Big Switch, Small Apology
A number of interesting comments have been left over the last few days, but I owe an apology to the posters in being so slow to approve them. I have an excuse, which I will offer further down the page, but it gives me an opportunity to write about something I am interested in. It comes out of another book which I enjoyed reading recently called the Big Switch by Nick Carr. The book is a good read and takes some well known ideas and, with good examples, argues that computing will become a commodity provided by utilities. The analogy he draws is with the electrical industry which was originally based on selling private generation units. I didn’t know this, but apparently Thomas Edison’s business model was to sell electricity generators to businesses and that is what sustained the early growth of the industry. It wasn’t until the costs of distribution started to fall and the switch to AC that this model was overtaken by that of the electricity company as a utility. Now of course we just expect to be able to plug in and get the power we need when we need it. So the developments in on demand and cloud computing look to him like the same “big switch” where computing resource becomes a commodity, provided by big utilities such as we are seeing with Amazon and others. So in this new world the idea that an organisation will have its own computers and software managed locally will seem as archaic as a company deciding to have its own power generation facilities. Of course this requires something of a culture shift for people to decide to do their work and store their data “in the cloud”, particularly where that data is critical to the business or maybe the academic reputation of the scientist. So why is this relevant to my feeble excuse for not approving comments for several days? Read on …Like a lot of scientists, I like control over what I am doing, which in my case is scientific software development and drug discovery. My colleague Damjan Krstajic and I have been working on agent-based systems automation applied to drug discovery for 4-5 years now with a technology we have called “the Discovery Bus” which is an implementation of a novel software architecture called Competitive Workflow. The software is on my own servers, in my own rack cabinet sitting in the entrance hall of my house. All of the software on all of the servers is open source and I have complete control of everything we do. My University research group uses these servers and in addition I recently managed to buy a second hand, 15 node MacMaster G5 XServe cluster for us to use for our virtual screening, shape matching and pharmacophore work. The cluster was a one time opportunity and I grabbed it without worrying about where to put it or any issues about software management. I couldn’t find space in a server room, and a local data centre quoted me a hosting cost of more (per year) than I paid for it. So I borrowed an office from a colleague, got it moved in and opened the windows.
The cluster runs OSX which I don’t want to upgrade (costs money, no control, not Linux) so the plan was to simply wipe and install Linux. Has anybody else tried to install Linux on a Power PC Apple? It took some time to get round the install hanging because it didn’t recognise the RAID and to get the Head node running, but then we thought we were through. Swap the nodes into the master frame, install, swap back, and so on … Trouble is the nodes didn’t boot when they were put back in the node frames. We worked out that they were hanging because they were looking for a graphics card. We decided to edit the BIOS settings to get round this and that’s when I found out that Apple’s don’t have a BIOS — they have “Open Firmware”, programmable using a language called Forth. At this point all this control over my computing needs didn’t feel so good. This got worse when, with feet up last Sunday, Damjan emailed to say that my home servers were down. Turned out my head server had had a disk failure and was completely dead. A few weeks ago my tape backup unit developed a hardware fault and I haven’t had it fixed yet, so no backups (yes, I know …). A new disk is on order and I live in hope.
On the XServe install, faced with learning enough Forth to write a script to alter the Open Firmware settings I quit and cried for help. Fortunately I found Terrasoft, specialists in Yellow Dog Linux for Power PC, phoned them up and bought their cluster management software (Y-HPC) for Apple XServe (and Playstations) which worked like a dream. We now have a running 15 node cluster which will keep us happy for a while. So thank you Aaron from Terrasoft!
Its getting a bit warm in the office though …
So that’s my excuse for not updating the blog and leaving the comments so long. Like James Herriott with a breech birth, I had my arms deep into things I really would rather not have to deal with. I still have total control of my computing needs, but with total control comes total responsibility.
Sign me up for the “Big Switch”.
Write a Comment