As a continuation on my blog concerning good enough systems and IT as a commodity, many people are buying IT, be it software, hardware or services from whoever can supply the cheapest price. Boy is this scary.

I understand that everyone wants a bargain but as my mother used to say, pay less costs more. I will never admit that she was right.

I just bought some new fuses for my house, technically Minature Circuit Breakers (MCB), Residual Circuit Breakers (RCD) or RCCD's for the fuse box. I bought from reputable European suppliers as my families life can depend on the ability of one of these to detect a short circuit (residual current to be accurate) and quickly turn off the current. Cheaper ones may not "trip" or switch off fast enough or the mechanism may stick after a while. You can look up all the specs, B, C class etc if you do not believe me.

In the same way I will not buy the cheapest food for my family as this could easily cause short term sickness and a deterioration in long term health. I also will not buy the cheapest wine, even though you get blue teeth free when you drink Lambrusco. I can save a fortune eating the cheapest frozen pizzas, but will this have any positive benefits. Or will a population that ate the cheapest low cost food be a burden on society and future generations.

Will a company in the same way that has fed itself on the cheapest IT, suffer from malaise, unresponsiveness and general bad management because of lowest price infrastructure. Will it take a couple years to show the deterioration in services and extra costs of buying decisions where the cheapest low cost IT was the main criterior.

So why do companies buy the cheapest computers. Why don't we employ the cheapest staff, get the cheapest coffee, have the cheapest food in the company restaurant. We used to call this a false economy.

Computers are now an intrinsic part of societies infrastructure, if we want society and commerce to run smoothly we have to invest wisely.

Who wrote this " A man who knows the price of everything and the value of nothing".

I was talking to an old colleague about what he sees going on in the industry and what some customers are asking. His answer was that a lot of people just want good enough systems, they do not request the best suited systems or what gives them any competitive advantage. Pretty sad as the IT industry always strived and tried to be innovative and make the best reliable and scaleable systems. Have we forgotten about this, is mediocracy the mantra of the day. Good enough systems are not only sad, but they cost companies and society time and money, you can always get more money, but once time is lost you cannot get it back.

Today I read that a countries online tax system crashed as it could not copy with the load at the end of year deadline for tax returns. So everyone has to spend their time doing their tax return again, to what cost to society and the economy. Was this calculated when good enough systems were specified. I hope this was not a good enough system, but some other problem. We do not need expensive technologies to fix this. The multicore CMT systems like the T5220 Niagara multi-core processors as a web tier and scaleable M-series enterprise servers as backend database engines can be used to avoid these problems. They can replace and be used to consolidate masses of good enough systems, that are not really good enough as we just saw. T5220 and M-series are good enough for every day use and can scale quickly to cope with peaks. We have the tools to avoid these embarassing IT problems lets use better systems.

If we put good enough brakes on our cars will they stop us in time when the car is full. To me that is not good enough. I want safety the system that meets the everyday and highest emergency requirements placed on them.

Good enough is not sufficent. We have the ability to do better, otherwise we have a serious malaise in society. How many of us will send our kids to good enough schools, if we had the choice and means to do otherwise.

If you only have the time to find good enough systems, I will save you time to find better ones. Here is a link to some better systems:

Consolidate several good enough small systems into this and save money and power. http://www.sun.com/servers/coolthreads/t5120

Consolidate many medium good enough systems to this and get more performance and reliability. http://www.sun.com/servers/highend/m8000

Who wrote this; "Mediocrity knows nothing higher than itself, but talent instantly recognises genius."

Well the reason I started blogging was to stop having to answer the same questions over and over again. I set myself a target of blogging or writing a blog every week. But I must be honest and admit that I failed. Towards the end of the year I started to miss whole weeks. However, I did get a large number of blogs 400-500 hits per day on some of the most popular ones. Also, some very good comments and discussions.

So for this year I will continue and try to see if I can reach my target of blogging once per week. As one of the blog advisors at Sun said that the more you blog the more people will read your blogs.

So a Happy New Year to all those that read my blog. As this is my first day back at work my blog counting starts today, so it is my first week and first blog.

Stockholm is foggy, +2 deg C and all the snow has melted.

On 6th Nov I wrote an article in this blog that to do virtualisation sucessfully, i use the term sucessfully as anyone can do virtualisation badly. Administrators, storage and system admins need to understand their applications before consolidating applications and servers. Historical capacity planning analysis is the only way to be able to size a new virtualisation server correctly. Well today this article appears http://www.internetnews.com/bus-news/article.php/3714351 which explains that there may not be sufficient network or I/O ports (HBA's for those in the business) on a server to cope with the capacity required by all of the virtualised servers and applications. From a storage perspective that the article refers to, it looks like people doing virtualisation have not understood the I/O profile and access patterns of the servers and applications that they are virtualising.

So everyone two choices, virtualise and hope for the best, hopefully no-one will complain as it is fashionable, a bit like the tight trousers analogy in my 6th Nov blog. The story about the emperors new clothes springs to mind. Or get out your historic capacity planning reports to see if all the applications and servers will fit together.

The process going on here is a bit like packing the car boot before going on holiday, some people fill the whole boot up with "just in case" things they may need. Others take the minimum. Some of the people that fill their cars up to the maximum never get to their destination as their suspension breaks when they hit the first bump in the road, they end up with a broken car at the side of the road. Yup they even have the roof rack filled to the brim. Overutilisation as we say. It is a comfort thing taking your house with you on holiday.

I am glad this article appeared now that the virtualisation fashion may start to calm down and people start to become more sensible. After all, it is not the first time in the IT industry that consolidating servers is a new thing, nor is server virtualisation. Maybe we are in search of something new, only to discover something old. Well for new stuff read about the coolthreads servers and X4500, in 20yrs I have never seen anyone make servers and chips that use less power and do more.

I had a interview with a journalist last week about automated tiered storage, here is the transcript. Interesting questions that he asked, which makes me think how people think about tiered storage.

Q: For what is automated tiered storage good?

A: Even there are many problems that can be solved by automated tiered storage. There are three main areas where tiered storage is good, saving money or running costs, reducing the storage management workload and legal reasons. Briefly, from a power, cooling and reliability perspective it is not economical to store archive or long term data on one tier, primary storage such as disk. So by moving inactive data from disk to lower cost tier/media you can save money. To reduce the storage management workload, by implementing tiered storage you start to manage your data more efficiently. If you can have your data stored on the appropriate cost and performance storage you will not run out of space as often as if you just have all data on one tier. From a legal perspective you also may need to store data for very long periods of time. In the case of legal or medical records you just cannot keep it on spinning disk as this requires power all the time, also the lifespan of a disk is about 5yrs, CD/DVD is about 10yrs and tape about 15yrs. So you can imagine that if you only have one tier such as disk and you have a large amount of data your staff will be doing nothing productive, but just migrating data and replacing disks, CD's or DVD's. Migrating and replacing disks is not a business that a company wants to get into as it does not generate any revenue, just costs money. To summarise if you have automated tiered storage your staff can spend more time on much more positive and revenue positive projects. You still have to migrate your data from old to new disks, replace them, but it may not be such a difficult task.

Q: For whom is it a good approach?

A: There is no such thing as a normal or standard computer environment, company or business model, so it is a difficult question. But really it is good for people or organisations that want to save money or have large amounts of data or cannot control their data. So it is good for established businesses such as Banks and it is also very appropriate for the new Internet companies which have lots of data that they have to keep but becomes inactive, that is data that they may need to read one day but they cannot delete it. For many companies nowadays data is the "lifeblood" of the company, the actual data has value. For example if new data mining techniques become available in the future companies may want to keep old marketing data or customer transaction history and run it through various data warehouses. So it is good for whoever wants to extract the maximum value from their data. You never know the value of something until you do not have it.

Q: For whom is it not?

A: In reality no-one. I would say that it is not good for people who have lots of money and do not understand computer systems, possibly home, small and medium business, but if they want to save costs and reduce complexity, then it is good. So many people at home copy old photos to CD or DVD. Even though that is not automated, people are still using a tiered storage approach. Home and SMB users often buy lower cost and slower data storage devices to attach to their PC's. They then move old data such as photos to these devices.

Q: What should companies consider before starting an tiered (automated) storage project? (How to get the most out of tiered storage)

A: Companies should understand their data and how it relates to their applications. They should consider that some data increases in value over time such as medical data, legal and insurance data. Other types of data decrease with time, temporary files & SMS have lifespans of minutes and outputs of test program runs you may keep for hours or days until you get the correct result or report. Sun has a very simple questionaire that all companies can use to see how mature they are in managing their data compared to others, this is called the Information Maturity Model. This has all been documented here: http://www.sun.com/software/solaris/optimization_report.pdf

Q: What are some of the pitfalls?

A: The greatest pitfall is when you do not understand who uses the data and how. Then just move it to another tier. This is what we call the access pattern, there is not point in moving data from primary storage e.g. disk to secondary storage e.g tape every night and then every morning people want to use it. You are just wasting computing resources moving the data between tiers. It is what we call thrashing. Not understanding your data usage and access patterns is the greatest pitfall.

Q: Is automated tiered storage The Future? Or are there other technologies/approaches that might be better?

A: Automated and tiered storage has been available for at least 20yrs, it was called HSM in the early days, we had solid state disk, spinning disk and tape then. However, automated tiered storage often tracks and goes up and down with economic cycles. This includes the recent increase or peak interest in eco-computing. Where eco means economic and ecological, these factors go hand in hand and are strongly interrelated. So while we are concerned with power and cooling and general eco-computing it remains important. If a new storage medium becomes available then the dynamics of the tiers and costs can change or stop the trend to automated tiered storage. However, as people will always be creating more data and there will always be different cost and performance characteristics of storage tiers, devices and media I believe that automated tiered storage is the future. Another approach that may be better is to delete the data you do not need, how many people or organisations do this ?

Thanks for now.

A statistician drowned while walking across the river, the average depth of the river was 2ft (60cm).

This is the best way to explain that, if we use averages when planning for server virtualisation or storage consolidation we may have performance and or capacity problems.

I am a very large user of the web, to read news, to buy things, to research subjects, pay bills etc. However, last night after a 12 hour long work day and trying to balance a last minute business trip with family obligations I went web dead.

Could not buy a present online, web slowed down, could not think of what to do, shops were closed that late at night.

Next time someone asks me why you did not get the present for your wife, I will say I went Web dead, not brain dead. Always, good to blame the internet for your problems. We could not do that 10yrs ago, even though I tried with my Mosaic browser.

In order to do virtualisation properly and safely without affecting the response time of the system during peak loads. We must have historic performance data for systems we are virtualising. At least 18 months, to cater for seasonal usage patterns. If not we may not be able to meet the Service Level Agreements (SLA's) due to bad response time and in the worst case a system could crash if overloaded.

System crash is not the bad thing, it is the affect this has on the people who depend on the computer system that cannot do what they wanted to.

So for people doing virtualisation correctly, they will need to monitor, analyse trends and generally get better at capacity planning.

I expect a silent boom in IT capacity planning.

Well, imitation is the best form of flaterry. Wyse have come up with a SunRay equivalent. Lots of people said that SunRays would never be popular, now other companies are doing similar things to Sun. For those that do not know about this technology, SunRays and thinclients are desktop machines that allow everyone to share a larger and faster computer. Sun can run Windows, Solaris and Linux applications on SunRays. Thin clients use less power (no disk, CPU or fan), are more secure (they have no data in them that can be stolen) and can be replaced in less than 10 minutes if they fail.

You can see the Wyse product here: http://www.wyse.com/products/hardware/thinclients/V10L/index.asp

Picture of the SunRay here: http://www.sun.com/sunray/sunray2

"A ray or enlightenment in a dark IT quagmire", cost $249 for a Sunray, Suntan lotion not included.

Hot chips and cabinets of disk, causing London to get hot under the collar. Blimey it went terribly wrong in London last week. Could of been a heated debate. See this report.

http://searchstorage.techtarget.com/originalContent/0,289142,sid5_gci1277677,00.html

Gentlemen never perspire. So how do the boys in old blighty fix the above problem.

Maybe, SunRay thin clients, low power servers, archive old data to tape. And time for a couple of cool pints of bitter as you have less and simpler support with Sunrays.

Years ago, I remember while sitting in a traffic jam in Shepards Bush roudabout, the outside temperature reached +40C. While driving 4 hours north of Stockholm a couple of winters ago the temperature reached -27C. Nearly had to chuck my English friends out of the car and tell them to push. They are too polite to say no (actually my friends, would probably throw me out).

Moral of the story, do not put your datacenter in the Shepards Bush heatsink/roundabout.

The IT industry is full of TCO and ROI calculators. However, I have not seen any Cost Of Exit (COE) calculators, this is how much it will cost you to get away from the technology you just purchased and implemented. COE is partially what open systems, now called open source was supposed to make easier i.e. to be able to move an application or data from one platform to another with minimum costs. COE could also be called Cost Of Lock-in (COL). There are serious implications here, just think how much extra your IT supplier is charging you if he knows that you cannot move away from that particular technology. But do not jump to something else as a quick reaction, that may cost you even more. Stick with open solutions, which give you options.

It would be interesting to see how much it costs to do get out of lock-in with the latest fashionable technologies, what is the cost too:

Migrate from Red Hat Linux to Suse, or any other flavour of Linux.
Migrate from a de-duplication solution to a standard backup solution
Migrate from a Content Addressable Storage solution to another archive solution
Migrate from VMware to XEN

Sun has some very open archiving solutions such as SAMFS see here: http://www.sun.com/storagetek/management_software/data_management/sam/index.xml

Any data you store in here you can access by any server using NFS. Also, the format of your data is not in any proprietary format, so you will always be able to read the disks/tapes using other systems.

When I first started to work with open systems, which is what we called UNIX in the 90's a Director asked me why is the security and management so bad. As he had walked past a network room full of running servers and the door was wide open. They only thing I could say is "thats why they are called open systems, the door is wide open so that everyone can use them", made him laugh, bought me a bit of time to fix the problem. Mainframe guys who did not want to let us in the datacenter, had to let us put the Unix servers in the datacenter after this escapade. Not too hard as I am an ex-mainframe guy, everyones nightmare. See, my discussion of computer tribes here:

http://blogs.sun.com/ValdisFilks/entry/it_tribes_why_we_behave

With Sun Solaris, if you do not like our SPARC servers, you can always migrate to Solaris on Intel or AMD servers made by other companies. If you do not want to buy Solaris from Sun, you can get it open source from OpenSolaris.

If we are financially smart, when we get a house mortgage or car loan, we check the small print to see if there are any hidden costs if we want to terminate or payback the loan early. We need to do this with the IT technologies that we buy. Too often with new technologies that solve one problem, they create another later. Good thing in IT is that, often those people who made the decision for a specific solution which had lock-in get promoted and move on. Someone else later has to pay or tidy up the decisions made earlier. Early adopters get promoted, late supporters get overtime to fix it. Senior management in companies who make IT Directors buy fashionable technologies that have lock-in, have to take some responsibility for this. Too often I see senior management say, my kids at home use this to do this, why cannot we do this at work. Simple answer, it is a hobby for your kids but a business for you.

Go for open systems, open storage and systems that do not lock you in to any device or software dependencies. Just think of the Cost Of Exit.

This is so obvious, if you store your document using a smaller font size they will not take up so much space on your disc drive. Try it and see.

Top tip, smaller fonts, save money.

One T2000 server can replace several old servers. You can move your email, database, web applications from them and consolidate onto one T2000.

One SL500 tape library can replace whole cabinets full of disk drives or network filers storing rarely used data. One SL500 tape library uses 10x or less power than the equivalent disk array or NAS filer.

One archive filesystem can move data transparently between disk and tape, you cannot tell the difference and never have to manage it yourself. This is what Sun Storage Archive Manager is for, to make you life easy, stop the disk vs tape discussion.

Tape or disk is not a binary choice, you can have both, we can store all of your data in an intelligent filesystem which migrates your data to tape when you want it to. Therefore you never know that you are using disk or tape, it all looks the same, like a file list in a windows GUI. All you realise is that the amount of money you spend on data management reduces. We can do this with the Sun Storage Archive Manager described here http://www.sun.com/storagetek/management_software/data_management/sam/index.xml

In both of the above examples, service and maintenance costs are lower, power costs are lower, you have more floor space freed up in your office or datacenter. You have more money in your pocket to invest in new applications or your favourite hobbies.

An economist would say that you obtain year on year (YOY) savings with a better total cost of ownership (TCO) and a faster return on investment (ROI) by using these Sun technologies.

I would say you have more money, can spend more time on fun things, become richer and have an enviromentally friendly computer infrastructure.

To me this means one thing, Solaris will be the only OS that enables Windows to run on it. Hypervisors excluded. How many platforms can do this. Shows Solaris leadership, again.

Lots of confused people philosophising about this. Please read my blog about tribes and listen to the music. Europe was at war for 1000's of years, now except for the east where we still get some sabre rattling after the iron curtain came down. People, or more precisely their leaders have learned to live with each other and it is better for everyone. I hope we do not forget that this is what the European Union is about, peace in Europe and not discussions about real chocolate, Swedish chewing tobacco being banned and english sausages not being sausages.

Some things are really simple, lots of Windows in the world lots of Unix (yup Linux is Unix), now Unix will run on virtualised Windows and Windows will run on virtualised Solaris/Unix.

The English will still drive on the left and drink warm beer (bitter) and implement the metric system inch by inch. Not everything can be solved or work together. But buying petrol in liters and measuring distances in miles is a good step forward. I like driving and measuring distances in kilometers, you get to the destination faster.

Where is the problem, Solaris and Windows continues to interoperate together. Just like litres and miles.

Very good presentations by Tim from Intel and Marc from Sun. They covered a lot of ground and material, brings back memories from my student days.

My notes:

From Tim (Intel), we will solve all the worlds problems if we re-write code to be executed in parallel languages using messaging passing. People need to agree on a parallel programming language, do not invent a new one, use what is here today JAVA, OpenMP just upgrade/fix it to be more parallel.

From Marc (Sun), we can solve some of the worlds problems today, using software in solaris libraries, Java, Multi-core and multi-thread processors and transactional memory. Need a system approach to do this, just getting one bit right will not improve things.

The two processors with the best multi-core/thread implementations available today are IBM's cell chip and Sun's Niagara. The fast CPU's in the last 7yrs just run bad code/programs faster and give you a bigger electricity bill, the number of transistors double every 24 months, increases in cycle speed (Hz, Ghz) is not longer sustainable due to overheating and internal (on die) connections becoming too small.

Interesting we did not hear anything about Itanium, shame 10yrs ago when I saw the first presentations from HP it was so promising. No news about Itanium multi-core, was hoping to hear about its future.

When I got home after the conference, the first decision that I had to make was. Whether to dial-in to work or feed the cat, I decided to solve the immediate problem which would have the greatest reward. I fed the cat, then I logged in. This was so that I would not get disturbed, investing in the future by feeding the cat out first. I look at this the same way about what Sun and Intel said. Sun can solve 80% todays problems, but not everything, fix 80% now, work on the rest in parallel. Intel, surprising for a hardware company, want to solve all problems by re-writing code first and get everyone to agree on a programming language. This is a good idea, not that practical but will solve many problems in the future/long run. But it will not fix todays data center and overheating CPU/chip problems.

Too summarise:

I do have lots of respect for what Tim said, but I think in the history of computing people just do not re-write code. Look at how many banks are dependent on the mainframe, because no-one can understand the code that runs on them. Most people believe the mainframe is around for some cost or technical benefit. I see lots of legacy applications that companies do not have the resources to change or they are scared to change.

Good for Sun to solve todays problems.

Good for Intel to rally the industry together to solve tomorrows problems (which are actually here today, since Sun's multi-core & thread Niagara chip can run this code and the old/existing code real well and it came out last year).

Strange when Sun is so practical, but leading in cores/threads/transactional memory etc and that Intel is calling for better coding. Very honourable, Intel. But it will be a long road, just as it was in Sweden when they changed from driving on the left to the right. It took probably 20yrs to get rid of the left hand drive cars. This joke/quote/allegory I must attribute from Dr Eric Hagersten.

My added value, we changed from driving on the left of the road to right here and still have traffic jams.

With Intel and Sun's new partnership, it is interesting to see roles being reversed. Sun the practical guy with leading processor solutions today and Intel the honourable guy encouraging the software community to stop arguing about programming languages.

History does repeat itself, just as Tim from Intel mentioned. 5yrs ago it was software i.e. JAVA from Sun and CPU leadership from Intel. Now it is CPU leadership from Sun and Intel leading the charge to re-write software.

Nice to see Sun and Intel together solving the performance vs power problem from different perspectives, also the joint Solaris announcement a month or so ago. It is a good complement, Sun's chips and Intel's software. Who would have thought this 12 months ago.