Avik’s Ruminations

Came accross this great comment from Erik Naggum as part of larger polemic:  

Scaling and maturation are not the obvious processes they appear to be because they take so much time that the accumulated effort is easy to overlook. To be successful, they must also be very carefully guided by people who can envision the end result, but that makes it appear to many as if it merely “happens”. Take a good idea out of its infancy, let it age without guidance so it does not mature, and it generally goes bad.

I think the only evidence of good  ‘design’ in in how software ages. The rest is noise.

As Morgan Stanley Matrix is released to clients, its groundbreaking nature is being finally revealed. A new and bespoke internet trading platform  with this amount of complex interactivity is a fascinating  design and architecture problem, from the usual suspects, to some unique challenges. Leading our team to architect and deliver  the services for Matrix  has been, and continues to be, a challenging exercise in solving issues of integration, scalability, stability and asynchronicity. 

Our  own Borre  Wessel is presenting the flex development experiences at Adobe MAX 09. The server-side story is equally interesting.

Talking of visualisations, came accross this new forex visualisation from FinViz. While we’ve seen and done better, this looks pretty nifty for being pure html. It ‘s performance is quite snappy as well. 

Came accross an old series of articles by Julian Browne on Space Based Architectures. A nice exposition if you are interested in the topic. Gelernter’s original paper on Tuple spaces and Linda is a good read, if only to understand the history of the idea. 

While the technology is easy to pick up for a reasonable competent programmer, the difficulty in doing something like this is, i think, in ensuring that the solution architecture is appropriate for the problem domain. In those situations, these kinds of tuple based solutions provide order of magnitude reduction in complexity when building large and resilient systems.

Google Native Client is a technology for running native x86 code in web browsers. “What!” I hear you exclaim, “why does the world deserve another instance of the ActiveX disaster?”

Not so fast. Building on the idea of Wahbe et al from the early 90’s, its basic idea is to build a verifier for  a slightly constrained x86 binary. The folks at Matasano Chargen have a good analysis of the security issues in this architecture. 

However, given the current prevalence of managed programming environments as browser plugins (in Flex and Silverlight), one wonders if there is a large user base for writing native applications for the browser. Games are of course one obvious candidate (the NaCl demos contain a Quake port.) In any case, this is something worth watching over the next couple of years.

I’ve been reading him for over a decade, but Nassim Taleb is probably the most re-discovered personality of the financial crisis (maybe along with Nouriel Roubini).

So its not suprising to see him pop up in the news cycle quite often these days. Two recent sightings are however particularly interesting given the state of our world. 

First, at the New Yorker Summit, he made the provocative point that the interconnected-ness in the world today makes it less likely to tolerate leverage. 

We have to be a lot more careful going forward, because we have globalization, the internet, and operational efficiency — which cannot accommodate debt.

Second, there is his recent paper that postulates that large financial institutions are more susceptible to ‘black swans’, the large but improbable risks. It uses a non-linear loss function to show that the risks of aggregation are usually hidden but often  outweigh the classical economies of scale.

So if you have a large job for Hadoop but don’t yet have a couple of hundred servers in your garage, Amazon has just launched Elastic MapReduce.

New mailing list @ietf for discussing all matters around the bi-directional web. Right away, an interesting conversation on Websockets. 

Get on it if you are interested in the subject… lotsa smart people talking in there.

Alan Greenspan wrote last week in the FT about how the risk management discipline derived from Markowitz has cracked recently, with the crisis coming from unanticipated directions that were not diversified away. He concludes the need for greater financial cushions, and therefore advises significant increases in bank capital requirements. 

David Bau takes the idea further arguing that blind adherence to Markowitz has created a culture where Wall Street believes that all risks are unsystematic and unknowable and that a bank’s primary mission is to hedge them away, rather than take real risks based on the idea that  they could identify sound businesses and genius ideas better than their peers.

There may be an element of over-generalization in that analysis, but have to agree with the essence of his argument that:

Once everybody believes you can’t beat the market, the market becomes stupid

So Sun seems keen to up the ante from its Sun Grid offering, announcing a new Cloud Platform recently, primarily built out of its Q-layer  acquisition.  The interesting part of the offering seems to be new REST-ful api for controlling the cloud, which they hope to standardise at some point.

But maybe it all will be moot soon? Or not?