Tuesday Jul 07, 2009

Congestion, Creativity, Capital and Competition

Despite the global economic downturn some businesses are aggressively spending for the opportunities of the future.  However the spending of today has conflicting objectives that some would argue are necessary.  Let's take wireless mobile providers throughout the world.  In this market, competition is fierce and beneficial to consumers.  The services offered to subscribers are plentiful and rich but they do come at a high cost for the providers.  Subsidizing the phones from the likes of Nokia, Apple, Samsung and Blackberry is one big cost to get the customer's subscription.

It's great that technology has enabled GSM phones to work almost everywhere in the world except Japan "where you'll need a special phone that either supports CDMA or uses the 3G standard UMTS in the 2100 MHz frequency band. Sony Ericsson, Nokia, and a few other phone manufacturers now offer multi-band GSM phones that also include support for UMTS 2100. Coverage also extends to some cruise ships." There is a crowded group of companies looking for the opportunity to connect to individuals to provide any and all content. It's as if companies have discovered another Gold Rush.

I'm excited to see wireless and cable providers compete and innovate for delivering services to all of us around the world.  Watching cable providers offering land line service over IP and phone companies offering internet connectivity is a good example of the competitiveness out there.  The days of just delivery of service or being only the data pipe are long gone.  Providers want to delivery the data but more importantly they want to create the applications that produce the data.  The telco, cable provider and handset manufacturer all want to own as much of the subscriber stack as possible.  Now that's competition!  Here in the U.S. Comcast and Verizon are aggressively competing to win one subscriber at a time for internet, phone and HD television service.  As a result both companies are making massive investments in capital expenditures.

In fact, despite the global recession, capital spending continues throughout the world by some companies as a competitive advantage for the rebounding economy in the future.  Having spent the past few weeks talking to customers in New Zealand, Australia, Germany and reading newspapers such as the Financial Times, I've collected a group of random data.  This data can be basically summarized into Charles Darwin's theory of natural selection and applies to business as well as nature.

  • world airlines have 866 Boeing 787 Dreamliners on order.  Each 787 averages ~$200M U.S. each!
  • new cargo ships ordered or under construction is ~50% \*more\* than anticipated need
  • telcos are making huge capital investments but they understand they cannot be sustained
  • will energy become so expensive that transporting it becomes prohibitive?
  • will multiple countries practice protectionism such that localization becomes attractive again?

I'm excited that new technology will be able to help address the above as well as new economic problems we will all face in the future.


Blog is available also at: http://bobporras.wordpress.com/

Wednesday Dec 19, 2007

FOSS = Low Exit Barrier as well as Low Entry Hurdle

FOSS is a check box item for new startup companies as well as enterprise corporations who are consolidating, upgrading or issuing new application deployments.  The high tech industry will continue to have companies acquire other's technology as part of alignment and pure business economics.  Some companies acquire open source software and their intent is to continue to FOSter the community with this software, while being able to monetize the asset.   Counter to this strategy some proprietary companies may be inclined to purchase an open source software stack simply to eliminate its growing popularity by customers.  The software industry should embrace, as have universities, that more and more new deployments require solutions based on open source software code bases.  The following table shows very large deployments of storage assets based on proprietary and open source models. Open source software does create a low exit barrier for unhappy customers, but it does enable a low hurdle for a company that wants to take advantage of the opportunity to engage.  If you have built your business model around open source software you have probably listened to your customers and have realized strategically where the software industry is headed.  On the other side of the coin if your business model is to stay proprietary you may be inclined to believe that open source software is a trend and you will be able to continue to differentiate in a commodity market.  The debate continues but customers vote with their purchases.  It is my opinion that os virtualization solutions both proprietay and open sourced will shed some light on the momentum or trend of open source software.  A robust, stable, enterprise OS that can virtualize other OSes as guests has an opportunity.  The market will embrace multiple choices for OS virtualization rather than have a single choice.  With the amount of vendors who have announced OS virtualization solutions that are both proprietary and open sourced the end results are still open for debate.  Who has the momentum?  I remember the VHS and Betamax debate and who tried to dictate rather than listen to customers.

Monday Nov 26, 2007

The Art of Coexistant File Sharing in OpenSolaris

Today any OS needs to provide support for the 2 dominant file sharing protocols in the industry (NFS and CIFS).  While there are good  implementations such as SAMBA which run  "On" the OS in user space, ideally you want both file sharing protocols to run portions of the implementation "In" the kernel.  Now that OpenSolaris has a kernel based CIFS server along side NFS, I want to give kudos to a few people who made it a reality. Keep in mind implementing any file protocol in any kernel is difficult.  When you introduce both Windows and OpenSolaris together at the kernel level one can appreciate the complexities introduced to coexist.  Coexist has the caveat that code gets added in a seamless fashion and that services are not negatively impacted by this new foreign object.  Fundamental items such as the file system, security, permissions, marshaling,  etc., play an important part in making the seamless coexistence a technical mountain to climb.  That said the OpenSolaris team solved some pretty significant technical issues to make the CIFS Server in OpenSolaris available to anyone.  Despite some self proclaimed experts insisting the feat would never be accomplished, I'll point you to a few folks who know how to collaborate, innovate and in some instances dictate how to "get it done."  You can see for yourself the numerous ARC (~35) cases sponsored and approved as well as some intricate details of how problems were solved.  You can hear from the developers directly and since OpenSolaris is open sourced-- peruse ~370 thousands of lines of code that are now part of OpenSolaris.  Let's start with Mike Shapiro and Alan M. Wright.  One is a Patriots and Red Sox fan (like me) and the other is fond of rugby (unlike me... I like the Patriots).  Both are top notch engineers who can architect with ease and crank out massive amounts of code... that just works.  Mike and Alan cracked a heck of a technical problem (yes these are the hard problems that motivate them both) with respect to "Unified POSIX and Window Credentials for Solaris."  Mike describes the problem in detail on a recent post at his blog.  Alan takes us through, via his blog, the evolution of how a fully integrated CIFS service was integrated into OpenSolaris.  He is very literate on topics such as SMB autohome shares and why they evolved.  Let's not forget about Afshin, Doug and Nico.  All quite literate as well on the details of the hows and whys of a fully integrated CIFS service into OpenSolaris that was not just munged to sort of work.  From an architectural perspective numerous design requirements were diligently reviewed.  Take for example endianness.  The CIFS protocol is sensitive to the x86 processor endian order of significance.  However careful care was taken to enable the CIFS protocol in OpenSolaris to work on both big-endian and little-endian architectures by putting intelligence into marshaling.  So that means that the OpenSolaris CIFS Server will run on a SPARC-based platform if a community member were to build an appliance or simply run the service as part of a general purpose server.  Yes Niagara-II and Rock microprocessors work as it is part of OpenSolaris architecture conformance.  What other open sourced OS supports a kernel based CIFS service that runs on both big and little endian machines?

Friday Oct 26, 2007

FISH -n- CIFS

There is motivation when skeptics like to preach that there is no way <substitute here> is going to happen.  Examples include come from behind victories in sports, new discoveries in science and cures for disease.  Good intentions are one thing, but following through on those intentions is a bit more difficult. 

The storage community at opensolaris.org has been busy this past year with community members that include partners, industry standard bodies and numerous code contributions.  The community has been on a journey to continue to enhance, integrate, develop, share and invent technologies applicable to traditional storage and more importantly hybrid storage solutions of today and tomorrow.  The Storage Platform for OpenSolaris Distributions has just had some recent contributions.  Keep in mind these contributions to the community are free, open and not proprietary.  Code is not read only, but available for modification, improvement and bi-directional sharing according to the opensolaris licensing terms.

Recent additions of code and new opensolaris projects include:

  • Kernel based CIFS server (Build 77).  That's right-- designed in as a first class citizen of the os with a kernel based protocol, tightly coupled with NFSv4, VFS, ZFS and Active Directory.  Windows Interoperability.  Another complimentary open sourced solution along side our friends from the SAMBA community.  This service leverages the os and its capabilities.  Need infinite snapshots of your CIFS files?  Want file compression?  Strive for encryption of your data?  Not a problem.  ZFS provides these data services IN the file system.  The kernel based CIFS server source code will post here today.  Need help with source code management tools click here.

  • NDMP service.  Table stakes for backup applications.  The code will be binaries only until the SNIA working group members complete their efforts.

  • Virus Scan service.  Another service that is tightly integrated with the kernel CIFS server and the ZFS file system.  This service provides ICAP protocol support for off-board virus scan engines.

  • NFSv4 Mirror Mounts.  NFSv4 Clients can now automatically mount shared file systems on a NFSv4 Server.

The above new Opensolaris project pages will be posting in the next few weeks so stay tuned.

Adding more and more content to the storage stack of opensolaris may raise some questions of what can be done with this stack.  Well first off it enables better integrated hybrid storage solutions.  The x4500 "Thumper"with Solaris has enabled new thinking storage solutions by leveraging the hardware and software unique capabilities.  One can continue on this journey and enhance the unique hardware and software capabilities.  In fact think of it as Fully Integrated Software and Hardware on a repetitive basis.  The tighter you integrate the incremental features, the more compelling solutions with commodity components using an enterprise open sourced os one can build.  I would think that if you could provide a software heath kit of an open sourced software stack to appeal to the masses you may have something worth investigating.

Come and participate.  Opensolaris is transforming itself from open storage solutions, xVM to Solaris install revisited also known as "Indiana." 

How can you pass up FISH -n- CIFS well prepared? Help spread the word.

Wednesday Sep 26, 2007

If you are painted as irrelevant are you a threat?

What is a better position to be painted as?  Company A: that alternative mouse trap or Company B: that former high flyer but "irrelevant" entity.  It doesn't matter what industry: automobiles, home appliances, beer industry, retail, etc. We have all heard the competition paint their competitors.  Sports teams like to use the "no respect" card to motivate the team.  Satisfaction is gained when the "team" shows the pundits and the odds makers differently.  In fact some like to be the underdog so they can fly under the radar and be taken for granted.  It is a great position for professional sports as any team can be beat on a given night when you are playing the best.  Roger Clemems was supposedly washed up 11 years ago when he was traded by the Boston Red Sox to the Toronto Blue Jays.  How many Cy Young awards did he win after the trade?  Four of them. How many Cy Young awards has he won in total?  Seven of them.  Being painted as all washed up and irrelevant was a motivator.  "He is too old", "Doesn't have the edge anymore", "Lost his spark", "Has too much money", etc. He saw it differently than his pundits and competition.

If you are a company and get painted the same way:  former high flyer, lost its vision, could not adapt after the .BOMB, financially unsound, etc. Those are certainly motivators.  If a company was in the software and hardware business what would they have to do show that irrelevance is determined by the customer not the competition at the end of the day. 

The company would build hardware that is optimized for virtualization on multiple fronts.  Virtulization is not a one recipe fits all problem.  Have your hardware be hyper visor aware.  Make sure your hardware can run the dominate industry operating systems even if one of them happens to be proprietary. Memory density and I/O connectivity should be features that differentiate.    Common industry standard components and subsystems that span across multiple microprocessor architectures is attractive.  If different microprocessor-based blades can share a common backplane all the better.  The company should then make sure that their software is virtual machine aware to be a great dom0 as well as run as a VM in someone else's dom0 even if it is proprietary.  This software would also allow different flavors of VM capability.  For example virtualize and entire OS 'on' your environment or virtualize particular applications 'in' your environment with zoning.  Lastly make sure your software environment runs on competitor's hardware. So far the company would have multiple hardware capability teamed with multiple software capability.  This sounds like motivation. 

Next the company would be able to put the two above items together and build compelling solutions for customers.  Imagine the products that one could build. For example build a server that can eat and serve storage easily because the architecture is closely coupled and the os is optimized.  Or a server that can horizontally consolidate multiple instances of linux by taking advantage of high cpu thread counts and the VM aware os.  How about taking it one step further.  The above 2 examples are based off of general computing.  So now let's extend the solution with embedded capability.  Create dedicated products with the same hardware and software from above-- but leverage their ability to virtulize...  That is both compelling and scary especially if the software licensing terms allow anyone to have access to all the code. 

Cy Young like stuff...

About

The blog of Bob Porras - Vice President, Data, Availability, Scalability & HPC for Sun Microsystems, Inc.

Search

Categories
Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today