HEAnet and Google gave talks last night to the UL computer society. They were very informative but not for the non tech head.
First up was Dave and Eoin from HEAnet who covered a whole range of issues from the historical to future directions. HEAnet is the ISP for the Irish third level universities and the talk described its development from a small group which provided the 512 k connection to the Internet right to the current organisation which provides 2 Gigabytes and is aiming to in to 10 shortly. It also revealed the new dark fibre structure which has been implemented. The talk covered more networking principles and related practical tools then any module on the topic so I must genuinely tip my hat to both Dave and Eoin for a top notch presentation that was interesting and engaging.
Thomas from Google Ireland followed with another brillant presentation. He took us from the company's first server to its current clusters on the hardware side, whilst on the software he started with crawling, stemming and word indexs before getting to the good stuff on PageRank as a technique, the query handling process and the Google File System. Within the context of a query he spoke about how the system design approached failure by the use of many parallel cheap computing resources and how programs like BORG where used for process transfer both within the cluster and between clusters. The main point was that files where stored in three copies or chunks and that this had worked as a policy when applied to day to day use. Another fun point is that no where within a Google cluster will you find screws as all the components are tied down with velcro for ease of access which dovetails with the policy of really cheap computer hardware for the cluster so if the part breaks its next but in a much more pragmatic manner due in a large part to the shear size of a standard Google cluster. A normal rack mounted cluster in a data centre will only use about 30-40% of the available racks but a Google setup use them all and when they take a room they fill the data center room as Thomas proudly said. Two of the Q&A topic where very interesting. The first dealt with the removal of the Firefox versus Internet Explorer statistics from Google's zeitgeist but the answer was well I can't really comment on that since we've been specifically told not to. The second was to do with privacy and data issues given the wide spectrum of information Google now has on many users with Mapping, Mail and Chat services as well as the normal query service. The scaryiest answer from this part was that in GMail when you delete a message it is merely marked as deleted due to the versioning approach taken with the file storage. Daniel wrapped up the talks and pointed us in the direction of the sports bar for refreshments.
Out of interest in the pub I asked Dave and Thomas how they found the Irish data retention requirements and how they deal with requests from the Gardai. The answers were of course we assist but the query must be formulated to our data. Interestingly both personally agree that the issues deserved more coverage and discussion so back to my last post and we'll see how Digital Rights Ireland can start carving a niche in this forgotten corner, oft the domain of secret cabinet orders or last minute additions to Dail bills (Garda bill I'm looking at you). A good night and much food for thought so roll on the next set of talks.