NetKernel News Volume 2 Issue 13

January 21st 2011

What's new this week?

• Repository Updates
• What is NetKernel, and when would I want to use it?
• NetKernel West 2011 - Programme Content
• Pimp My Laptop

Catch up on last week's news here

Repository Updates

There are no updates this week. Steady as you are.

What is NetKernel, and when would I want to use it?

That Brian Sletten bloke has been at it again. This time he's written a very articulate and independently authoritative answer to a question on Quora.com...

Q: What is NetKernel, and when would I want to use it?

A: Answer here

With this, and last week's REST refcard, it begs another question...

Who's Brian Sletten, and why does he keep plugging NetKernel?

... so I thought I'd better step in to rescue his reputation for him.

Firstly Brian is wholly independent. Other than mutual respect and admiration, there is no formal relationship between Brian and 1060.

Brian splits his time between his day jobs as independent consultant architect/author and his passionate evangelism of advanced software and information engineering - mostly via the No Fluff Just Stuff conference series.

Currently, and amongst other things, he's the consultant architect for the BestBuy.com resource oriented BBY Open Commerce API. No small gig, BestBuy.com is one of the largest ecommerce sites in the world.

Brian is a member of the Beverly Hills glitterati and regularly plays racket ball round at Brad Pitt's. (Well, he does spend about 3 days a month at home in Beverley Hills, that much is true).

Honestly, you'd have to ask Brian why he's so passionate about evangelising NK and ROC. For us it is very much appreciated. You can ask him yourself, he has signed up for and will be a speaker at the NK conference in April...

NetKernel West 2011 - Programme Content

The details of the conference programme content are now published. There are also further details on registration options and sections on "Follow-up Friday" and "Social Saturday"...

http://www.1060research.com/conference/NKWest2011/#Details

As in previous years the aim of the conference is purely an ego trip for yours truely. Crap I've gotta get a grip on this Honesty Tourrets. Actually the programme intent is firstly to cover the technologies that have been surfaced over the past year and a bit.

So, for example, there'll be time spent showing how NetKernel Protocol works, and what new distributed ROC architectures are engendered. It was slipped out with little fanfare, but there's a heck of lot of potential to show here.

We'll also be releasing the all new compositional development environment (CDE), showing how it works and how you can design your compositional tooling to exploit it - if you've wondered what Tony's been doing recently, all will be revealed!

In the last quarter of 2010 Tony and I also got around to brainstorming and documenting all of the key ROC design-patterns that we've used/come-across during the last 10-years. Over a couple of sessions, we'll be sharing these in bite-sized chunks with examples and takeaway references.

You'll be pleased to know we've learned our lesson from previous years and whilst there's a hose-pipe of stuff from the horses-mouth - we'll ensure plenty of breaks (we'll even provide restrooms this time - no more queueing for the plant-pot in the corner).

There are several talks from 3rd parties with an eye on practicalities and you can be sure to get an impartial picture of their journey into the world of ROC.

There's also opportunity on the Wednesday for lightning chalk-talks for the experienced NK community to share what they're up to - these have been really very interesting in previous years, you guys are doing some exciting stuff.

Don't forget you've got just over a week for early-bird rates which closes midnight GMT on Monday 31st January. Our prices are a little more than in previous years, reflecting that using a Hotel venue for the first time comes with increased venue and ancillary costs. But, with the great accommodation rates we negotiated, we hope its within reasonable budget. We certainly plan to deliver the value for money you've grown to expect from previous years.

See you in Colorado!

(Don't forget, if you're new to NK, a one-day bootcamp precedes the conference and provides a fast-track into NK and ROC).

Pimp My Laptop

This week I had a day of transition from my old laptop (3.5 years old, heavy duty use, display has some weird green lines and the backlight is dim on one side, time to change before it becomes critical). Laptops are commodities and pretty dull but even this apparently inane subject offers some insight on ROC. Let me elucidate...

A Snippet of Biography

In the mid-90's before I decided "to figure out this Web thing", I had a couple of years working in HP's laptop division. I'd got myself a really cool job - I had freedom to keep one foot in HP Labs research and one foot in the division (based in Cupertino Si Valley). My role was future technology strategist.

My direct boss was Bill Wickes, any ubber-geeks out there will know him, he's the father of HP's scientific calculators, he's also an ex-Princeton physics professor. (For a junior nerd like me at the time, what cooler job could I have? Oh, it gets even better, his dad built the Saturn V rockets! If you're coming to the conference you should corner Kathleen Dollard, it turns out her Dad was an engineer on the Saturn's too! ROCket science, in every respect).

This was in the days when HP decided to really target the PC business and, rather touchingly in retrospect, it was still engineering-led. It would "break the PC business with technical innovation".

My job was to extrapolate from the current product roadmaps, to look at suppliers roadmaps, to look at the future market growth, to look at the research projects inside HP and to foster and if necessary initiate technologies. As well as being a great technologist, Bill's a great manager, he gave me free reign:

I could look at end-user product technologies - for example I initiated research into embedded fingerprint readers and led business relations with possible suppliers and talked with Microsoft about integration in the operating system - all of which are now standard features. I was involved in the discussions between HP, Ericsson and Intel on what became Bluetooth. I was at Intel when the first clock-scaling Geyserville chips were planned (now called Speedstep or Turbo something or other).

They could be production engineering - I commissioned an HP labs project to develop a laser optical display interconnect to reduce manufacture and repair costs.

They could be indirect peripherals - I developed a detachable laptop display and stand, and, the simpler, adjustable angled laptop stand. And the "legacy-free dock" based on just USB and Firewire buses.

We even planned beyond the PC to "always connected tablets". I had this concept of an "info-scoop" (catchy name eh!) - it was clear that storage costs could be predictably forecast and you could carry everything with you everywhere. The first thing you'd take would be music, then voice calls and web, then documents, books, small apps etc etc.

I was drafted into a special "e-music" strategic corporate initiative and I hacked up an MP3 player on an HP palmtop but couldn't get traction. Trying to pitch the extrapolated idea of the progression iPod -> iPhone -> iPad was, at the time, harder than evangelising ROC!

So an interesting career step, and great experience in many ways. Plus I stacked up a world record number of air-miles. Over a beer there's more stories here.

Fast Forward: The Present

So you see, choosing a new laptop is a matter of professional pride. As I said, they're commodities (like Pork-Belly's) and I have other things on my mind these days. But dipping my toe back into the market, its good to catch up on progress.

Laptop: Select and Pimp

My laptop is used full-time for about 14 hours a day. At home its on my knee, a desk or lugged to the kitchen table. In the office during the day its on a tilted stand (probably somewhat perversely, because getting one of those on the accessories list was one of my first serious contributions to HP's bottom line).

With an IDE constantly open, and with all the spreadsheet and business stuff too, I have always aimed to get the most pixels I can possibly achieve with each generation. Last time I had 1680x1050 which was my only display - it cost a little more but was no way a premium option.

Imagine my concern then to discover that high-resolution displays have migrated into the high-end segment. Everywhere you look in the low to mid-range you see a standard format display 1366x768 - with variations on size from 13.3" up to massive things but very few, or premium priced, options for higher resolution. What the hell was this?

My unwillingness to enter the upper tier market is not about the money. Rather, I think having bleeding edge performance would be dangerous to us. You may know by now that, with NetKernel, we are obsessed with computational efficiency. So to ensure that we don't get lazy, it makes sense for us to use mid-range kit. So I like to buy just above the crappy basement segment but at the low-end of middle market. A variation on the Goldilocks principle.

Well many of you will know that I've had my head firmly planted in a place with little sunlight. You know that whilst the laptop has become a commodity, so too have monitors. I hear that LED backlit monitors now come in cereal packets in the USA. I put two and two together - oh right, small display for travelling, very large format display for desk/home.

Which, dear reader, is my usual long and rambling way of getting to the point. I bought a cheap-as-chips Toshiba L630-14J and a 1920x1080 LED backlit monitor. (So many pixels I have to grip the desk tight to cope with the agrophobia).

Being of a practical bent, I also like to buy the cheapest set of ancillary bits in the product range - base memory of 2GB, base disk of 250Gb, base CPU Intel i3 370M. I then proceed to gut the machine, throw away the memory and swap the drive, get rid of Windows 7 Home Premium and put on Ubuntu 10.04 LTS.

To whit, I am the custodian of a 13.3" Tosh notebook with custom 8Gb RAM and a 60Gb solid-state disk. (I stacked it with RAM, not because NK needs it, rather it means I can now play with a development cloud of clustered NKs on virtualized Linuxes where each one, including Java/NK, comes in at 300MB). The integrated Intel chipset is a Core i3 370M with dual core and dual hyperthreading - logically 4 CPUs to the operating system.

Including the monitor this lot came in $400 below a large-format premium model, and well under $1000. All of the bits, including the solid-state disk are commodity items. This is what you get at the low-end of mid-range today.

As ever my first thought was to find out how NetKernel goes?

NKMark10

You'll recall that one of the tools we introduced to NK4 last year was the NKMark10 benchmarking tool. Here's the news items where it was released including some representative scores for some platforms we tried, together with the, at the time record score, on an 8-Core 2.5GHz Xeon server.

Inside NK, the tool is available from the control panel tab here...

http://localhost:1060/tools/nkmark/

You'll remember that one of reasons for these tools and measurements is to show off NK's linear scaling - which is reported in detail on the main website here.

PJR Laptop ShootOut

So as meaningless comparisons go, this is meaningless. (Well OK, it does give you a relative feel for the progression of the low-mid-range in 3.5 years).

First my old laptop (2GHz dual core, AMD Turon) NKMark10 score (click the images for high-definition image)...

Old laptop NKMark10

And the new one...

New laptop NKMark10

I've got an NK box with a score of 23.0! By comparison, last year when we released NKMark10 we gave the score for our staging 8-Core 2GHz AMD Opteron test box it achieved 16.9

So a disposable commodity laptop now ranks as a high-end server.

The Sermon

Is there any message here? Well, to me it feels like our classical approach to software does a disservice to this machinery.

Bill Wickes could shoe-horn a complete symbolic math engine into kilobytes of RAM on an 8-bit chip in the 1980s.

Today, we we see OO-frameworks on top of frameworks on top of frameworks. We feel like its progress. And it is. Linear progress. Meanwhile the platform is exponentially better.

We are witnessing software's "middleware middle-age spread".

We have the capacity to be doing fabulous, unbelievable, "Saturn-V scale" things with our information.

As you become familiar with ROC you realise that resource-oriented development is about working from the edges (top and bottom) inwards. More often than not, you find that there is no middle.

As Brian keeps doing, tell the world, ROC is "every-ware" and "no-ware".

Lets go to the moon.

Have a great weekend.