Tuesday April 19, 2005

Google game - McDougall's

Richard McDougall (one of my favorite natives of Oz and all around Solaris mega-uber-geek) and I were talking about my little Google Challenge at SUPerG today. In looking at his blog, it was painfully obvious that he is way too precise, and lacks some of the whacky metaphors and humorous monologues that make the game easier. I was able to find Richard a nice two word entry though: Fireengine, dwarfs

bill.

SUPerG - Chips and Salsa

Wow, Marc Tremblay rocks. From a medium to high level of detail, an excellent hour of explaining just *why* chip designs are the way they are. Instead of bragging about how cool, small, and fast some chip is, he takes the time to explain how caches work, and how they don't work. His presentation included some great visuals showing cache hit latency measurements, and penalties for cache misses. He then tied that all back to decisions made in SPARC chip design, and showed where the chips meet the code moving forward (nice trending).

bill.

SUPerG - Off to a Good Start

The Sun Users Performance Group (SUPerG) conference is in Arlington, VA this week. I have presented at many of these biannual geek-meets, and attended almost all of them over the past 7 years or so. If you have the chance to attend, and have any interest in performance topics (from architecture to chips to grids to storage), you should consider attending this geek-fest.

SUPerG is historically (to quote Shahin Khan) intentionally very low in marketing content, and intentionally very high in good technical content. This high signal to noise ratio is a welcome change to the standard vendor sponsored conferences filled with warm fuzzies, and thinly veiled sales presentations masquerading as technical content.

First on stage in the morning general session was Steve Campbell. A rehash of the standard Sun vision presentations, but he kept it moving, and put it into context very well. No "rah-rah, Sun is great" stuff here, just "this is how the geeks in the audience and Sun's stuff fit into overall architectures in the past, the present, and looking forward.

Andy Ingram was second on the stage. Andy kept things at an "architecture" level, talking about trends, Service Oriented Architecture (SOA), the Service Oriented DataCenter (SODC), and drilled his way down to the (slightly marketing fluff) capabilities and features provided in the Sun components of the stack. Andy is *brilliant* at presenting technical details as similes and metaphors. This makes complex ideas seem familiar and "safe". The idea of multi-threading as a waterfall, and large parallel processing as Niagra Falls versus some small (but incredibly high and thus quickly flowing) jungle waterfall. Great slides with simple blobs of needs, categorizations of business needs, and processing capabilities.

bill.