http://arstechnica.com/gaming/news/2010/08/microsoft-beats-intel-amd-to-market-with-cpugpu-combo-chip.ars

“At Hot Chips today, Microsoft’s Xbox team unveiled details of the system-on-a-chip (SoC) that powers the newer, slimmer Xbox 360 250GB model. Produced on the IBM/GlobalFoundries 45nm process, it’s fair to say that the new SoC is the first mass-market, desktop-class processor to combine a CPU, GPU, memory, and I/O logic onto a single piece of silicon. “

What does this mean?

Apple is toast…Google is getting crisped! Microsoft at this point has redefined the computing landscape for all entities, hardware and software included, for the next decade. By defining and producing a piece of CPU + GPU, they have delivered on a new vision for computing for the 21st century. What’s more, if they rearchitect/optimize Windows for this platform, it could signal the end of “Unix.” This also means that Apple’s approach of OpenCL and nVidia’s approach to GPGPU are dead. Further this kills ARM technology since nothing in the ARM platform can deliver GPU level performance — and this design is already power efficient as well. Virtualization / cloud computing is dead as well!

The rapidly increasing core count in hardware is pushing virtualization technology to its limits. The latest VMWare vSphere hypervisor, in its enterprise edition, can only do an eight-way virtual SMP. That is, it cannot even support a single Intel processor which has 16 hyperthreaded cores in totality!

And that’s the tale of the market leader

So which company will want to entrust their enterprise computing platform on virtualization that cannot even support two Intel Xeon CPUs (32 cores with hyperthreading)?

Also, the era of General Purpose Graphics Processor Unit (GPGPU) computing is upon us. This marks a watershed event in the computing world. Even though this started a few years ago (2007-8), the phenomenal increase in GPU core count has made it a reality that cannot be ignored in 2010, especially with the shipping of Snow Leopard by Apple.

Open CL merely plays a primitive load balancing act between the CPU and GPU.

2010 will be a landmark year in the history of computing as most software technologies have been shattered by a tsunami of ‘cores’ — virtualization, JVM/CLR (i.e. Java/C#) — even C/C++ & Unix are barely hanging in there. None of them are able to keep up with this huge wave of cores and are getting choked by a vacuum of conceptual/logical architecture.

The onslaught of GPUs will slowly but steadily erode x86 dominance.

Along with the fall of virtualization, one can clearly visualize the big ‘crash of the clouds’ i.e. large scale failure of ‘Cloud Computing’ — which is still being peddled as the magic platform — resting on the ’solidly shaky’ foundation of virtualization.

What will rise out of this destruction of procedural and object paradigms is functional programming. Procedural programming becomes too expensive due to the overhead of threads. Objects suffer and the additional burden of memory falls on large scale. Functional provides a clear break between the in models of computational logic (what) and execution logic (how). Procedural is all ‘how’– object-oriented is all ‘what’.

Once Hypervisors and/or OS’s (be it Unix or Windows or OS X or something else) are rebuilt on this partitioning principle, only then can software be back on the road to regaining the productivity mantle.

After all the ’song and dance,’ none of the public cloud computing offerings seem to be able to handle deploying high-availability (i.e. clustered) database environments or high-availability (i.e. clustered) app/web servers.

These offerings do not provide the underlying hardware infrastructure or show real ’hardware utilization metrics.’ Only web apps that are horizontally scalable by design are easily deployable on public clouds. This is to say that only small to mid-size apps are suitable with the current generation of public cloud offerings.
Therefore private clouds and hosting solutions are always going to have a huge edge when it comes to providing high-availability offerings with ’transparent hardware metrics.’

Amazon:
http://aws.amazon.com/ec2/instance-types/
http://aws.amazon.com/ec2/pricing/

Google:
http://code.google.com/appengine/kb/business.html#cost

Microsoft: (No production offerings!):
http://www.microsoft.com/windowsazure/offers/

Microsoft vs. Google:
http://www.businessinsider.com/microsoft-azure-vs-google-app-engine-vs-amazon-web-services-2009-7

IBM vs. Google:http://blog.taragana.com/index.php/archive/ibm-tries-to-blow-away-googles-cloud-computing-campaign-with-low-cost-e-mail-for-businesses/

The next major Java SE release, Java 7, could potentially usher in a maintenance wave starting Sep 2010. Java7 has lot of good features . . .

http://openjdk.java.net/projects/jdk7/features/So much so that it almost feels like a new language . . . e.g. Multicatch, Jigsaw . . .

http://openjdk.java.net/projects/jdk7/features/

http://code.joejag.com/2009/new-language-features-in-java-7/

This could trigger a massive rewrite of the entire Java codebase.

Take this with C# 4.0 new features . . . and F# the new successor . . . the entire .NET platform could get a rewrite as well.

Business Process Intelligence (BPI) is an area that is quickly gaining interest and importance in all industries alike.

But what is Business Process Intelligence?

BPI refers to the application of various measurement and analysis techniques in the area of business process management. BPI increases the quality of process execution through features such as analysis, prediction, monitoring, control, and optimization.

http://is.tm.tue.nl/bpi09/

Some early research papers:

http://www.hpl.hp.com/techreports/2002/HPL-2002-119.pdf

http://www.prism.uvsq.fr/~grig/These/ci.pdf

A practical explanation:

http://www.ebizq.net/blogs/nari/2008/09/business_process_intelligence.php

What is Business Process Analytics?

http://www.wfmc.org/business-process-analytics-format.html

What is XPDL?

http://www.wfmc.org/xpdl.html

Who’s using it?

http://www.wfmc.org/xpdl-implementations.html

Free BPM tools

http://bpmfundamentals.wordpress.com/2009/07/10/free-bpm-modeling-tools/

In a major development, James Gosling, the father of the Java programming language, has left Oracle Corporation.

http://www.infoq.com/news/2010/04/gosling_leaves_oracle

There are quite a few lessons to be learned here:

• You can invent Java and still be out of a job!
• No language lasts forever.
• Nothing will replace “select *” aka SQL. It has nothing to do with computer science. The foundations of mathematics are built on set theory — unless some other language can be invented on top of set theory that betters SQL.
• Time to go functional/dynamic: The object-oriented era is coming to an end.

Per the grapevine, Gosling is probably looking at the following choices:

1) Join Google.
2) Join IBM.
3) Move to academics (like Bjarne Stroustrup did).

It would indeed be something if he joined Microsoft instead! Don’t think that’s likely though…

It seems like Solid State Disks (SSDs) on servers have finally arrived. Since most environments use storage area networks (SANs), it’s a huge performance boost to use SSDs unless there is a need for storage in excess of 250GB on the local drive for servers.

What is a Solid State Disk?

http://en.wikipedia.org/wiki/Solid-state_drive

HDD vs. SSD

http://www.overclockers.com/hdd-vs-ssd/

Going by the above article, SSDs promise 10 times faster reads/writes.

Reads 2x SSD Raid0 = .072ms * 1000 = 72ms 2x HDD Raid0 = 9.590ms * 1000 = 9590ms.
Writes 2x SSD Raid0 = .104ms * 1000 = 104ms 2x HDD Raid0 = 1.167ms * 1000 = 1167ms

Availability:

http://www.storagesearch.com/ssd-buyers-guide.html

1TB SSD:

http://www.pcworld.com/article/194327/want_a_cheap_ssd_oczs_4000_ssd_may_not_be_for_you.html

Budget options:

http://www.google.com/products?hl=en&q=SSD&um=1&ie=UTF-8&ei=94XIS5mDLoL78AaVmLCHBw&sa=X&oi=product_result_group&ct=title&resnum=3&ved=0CDYQrQQwAg

Java, it seems, has evolved from a groundbreaking, revolutionary language platform to something closer to a modern-day version of COBOL. In just 15 years, it has moved beyond maturity into a silver-haired stage of staid dependability. Java offers stability, not agility; reliability, not innovation.

http://infoworld.com/d/developer-world/geriatric-java-struggles-stay-relevant-700
Oracle could end up losing the “Java war.” Oracle cannot push Java at the expense of alienating all other devs from Pro*C/C#/Python/PHP/COBOL etc. Yes, Java runs inside Oracle, but it’s still 2nd tier compared to PL/SQL because of its syntax i.e. classic procedural vs. set-oriented.

So where does that leave Java? It’s a great language for enterprise development, but not innovative anymore. Even James Gosling, best known as the father of the Java programming language, encourages development of other languages on the Java Virtual Machine (JVM).

http://polyglotprogramming.com/papers/PolyglotPolyParadigm_v5.pdf

There are powerful alternatives including Scala, Clojure, and JRuby. What’s more, these new languages bring “functional programming” capability to the table, which Java cannot do.

Here’s a funny aside regarding XML. Gosling has remarked, “I’d really like to never see XML again. We are 99% of the way there.”

http://www.theserverside.com/news/thread.tss?thread_id=59733

The F# programming language will become a major .NET language with its formal debut in Visual Studio 2010, as it transitions from a Microsoft research project. F# has become a surprise hit among .NET developers. Those that can make the leap will solidify their skills and employment opportunities for the next decade and beyond.

F# also “hugely fits” the Midori programming model which Microsoft is working on

http://www.sdtimes.com/link/34251

Reading from the above, there’s no question that Microsoft is seeing Silverlight as the lightweight platform for delivering applications (web-based and mobile). Also, Microsoft might launch Midori as a new operating system for cloud data centers to up the ante against Google.

An important topic related to both Ruby & Python is the Global Interpreter Lock (GIL). Because of GIL, pure Ruby/Python languages can use only one CPU (1 core) regardless of the available number of CPUS/Cores.

But GIL is a non-issue in at least three ways:

1> Common automatically eliminates GIL Language Runtime /Java Virtual Machine (CLR/JVM)
IronRuby/IronPython/JRuby/Jython/MacRuby have “no GIL” issues, because they map corresponding “green threads” to CLR threads or JVM threads or native threads.

2> Asynchronous Programming Approach

Even if using pure Ruby/Python, using a library like Twisted (for Python) or EventMachine (for Ruby) eliminates GIL/concurrency issues.

http://www.artima.com/weblogs/viewpost.jsp?thread=230001

http://www.igvita.com/2008/05/27/ruby-eventmachine-the-speed-demon/

3> Using Separate Libraries

http://pypi.python.org/pypi/processing

http://www.parallelpython.com/

http://github.com/grosser/parallel