Virtualization? What’s that? And why? (Update!)

By , June 23, 2006 05:42

Jörg did provide a translation of his article. You can find it here:

Does virtualization solves any substantial problem?

Virtualization? What’s that? And why?

By , June 22, 2006 06:24

Colleagues did start a discussion on the usefulness of virtualization technologies, as available today. Sadly, it’s in german only, so I will summarize here, after providing the links to the blogs:


Ingo states, that virtualization today is mainly seen as hardware or server virtualization, and refers to a couple of articles (mostly in english!), that state, that this approach add problems, and does not solve problems, because one needs to learn the abstraction layer also, which adds to the complexity.

He continues, in stating, that this also opens new problem areas, like billing, but also adds complexity to change-, configuration- and release-management. Problems, for example in the billing area: Who does pay for the ressources needed for the virtualization layer?

Jörg adds to this, and states, that Ingo’s approach in describing the problems is too honest. Jörg simply states:

Virtualization as done by tools like VMware simply replace a box by a virtualized box. This solves nothing really important (and, we all know, HW prices keep falling, and only are about 20% (mostly even less!) of the overall costs of “a project”).

So, he continues in stating, that the real problem is provisioning (see my older entries) of SERVICES, and not the SERVERS.

And, once you have put a “model” around the description of your service (aka: you have virtualized the service), you no longer need server-virtualization (or: you can benefit more! You can make better use of the HW), because you are then able to quickly “re-deploy” your service (and not the server!) from one system to a different system (and that can even be a different OS, different CPU-Architecture, you name it!).

Thanks to both, Ingo and Jörg for starting this discussion!

5 Nines?

By , June 22, 2006 05:26

Yesterday I was invited to talk to the “AK Server und Betriebskonzepte” (Work Council Server and Operatingconcepts) from BITKOM on 99.999% availability and how to achieve it.

As this is a cross-industry council, presenters from all other major IT vendors were also present.

We all agreed on a couple of major points:

  • 99.999% availability is achievable
  • 99.999% availability is NOT a technical problem
  • 99.999% availability REQUIRES cooperation/including of people/processes

And, most importantly:

  • Do not talk about 99.999% availability, talk about SLAs, etc.
  • Do not talk about technology, do talk about business impacts, procedural requirements, etc.
  • Do NEVER believe, technology alone (regardless of the technology!) delivers 99.999% availability.

As this was common sense among the participants, the team still considers it a requirement to educate deciders, non-IT-management, consultants on these simple facts. That’s why I added this entry to my blog… 😉

Support? What’s that?

By , June 20, 2006 08:41

We all in IT do have a phrase, that’s creating lots of headaches:

Not supported!

Still, no-one really knows, what this implies or means. Or: Everybody has his own interpretation/meaning of this term. So, let’s try and start to define, what the term “supported” means, what it implies, and what we all really want. Because clarification helps communication, and that leads to more relaxed conversations.

Support, as defined (today) in wikipedia:

Support may refer to the following:

  • Support (mathematics)
  • Support (mobile framework), in mobile computing
  • Support (technical analysis), in security trading
  • Military combat support (see combat engineers, anti-tank, artillery)
  • Military service support (see combat medic, military intelligence, military logistics)
  • Sympathy (emotional support)
  • Supports in engineering and construction include arch, beam (structure), column, balcony

From that, we all see, that the thing, we think about, isn’t even mentioned here!

Let’s start with mathematics (because that is the only exact science, right?):

In mathematics, the support of a real-valued function f on a set X is sometimes defined as the subset of X on which f is nonzero. The most common situation occurs when X is a topological space (such as the real line) and f is a continuous function. In this case, the support of f is defined as the smallest closed subset of X outside of which f is zero. The topological support is the closure of the set-theoretic support.

In particular, in probability theory, the support of a probability distribution is the closure of the set of possible values of a random variable having that distribution.

Now, we even know less then when we started, right? So, let’s try and find a different one:

Webster states:

Entry Word: support
Function: noun
Text: 1 something that holds up or serves as a foundation for something else (if you don’t add a couple more supports to that tower of blocks, it’s going to fall down)
Synonyms brace, bulwark, buttress, mount, mounting, shore, stay, underpinning
Related Words column, pedestal, pilaster, pillar; arch, bracket, cantilever; crutch, mainstay, peg, post, stake, stanchion, stand, stilt, truss; base, foundation, frame
2 an act or instance of helping (the team’s victory owes a lot to Joe’s strong support in left field) — see HELP 1
3 something or someone to which one looks for support (Grandfather has long been the extended family’s emotional and financial support in times of trouble) — see DEPENDENCE 2

Entry Word: support
Function: verb
Text: 1 to promote the interests or cause of (my parents support the local schools both by volunteering and by fiercely opposing funding cuts at town meetings)
Synonyms advocate, back, champion, endorse (also indorse), patronize
Related Words adopt, embrace, espouse; abet, aid, assist, prop (up), second; bolster, boost, buttress, reinforce; bail out, deliver, rescue, save
Phrases stand up for
Near Antonyms baffle, foil, frustrate, interfere, oppose, sabotage, thwart; desert, disappoint, fail, let down
2 to pay the living expenses of (a young widow supporting a sick mother as well as two small children on a teacher’s salary)
Synonyms maintain, provide (for)
Related Words finance, fund, stake
Phrases foot the bills for, take care of
3 to hold up or serve as a foundation for (pillars supporting the bridge)
Synonyms bear, bolster, brace, buttress, carry, prop (up), shore (up), stay, underpin, uphold
Related Words steady, truss, underlie
4 to continue to declare to be true or proper despite opposition or objections (we support the students’ right to speak out on local issues that affect them) — see MAINTAIN 2
5 to give evidence or testimony to the truth or factualness of (her grades don’t support her claim that her after-school job isn’t affecting her grades) — see CONFIRM
6 to provide (someone) with what is useful or necessary to achieve an end (sent reinforcements to support the troops already in the thick of battle) — see HELP 1
7 to put up with (something painful or difficult) (he could never support the thought of having to go on living without his beloved wife at his side) — see BEAR 2

So, just for a starter, let’s use definition 6 of webster’s verb:

to provide (someone) with what is useful or necessary to achieve an end (sent reinforcements to support the troops already in the thick of battle) — see HELP

So, it boils down to “help”. Should it be that simple? No, it is not, because there are a few constraints attached to that:

  • In order to help, you need control
  • In order to help, you need ressources

In IT speak, this means: You need to “own” the stuff, otherwise you can not support. You need a contract, because otherwise you can not get ressources.

The most critical part is the “control” part, because here, most mis-understandings occur.

Some (mostly software) companies claim support, if they allow the usage with something, they do not control. That’s clear and good, because otherwise, they would never achieve broad adoption (Example: Microsoft Windows). Some other, more system-oriented companies only claim support, if they control the complete stack (example: Apple MacOS X, runs only on Apple Hardware, they even thought about putting in hardware to PREVENT mis-use).

Still, for the average end-user, these differences are not transparent, because they all use the same word: “support”.

So, we all should be more precise in USING that term. It might be good to replace that phrase appropriately by things like:

  • will (not) work
  • (not) allowed
  • might (not) work
  • (not) tested
  • (not) certified

And always add, which pieces will be covered by a support contract, because these are only the pieces, that are “owned”, and can therefore be “patched” (which opens a different can of worms), or maintained.

With that: Floor open for discussion!

Consolidation, a challenge!

By , June 19, 2006 04:55

Last week I was talking to a customer. He has a big problem.

The topic of the meeting has been consolidating, and benefitting from Oracle’s licensing w.r.t. Sun. So far, the meeting went well, and then, during a small break, customer described his problem and asked for solutions to the following:

In his datacenter and under the desks in the offices, he often finds (and: Don’t we all do that?) small systems, running for years, some even decades, and doing real important stuff. These systems have never been part of the “real” datacenter, but still, can not be replaced, because they are vital parts of the overall environment.

Now, he faces the following problem:

These systems, as stated, are old, some are still Intel x386 boxes (mostly running Linux apps!). These systems are not fully loaded, but seem to run at some 20-30% usage (which we all know, is average, even today!). So, replacing these systems one-by-one with current boxes will not really yield a better CPU/performance ratio, because then these new systems would be used in the below 1% range. That’s a waste of resources, regardless of the fact, that such systems do cost way below 1000$ today.

So, he is looking for solutions to consolidate these systems, and have a proper way of doing accounting on the different workloads put onto these systems.

I simply stated, that we can do that, and he agreed to have a different meeting later this summer. Still, I’d like to outline, what we will be proposing:

  • With Solaris 10 we have Zones/Containers.
  • These can be put under control of the Solaris Ressource Manager and the Fair Share Scheduler.
  • With Solaris accounting tools, augmented by tools like TeamQuest, there are really flexible ways to generate accounting infos that can be processed by standard billing systems.

So, bare in mind: Consolidation does not always mean: Putting smaller systems on big systems. It can also mean: Making small systems even smaller!

So, small systems can now become even smaller, and I leave you with my initial answer to this customer:

“I do have an answer, but the initial statement might not be, what you expect:

The problem is not the technical solution, the problem is the acceptance of ‘partitioning’ OpenSystems. In the Mainframe world no-one really worries or even thinks about “sharing” pieces of the OS, or the hardware with a different business unit. In the OpenSystems world, no-one today really trusts this. With the certification of Solaris 10 as trusted, at least, we have the necessary technology and approvals to do just that.”

So, it’s up to you to start consolidating all your “long-forgotten” iron… 😉

Happy Birthday, OpenSolaris!

By , June 14, 2006 12:45

OpenSolaris 1 Year Anniversary

What shall I say?


I’m now working at Sun for more than 8 years, and ALWAYS did use Solaris/x86 on my laptops. Starting with Solaris 2.5.1, quickly updated to Solaris 2.6 on a Toshiba 480 CDT (still running on it!), then Solaris 2.6, 7, 8 AND 10 on my Toshiba Libretto 110CT (yes, I have Solaris 10 03/05 up and running on that small box, with only 64 MB of RAM, and only 233 MHz! When I find more time, I will report here, how I did it (Thanks, Casper!)). Then Solaris 8, 9, and 10 on a Dell Inspiron 5000, and now Solaris 10 and OpenSolaris (dual boot) on my Toshiba Tecra M2.

Before I joined Sun, I also did use Solaris/x86 on a couple of servers at the company I have been working for, so, honestly, Solaris was one of the reasons I joined Sun.

During all these years I did buy some copies of SuSE Linux, but never really used them. I found Linux too much of a hassle in patching and keeping it up-to-date. So, SuSE got “money for nothing” from me… 😉

Although it sometimes was difficult to get all work done under Solaris, today, I don’t really have anything missing (OK, I have to admit, we all MISS Adobe’s Acrobat Reader for Solaris/x86! Johnny L., fix that!). That can easily be seen at the messages, I get, when I sometimes, just for fun, boot the laptop into Windows. I always do get the “your virus protection definition file is out of date for more than 2 weeks, please update”…

So, please all, let’s get up, and sing:

Happy Birthday to You, OpenSolaris!


Wagner 2: Lohengrin, Nagano, Lehnhoff, Baden-Baden

By , June 8, 2006 05:20

Yesterday we spent the afternoon in Baden-Baden, enjoying the highly acclaimed and praised performance of Wagner’s Lohengrin by Nikolas Lehnhoff, conducted by Kent Nagano. This was the third and last performance in Baden-Baden, I don’t know, if the participating operas (Lyon (Opera National) and Milano (Scala)) will host this performance also. If yes, go there! Or if any of these performances will lead to a DVD (which I would welcome!).

We again did prepare us by watching and listening to older videos and CDs of former performances, and by also listening to the introductory lectures given by Stefan Mickisch. We can highly recommend his CDs, they are fun listening to, and do provide a profound introduction to the theme and the music. The sad thing is: They are german only. But, those who love Wagner should also be able to understand german, at least a bit… 😉

We also have to report some sad news: Yesterday, both Waltraud Meier and Klaus Florian Vogt (Ortrud and Lohengrin) were sick, and were not allowed to sing. Waltraud was replaced by Anette Bod, and Klaus Florian did the acting, and Stuart Skelton did the singing from the right side of the stage. Stuart was the Parsifal in Frankfurt (check my former blog), and we both agreed, that his voice was much better yesterday during his performance of Lohengrin, then during his performance of Parsifal. Although I’m not a fan of Waltraud, my feeling w.r.t. the performance of Anette was: She was “shrieking” to much. OK, this is also attributed to the role, but still, I hadn’t been completely convinced by her performance. But we missed Klaus Florian most, because he was praised that much (and I saw small snippets of his performance in TV shows in the last two days, which were really convincing!), and we were eager to listen to his performance. But: People (and singers) are human, and sickness can happen to anybody.

Putting these sad things aside, let’s look at the overall thing:

Highly recommended!

First: Kent Nagano made his orchestra, the Deutsche Symphonie-Orchester Berlin, perform clear and precise. I was able to hear things. I never heard before! This also made hear-able discrepencies, and tensions, that some other performances try to neglect.

Second: The setting was more modern, and as such allowed to concentrate more on the emotional problems the characters have (Elsa, for one: The driven, the victim). This also reflected in the movements of the people (not that much of “showing”, but more “human”).

Third: Singing was great, best yesterday was Tom Fox as Friedrich von Telramund.

Forth: The scenery fitted well with this “modern” setup. They placed a chair (backward facing) and a lecturn on a round plate, surrounded by an amphitheatre, so that all people “talking” were standing behind the lecturn and “announcing” their things and thoughts. This made the impression of a university setting from the 19th century. This was convincing, as we are dealing with a “high court”. The second movement had a stair-flight, which fitted well with the actions, and the third movement returned to the amphitheatre setting.

Short: We loved it!

Open? OpenStorage

By , June 6, 2006 09:40


long time ago, but still not widely known… 😉 So I need to remind you:

Open Storage Program for SunCluster (OSP)

Around christmas 2001 (yes, really, that long ago), I did ask a simple question on a Sun-internal technical (not sales!) alias dealing with SunCluster. The question was something like:

How much more Sun Servers can you sell, if we would simply say “YES” instead of “NO” to the question of SunCluster in connection with non-Sun storage?

Within a week, I got listings of opportunities summing up to a 3 digit million dollar number (and remember: servers were more expensive in those days).

That then simply started the OSP. Today that OSP covers nearly every major storage vendor, so, SunCluster can be used in conjunction with nearly every thinkable storage solution.

And, what’s more: These configs have higher value, because all involved parties did testings, and certifications, and do support these configs.

So, again: Just check:

Open Storage Program for SunCluster (OSP)

Provisioning: Sun Grid, Outsourcing and Outtasking…

By , June 6, 2006 06:05


continuing the last entry on provisioning, there are also different currents in the field.

Some time ago I visited a large bank and insurance company’s IT service provider. They have the idea to outtask everything, that does not offer value and that does not contain their IP. The interface to the delivering partners shall be handled by agreed-upon blueprints. So, the pieces that will be delivered to the datacenter will be preconfigured systems, including software from different parties (up the stack from OS crossing middleware up to the app-server tier), configured to the specifications of the IT service provider.

These blueprints shall be defined and updated on a regular basis by the IT service provider and the delivering parties.

So far, this sounds highly reasonable, and has been best-practice for a couple of years (even across the whole industry, not only this customer).

Still, I see a couple of problems with this approach:

  • The delivering party gets large control of what the IT service provider can do, because the delivering party can simply state: “Can not be done.”
  • The process of defining the Blueprints is repetetive, and every single repetition gets more and more boring. And could be automated also… 😉
  • It does not reflect the current market trends (if you believe our SunGrid vision), because these blueprints define the building blocks of an internal grid, and the process of putting the IP-laden apps onto these building blocks is partially done by the delivering party, and partially done by the IT service provider (and is not part of the blueprint), which does not give a consistent, exchangeable, replaceable, repeatable way of operation.

If we look at what the trends are, I would recommend a different approach (which other parties have already also started years ago. I for example was hired as a technical consultant by a different bank helping them in keeping their RCM (Reliable Configuration Management) system up to date. This RCM is still in place and works perfectly for this customer even after nearly a decade of operations):

  • For the hardware, use so-called staging centers. These are facilities of the delivering party, where all components are pre-assembled, put into racks, and loaded with the delivering parties’ software (mainly: OS)
  • If you want more, you can replace these by getting fully configured application environments, using Sun’s DCRI (Data Center Reference Implementation), with the accompanying services SRDCRI (Sun Ready DCRI). Here, complete environments are preconfigured including custom apps.

These two solutions underline the classical approach of outsourcing/outtasking, but still contain the risk of not being “flexible enough”, or “vendor influence”. Still, they add a lot of value, as they take out the risk of doing everything yourself, and the vendors do support these setups. And, as the vendors do this often, the benefit from automation is paying, will say, it’s cheaper, then doing it yourself.

If you want to have more control or more flexibility, you need to take the driver’s wheel into your own hands, and might consider provisioning. This can be additive to the above, or could also be a replacement (we might offer DCRI capable modules for N1 SPS, for example. JET, our utility for augmenting the JumpStart network installation method, is already part of N1 SPS, so your investments there can be saved and re-used).

This way you achieve the following:

  • Control of the definition process, because provisioning can be seen as a tool to map definitions onto hardware. And the blueprints do exist in machine-readable and executable format, because they are part of the provisioning system.
  • Flexibility, because you can easily replace the underlying hardware, because upper level configs and provisioning tasks do not change (putting an EAR into an appserver is an app-server task, and has no hardware or OS specifics to it)
  • And, last, but not least, you do set up yourself for what’s coming with stuff like the Sun Grid, because you surround your processes by tools. And the target of the provisioning task or tasks can also be changed easily. So, if it is a landscape of servers in your datacenter or a landscape of servers on the SunGrid is irrelevant to the provisioning itself.

I hope I gave some insights into the thinking, we at Sun (at least myself) have w.r.t. what provisioning is meant to achieve. I do hope, these small snippets here do help you in putting all the market hype around provisioning into a context, that can be used in your environment. And if you start considering re-thinking former decissions, and aligning your IT processes to the ever changing world, feel free to get into contact with me.

Provisioning should not force yourself to adapt to specific vendor’s ideas, It needs to be able to adapt to your needs, and still be capable of adopting to market driven demands (therefore enabling yourself to benefit from the things that are coming).

And to answer one of the comments/questions to my last entry: Sandbox is not as easy to define as “only in development” or “only a small project, but therefore the complete lifecycle”.

The reason for this not-so-easy answer is simple: Still every customer, every prospect, every project is different. The most significant indicator of success is: The less interfaces, the easier to use it as a sandbox. So, if you have a project, that has a well-defined scope, and that does not influence other departments, then such a project can be used as a sandbox. If the people from the customer side involved in that project do like the project, and subscribe to the benefits, this will be an injection point in the customer, because they will do the “word-of-mouth” type of internal advertisements. This is even more the case, if that project has big external visibility (for example a life-cycle management of a web-server infrastructure, where the provisioning can start by being used as a CMS (content management system)).

But, most important is: You need the buy-in of the people, that are involved, because finally, they have to maintain the system afterwards. And as the introduction of provisioning aims at simplifing the processes, it is imparative, that the people who shall use these systems need to “love” them.

Makes sense?

Provisioning: Why customers have difficulties to really go that route…

By , June 2, 2006 07:08

I’m highly involved in provisioning tasks (we call all that “N1”) in the german market, and what I have been learning during the last 3-4 years from talking to customers and doing PoCs in different accounts can be summed up quickly as follows:

  • Larger accounts tend to not decide, because a decision towards one single product puts them into the hands of a single vendor.
  • Larger accounts also tend to not decide, because their organisations are larger, and deciding to do provisioning also might need to be prepared by internal organisational “adoptions” or “adaptations”.
  • Larger accounts also tend to have a “dual-vendor” strategy, which prohibits a decision for a single vendor in the arena of datacenter provisioning. As long, as there are no real interchangable products (and I think, we all will wait long on these!), a change from one provisioning tool to another will always be a large re-implementation. The sad thing here is, that the companies still live the old, sneaker-net way, and will not benefit from the advantages of provisioning. Remember: Even doing it with one vendor will benefit in overall cost savings, and even a later change would not be as expensive as perceived, because the structural and organizational changes will be re-usable!

But: This only applies to companies, that want to start to provision the complete datacenter. With many large accounts we are doing provisioning, but this is for smaller pieces of their environment. And that works very well, if it is contained in the small “sandboxes”.

  • Smaller companies on the other hand are normally more willing to go this route, as they can adopt quicker to changes, and so we see larger adoption here. And: These smaller companies are willing to go and also are going the complete way. Remember: although these companies are smaller, the amount of work needed to be done is not smaller, if we look at the configuration and setup of the tools.

What’s your feeling w.r.t. these topics?

Are you willing to start provisioning today?

Feedback welcome!

Panorama Theme by Themocracy