is the preferred platform for data warehousing and business intelligence And we haven’t stopped. We have seen over 5,000 new HP-UX 11i applications in past 5 years. HP-UX 11i v3 provides a performance boost of >35% over v2. The HP-UX public roadmap extends beyond the end of the decade. The important fact is this is the longest future public roadmap of any operating system in the industry: Longer than Windows, Linux, Solaris and AIX We have had 25 years of HP-UX innovation with multi-billion dollar investment for the future. Thank youCelebrating 25 years is a major accomplishment and I want to personally thank you for making HP-UX what it is today. Without you, our customers none of this would have happened. HP-UX 11i has been the foundation for meeting your business critical computing needs! It is an industry icon for yesterday, today – and most importantly tomorrow. If you thought the first 25 years were cool, just wait until you see what’s next. See special 25^th Anniversary of HP-UX 11i website at: www.hp.com/go/hpux25th

Thank you again,

The Windows, Business Intelligence and Itanium Processor Story

Martin Fink — Tue, 09 Dec 2008 18:08:00 GMT

For a while I have been talking about the state and direction of UNIX in the industry. Now, I would like to change gears. I’m often asked is there truly the need for Microsoft Windows and SQL Server to be running on Itanium-based systems such as HP’s Integrity servers. The answer comes down to the type of application you are looking to support. For business intelligence there is a huge need.

Today’s organizations are looking to providing business insights to all employees leading to better, faster, more relevant decisions on business critical matters. As such IT departments are facing a continuous increase in the need to gather data and provide services for storage, reporting, and analysis of this data. This is vital to stay competitive. This creates 3 unique challenges that are well suited for Itanium based platforms.

First: Availability

Business intelligence applications are becoming increasingly embedded in the fabric of your business. As such you can’t afford them to be down for even a limited amount of time. Every time you lose access you lose the ability to react quickly to market or operational changes. As such, availability needs to be built in from the processor level all the way to the data center to ensure the availability of the data warehouse. This includes incorporating RAS features in the processor, building redundancies in the systems and clustering across the data center.

Second: Scalability/Performance

You can see the importance of business intelligence in the industry just by looking at the amount of stored data. Did you know that the amount of stared data almost doubles every 12 to 18 months. In addition to the data, there is an exponential growth to the number of users that require access to business intelligence in their daily activities. As such, throughput and performance become important. Itanium-based Windows systems provide the single system horse power and scalability not present on other platforms. The largest systems support up to 2 TB of memory and 64 cores which enables them to support the bigger data warehouses. One side note: Today we support 64 cores with a single instance of Windows, but we recently demonstrated support for 256 logical processors with an Integrity Superdome and next generation version of Windows Server and SQL Server. Customers will be able to support these next versions of MS software and take advantage of the full capacity of Superdome without changing hardware.

Third: Cost

At the same time, today’s economy puts IT departments under constant pressure to deliver the right information at a right time and very importantly -- for a right price. As a result, IT organizations are looking to consolidate and automate their environments. They want to:· Centralize of all components of the BI solution, which comes standard with SQL Server, including servers. · Vastly improve integrated solution of SQL Server on analytical and integration side· Consolidate physical servers and storage with the same application types on fewer but larger systems · Consolidate data from multiple sources and formats into a single integrated solution· Consolidate different types of workloads on systems · Utilize storage systems that are few in number but large in size This creates a need for larger, mission critical servers and storage such as Itanium-servers.So let’s readdress the question: is there the need for Windows and SQL Server on Itanium-based servers? Depending on the application, the answer is absolutely yes. Business intelligence is the prime example. Business intelligence is becoming a vital component of most organizations. You could say that it is mission critical. As such, availability, scalability/performance and cost are being key attributes of the infrastructure. These attributes are the strengths of mission critical servers.

The UNIX Paradox - Innovation

Martin Fink — Tue, 04 Nov 2008 20:41:00 GMT

In previous blog entries I’ve discussed the need for stability in a UNIX operating environment. Let me be clear here, this in NO way means that to stabilize we have to stop all innovation.

The major UNIX operating systems have been around for a while. HP-UX is coming up on its 25^th anniversary. The first version of AIX came out in 1986. SunOS was initially released in 1983 and followed by Solaris in 1992. Those 20+ years have enabled each of these operating systems to include a large amount of the key functionality into their respective environments. So after 20+ years are we done? No. To start, there some basic needs for continuing to enhance the operating system. For example, operating systems need to be enhanced to support new processors or memory, new I/O technologies or even a new security standard.

In addition to the basic requirements that make it necessary to continue to enhance the operating environment, I believe there are a couple of areas that users are looking for further innovation in. These focal points for innovation are ease of use and lower total cost of ownership (TCO).

1. Making the operating environment easier to use.

Ease of use starts from the initial ordering of the operating system to the deployment and finally ongoing support of the environment. First, operating systems should be bundled with the functionality and the software that the organization requires. By prepackaged functionality, you are ensuring that the solution matches the customer needs, simplifying the installation process, and thus reducing the overall risk of an upgrade. Secondly, as much of the installation process should be as automated as possible and utilize self discovery mechanism to incorporate the new image of the operating system into the current environment. With new package and automated deployment, you can make the UNIX environment significantly easier to set-up. For example, with the new packaging offered in the HP-UX 11i V3 operating system we are able to reduce the number of installation steps to 9 and the number of reboots to 1. Once the environment is up and running making the environment easy to manage becomes critical. The more manual the process is the more complex my UNIX operating environment is. So for starters, I require a set of tools to manage my environment such as the tools offered in the virtualization suite (VSE), in the management area with Systems Insight Manager (SIM) tools and with the disaster recovery capabilities in Serviceguard. While these tools make it easier, the key to innovating is to make the different capabilities and integrated and automated as possible. Such that the management tools interact with the virtualization tools which communicate with the disaster recovery capabilities and so forth. By integrating and automating the time to manage is lessened and the risk of a problem due to an operator error is minimized.

2. Lower Total Cost of Ownership (TCO)

In today’s environment IT departments are being asked to do more with less. They have to justify all of their expenses. So how a new environment impacts the bottom line is critical and anything that can be done to lower the total cost of ownership is seen as a plus. There are a number of ways the total cost of ownership can be impacted.

A. Increase the capabilities of the system (i.e. increase performance so I can get more out of my machine)

B. Maximize the utilization of the system (this prevents having to buy additional systems while I have capacity sitting idle)

C. Automate the system (redeploy operations staff to higher payback projects that are no longer required to maintain the environment)

D. Reduce the amount of planned downtime

E. Make it simpler/easier to deploy and manage

Most vendors will focus solely on the first item by raising performance. Unfortunately, this overlooks the fact that labor costs are one of the biggest expenses in the data center. As a result, vendors truly need to focus and innovate around automating as much as possible and that which isn’t automated needs to be made simple. In addition, the focus needs to be on how I get the most out of the environment (maximum utilization and availability).

As I said at the beginning, I still believe there is the need for progress/innovation within UNIX operating systems. Based on the state of the world today, the biggest thing the UNIX vendors can do for their customers is focus on making the systems as easy to use as possible and lowering the total cost of ownership.

Mainframe Claims

Martin Fink — Mon, 13 Oct 2008 19:14:00 GMT

In looking at IBM’s website I have noticed a number of claims about our product and their offerings that frankly aren’t true. As such, I wanted to give you my thoughts on what IBM is claiming:

IBM Claim 1: z10 mainframes can outperform HP Superdomes on power consumption by a factor of 2.5x or greater.
Faulty assupmtions: Per IBM, 1 mainframe is equivalent to 4 64-way Superdomes, with no reference to source given.
Reality: HP estimates that a fully loaded HP Integrity Superdome is 33% more powerful than a fully configured z10 EC. That changes the equation a bit, doesn't it!!

IBM Claim 2: There is no mainframe skills shortage
Faulty assumption: Mainframe reduces IT staff requirements over "distributed" environments. I honestly can't believe they are serious here. Any time IBM uses "distributed", their basis for comparison is servers that are 5 years old, grossly under-utilized (rougly 5%), non-virtualized and running one application per server. This is the only way IBM can show TCO saving versus the mainframe. IBM also touts "thousands of mainframe-trained graduates", but includes those that have taken just a class or two. Are these the people that companies want to run their mainframe?
Reality: Virtualized environments and the tools from HP reduce IT staff and enable new employees to come up to speed quicker. On the staffing front, analysts say that IT stafffing for legacy systems such as the mainframe is an area of caution.

IBM Claim 3: NonStop and Mainframes have the same availability
Faulty assumptions: Used general system availability used by Gartner to show parity.
Reality: Actual customer environments, as measured by Standish Group, show that NonStop exceeds mainframe availability even if the mainframe is clustered for higher availability.

IBM Claim 4: Mainframes have better TCO than HP Integrity
Faulty assumptions: The study used was one "conducted by IBM", but does not show the basis for comparison to the HP Superdome.
Reality: If IBM really wanted users to compare system performance, they would use industry standard benchmarks to do so.

My advice is don't take what they are saying for granted. Kick the tires and look under the hood to make sure the claims are actually true.

The UNIX Paradox - Stability

Martin Fink — Mon, 15 Sep 2008 22:17:00 GMT

I completely understand the desire for stability in one’s production environment. If an environment is performing activities critical to my business (booking orders, controlling my manufacturing facility, hosting a data warehouse, etc.) and it has been doing it successful for months or years why would I want to take the risk that comes from changing something/anything. Yet there are times that I need to make changes. I may need to add the most recent platforms to accommodate user growth, install the latest security fix, or take advantage of the latest Integrity VM capabilities. So how do I do this while keeping the environment stable?

1. Minimize the number of big releases.

Recalling the past, customers would never touch a .0 release because they always wanted someone else to do it first. Today we've enhanced the process with better testing and there is the ability to roll out new functionality that can be picked up anytime along the way…the customer's way - less disruption with improved release processing. Big changes make users nervous. They don’t want to implement anything that will destabilize the environment. The more change, the more risk in a customers mind. Some vendors will continuously come out with new versions of the operating environment and force customers to adopt it. I personally believe, however, that it is better to make major version changes less frequent and then provide customers with new functions and fixes along the way in smaller bundles. In addition, customers are looking for predictability in the patches/enhancements so they can plan for it. Some of the more recent operating environments release security patches almost every other day driving customers nuts.

2. Binary compatibility

One of the biggest challenges in moving to a new version is the time, effort and risk in recompiling and certifying applications. This is a big reason for staying on older versions unless there is some compelling benefit. If, however, the environment is always maintained independent of any change to the release, then the customer’s solution is much more stable moving forward.

3. Ala carte

Customers today like to have it their way. Users don’t want to be forced to newer releases. Users don’t want to adopt new functionality if they don’t need it. In other words don’t change just for the sake of changing. Instead, let me pick the changes that I want. The key here is to give the customer the option to pick and choose which patches and enhancements they want to apply to their environment. Don’t force them to move to the latest environment just to make it easier on the vendor. On top of this you add the concept of bundling that makes ala carte selection easier for the customer. In other words, give them “all new features”, “all fixes”, “all changes to enable new hardware”. This ease of selection makes it much easier for customers to adopt and deploy, and most importantly gives them ways to manage the risk. Additionally, provide customers tools that help them understand whether they’re current on security fixes, etc. In short – ala carte, ordered and served in many different ways!

4. Longer support lives

Often vendors will push customers to the newer release by cutting the support for previous releases. Instead customers that aren’t ready to move or don’t see the need to move want to be supported. They want to ensure that their environment is protected.

The combination of these four items is what customers are looking for when they want to stabilize their environment. If I’m missing something please let me know. These are part of the key guidelines we drive towards in the development of our operating environments.

Don’t, however, be mistaken that by focusing on stabilization I am neglecting innovation. Innovation is key to the future of our operating environments and in my next posting I will discuss areas that I believe still need progress within the UNIX operating system.

The UNIX Paradox

Martin Fink — Wed, 27 Aug 2008 22:43:00 GMT

Over the past few weeks I had the opportunity to visit a number of UNIX customers. In my discussions with them they brought up differing needs. There is a set of customers that are looking to stabilize their environment. This type of customer does not want to have to go thru a transition, having to recompile applications, or even retraining. As the old adage goes “don’t fix what isn’t broken.”

Other customers are looking for new innovations in their operating environment. Whether it is supporting new technologies, added security or even making easier to use there are new features that they want or need.

This creates a major balancing effort for UNIX vendors between stabilization and innovation. Which side of the scale that a customer is focused on often depends where they are in the application life cycle. If I’m just bringing up a new application, a new version of the operating system or even a new operating environment, I’m most likely to want more innovation/new functionality in my operating environment. As time progresses and my application environment and/or operating environment has been running for a while and I don’t want to change. We have all heard the stories of customers that have had their environment up and running for years.

So for a UNIX vendor this creates a paradox. How do I create an environment that is stable while continuing to innovate? Over the next couple of posts I will give you my thoughts on both sides of the scale. I will first discuss how I believe we can provide customers with a stable environment and what is required to do this. Then we will talk about how you can continue to innovate on a stable environment and what types of innovations are still needed for the UNIX environment.

I encourage you to post your thoughts. Which side of the scale do you fit on? What are the important areas you are looking for from a UNIX environment.

Contributions

Martin Fink — Wed, 30 Jul 2008 17:23:00 GMT

If you didn’t catch the press a little while back, HP announced the contribution of the AdvFS source code base to the open source community. A lot of times I get asked "why doesn’t HP open source software XYZ". The reality is that open sourcing stuff is really hard. In many cases we have third party licenses that won’t allow us to do that. In other cases, the code is structured or documented in such a way that would make it very hard to release in an atomic way. And in other cases, we just don’t think we can build a community so it doesn’t make sense. In this case, a few smart people in HP saw a huge opportunity. We had a code base that HP wasn’t going to use. We owned all the code and the IP. We did the work to make sure there wasn’t anything in there we couldn’t release. Finally, the need for IP in the open source community that would help deliver a robust, mission critical file system was there. While it would be possible for someone to do a port of AdvFS to Linux directly, that wasn’t really our intent. We wanted to share "tricks of the trade", and also release a huge amount of prior art. In the end, we hope this accelerates the delivery of a new, modern and mission critical file system for Linux. I know I’ll be impressed with what the community delivers. I can’t help but to comment on the fact that we released the code under the GPLv2. We didn’t create a new license, and we made sure we were Linux compatible. Take a hint Jonathan….

Convergence

Martin Fink — Wed, 30 Jul 2008 17:18:00 GMT

IBM recently announced their latest quarterly results with what looked like impressive growth for their P-Series systems (29%). What they made sure they weren’t too overt about was that this growth was for "converged System P". See, a few month ago, IBM decided to combine their I-Series (AS/400) with their P-Series systems. So now, this seemingly high growth is actually due to the fact that they add the count for all those systems that used to be i-Series and add them to P-Series to make it look like they have really high growth. Cute marketing stunt. Just wanted to make sure the facts were a tad more visible to you than IBM wanted you to see.

Hidden Freedom

Martin Fink — Mon, 28 Jul 2008 21:23:00 GMT

A few month ago at HP Technology Forum in Las Vegas we announced the launch of our new Integrity Nonstop Bladesystem offering. Twice the performance, half the footprint, 100% Nonstop. As part of our launch we also announced a marketing program to give away a free Nonstop to customer who want to get off of a mainframe. You can get the details here: NonStop FREEdom. I don’t want to rehash what we launched, but since we did this, I’ve gone to a number of customers and it seems as though IBM has been offering a few folks some free mainframes. The problem is that I can’t find any public marketing program that talks about these free mainframes. It seems you might have to specifically ask for one. At least we made our Nonstop FREEdom program a public and visible program. In any event, if you happen to need an IBM mainframe, and you’re not getting yours for free, make sure you push a little. Maybe we need to provide you a competitive quote to help you out.

Let’s get started

Martin Fink — Mon, 28 Jul 2008 21:19:00 GMT

Hi there. Welcome to my blog: Martin’s Musings on Mission Critical Computing. For those of you who don’t know me, I run what we call “Business Critical Systems” at HP. Whenever you think of high-end and mission critical computing, it’s most likely my team that takes the lead for HP. My goal with this blog will be to comment on events and happening in the mission critical computing space. I’ll share my thoughts on current events, and where I think things are heading. I’ll try to avoid blatant marketing stuff (we have other venues for that), but I’ll make you tolerate a few pokes at my main competitors (hopefully in a fun and lighthearted way). If for some reason I feel the need to take a break from my day job, I may also comment on a few of my main hobbies: Home Theaters and Digital Multimedia creation. I expect disagreements along the way, so feel free to share. Blogging is about sharing, sometimes we’ll agree, and sometimes we won’t. – Martin