My good friend Joe Weinman who is a cloud junkie and obviously obsessed with the cloud, sent the best (and nerdiest) email on my decision to exit from the news reporting business. Joe’s secret desire is to be a standup as you can find out at Structure SF where he is our MFMC (most favored MC). Anyway here are some bits from his amusing email/letter:
Om Malik recently announced the release of Om 2.0, and limited ongoing support for Om 1.0, which has been deprecated effective immediately. We thought we’d delve into some of the speeds and feeds of the new model and contrast it with the old.
Hardware – The hardware is based on a popular crowd-sourced open design that has used evolutionary metaheuristics and a quaternary code to evolve slowly over millennia, but arguably remains essentially unchanged. Although Om’s basic hardware design is now several decades old, it has stood the test of time. Various modules (e.g., stream processing, oxygen intake, visual and auditory data collection sensor subsystems) are replicated for high availability. The chassis is largely calcium-based, and the entire system is air-breathing, water-cooled, and is powered largely by carbon-based molecules, including caffeine. It has a top-of-rack switch and a dual-core neuromorphic parallel-processing computing architecture with the processing power of close to 100 billion neurons arranged in a three-dimensional matrix. The kernel enables limited multitasking. All of this capability resides on a mobile platform, which fits conveniently into the space needed by about one rack.
Software – Software is where the real differences between the first-generation Om and this new release are apparent. The first release came to market in 2006, and was largely devoted to information processing. This 8-year effective life compares favorably with other popular software products such as Windows XP, for example, which was upgraded after about 6 years. Based on a wide-area network enabling interfaces to thousands of data collection points, Om 1 was focused on collecting data, processing it using advanced trend analysis and information processing algorithms and unstructured data processing heuristics, archiving the data in a column store in human-readable format via WordPress, and then distributing results globally over a proprietary content delivery architecture using a public HTTP/HTML RESTful interface. The new release will still have the same data collection and data processing capability, but the entire output philosophy has been completely rethought and retooled to enable high-speed switching of financial flows to appropriate targets based on voting algorithms which are being maintained as trade secrets. This upgrade, which has been in beta for a few years, has now been officially launched effective February 20th.
Operations and Management – This new software upgrade has enabled substantial modifications to hardware operations. While Om 1 was designed to be “always on” and was highly interrupt-driven, the new hardware will now be operated so as to enter sleep mode on a daily basis, reducing power requirements and chassis stress, as well as enabling preventive maintenance. This is expected to extend the life of the system substantially.
Although the new system will still interface to the GigaOM board, it will largely be operated behind a firewall and out of public view. However, it is still backward compatible with older applications from its first release, and will be demoed occasionally at the GigaOM.com site and at GigaOM events such as Structure Data and Roadmap.