The Deliverator - Wannabee

Theo and I had a discussion the other night that amongst many many topics landed on the subject of cloud computing. I argued that current outsourced storage and computing services just weren’t reliable enough for mission critical enterprise use. While you might save money in the short run, the lack of direct oversight and control over how those services are provided on the back end could seriously bite your business in the bottom line. Today, one of the highest profile of such services, Amazon’s S3 storage service (along with several other of their cloud computing initiatives) was down for close to 2 hours due to some sort of (claimed) ddos attack against their main commercial e-commerce site. To put this in perspective, today Amazon was down for longer than they claim their service should be down in a whole year. Their own S3 service level agreement states 99.99% availability. According to this wikipedia page about the poor use of 9’s for evaluating service availability/reliability, four nines of availability works out to about 52.6 minutes of downtime per year. FAIL

P.S. No I couldn’t do better :)

I first wrote about MRAM two years ago when Freescale Semiconductor announced that they were actually shipping product, albeit extremely low capacity (4 mbit) product. Now it looks like Toshiba has jumped into the game and will shortly be shipping MRAM chips in 128 MB per chip capacities. While this is far from replacing a hard drive, it may be able to replace DRAM in computers and other devices and allow for extremely low power instant suspend/hibernate modes. This could result in very long battery lifetime in devices like PDAs and cell phones while still letting users have instant-on convenience. It also sounds like IBM is jumping into the MRAM game, so it looks like there will be an active marketplace for this technology and support from industry heavy-hitters.

I needed to get away for a while and took a roadtrip. Here are some select pictures. Here are some panoramas. If you want more, I have another 900 or so if you are willing to sit through the slideshow.

Posted here is my schedule for SIFF 2008. If you want to join me for a flick or 20, let me know.

One of the things I like best about my n810 is its ability to serve as a remote access client using RDP, VNC or SSH. With RDP and VNC, the bandwidth requirements are such that a WiFi connection is needed to work efficiently. With SSH, the bandwidth requirements are much lower, such that I can effectively use my N95’s EDGE wireless wan connection. While T-mobile’s EDGE coverage around the Seattle area is good, it is simply not practical to keep the connection up all the time. Even if the coverage were perfect, battery life issues would prevent a continuous connection all day long.

Unfortunately, by default SSH only gives you a single terminal session in which to launch applications. If the connection is interrupted, so is any app you might have been running. What is needed is a way to multitask and keep the application running. Screen fits the bill nicely. With it, I can ssh to a server from my n810, start up multiple command line tasks and before disconnecting, decouple any apps I wish to keep running from the session. I can then move elsewhere, reconnect to the server using ssh and with the simple command screen -r, reconnect all those running apps to my new session. This ability comes in really, really handy. I can leave things like IM clients, IRC clients, email, network monitoring apps, backup jobs, long software compile operations, etc. running all day long and just connect to them as needed. If you haven’t given screen a try, I highly recommend starting off by reading this introductory guide, as official documentation is sparse.

I took some time this evening to reinstall Gallery. A large number of sub-galleries had some level of corruption, and although I could have restored many of them to working order, the degree of time needed to do so was more than I was willing to accommodate. So, I am starting out fresh. I have virtually all the images contained in my original gallery, but naming and organization will change somewhat. Due to Silverfir’s limited bandwidth and my being too lazy to physically visit the server, it will likely take a few weeks to get all the pictures up again. If I am feeling particularly bored some time, I may go back through all my blog entries and re-link to the new image urls. Going forward, Silverfir now has a backup drive, of which I intend to make frequent use.

The first gallery to go up is of pictures from the 2008 Seattle FIRST Robotics Competition.

TRC won 2 matches and lost one this morning. One loss and one win happened without us showing up (something I had feared). Getting the new mechanism working and getting the robot reinspected proved a longer process than the students counted on. We showed up for the third match though and played well, although the new shooting mechanism remained disabled for the rest of the tournament. The new dislodger worked quite well, however. We finished the qualifying match seeded 11th of 30 someodd teams. Because several of the top ranked teams chose each other, we ended up as alliance captain #8. We picked teams 955 and 2147 for our alliance partners. Lunch is finished and playoffs are beginning in a few minutes. More later.

Update: The TRC went on to win one match against the number #1 ranked seed, tie another and then lose the final two to get eliminated from the quarter-finals. In an echo of Portland, one of our partners suffered a catastrophic breakdown and couldn’t continue, so we called in a sub, which ended up being team 753. The replaced team, team 955, stayed on the field with us for the rest of the matches and was obviously lending a hand tweaking robots between matches. Thanks for being great sports 955! All in all, I am very happy with how the season went. Never before in my experience have the students been as directly involved in all aspects of club operations and the building of the robot. We finished strong enough at two different regional competitions to be an alliance captain. The drive team played some awfully scrappy matches, making good use of the robot they had, vs the robot they wanted. The team made an extremely gutsy decision to completely redesign the robot between the Portland and Seattle regionals and then worked methodically towards building their lightweight (24.9 pounds!) functional design. In 20 hours plus time in between matches in Tacoma, the TRC built nearly a completely new robot. I am extremely pleased at the teamwork displayed in the pits and the clear desire to realize their vision, even after recognizing that it likely wouldn’t be fully functional till after the competition. When FIRST has instilled in a team something other than the motivation of a quest for another shiny medal, then it has done its job.

I originally wrote this as an update to the previous entry, but later felt it deserved its own.

Turns out the TRC had one more match today at the end of the day, of which I was not immediately aware as I was already back in the pits working on the new mechanism at the time it occurred. Evidently there are some hurt feelings over this one, not so much over the loss of the match, but because one of our alliance partners, rookie team 2555 wasn’t doing well in the tournament and just packed up and left. I believe in the four years I have been involved, this is the first time I have seen a team just outright leave. I have seen teams with completely non-moving robots still place it on the field in order to cheer what they built or just to cheer their partners. I’ve also seen teams suffer major robot breakdowns, leaving them unable to compete, stay through the end of the tournament and then win an award. We have now had a couple matches where a team hasn’t showed (it is usually because something broke between matches and they were back in the pits repairing something), but to see a team (or not see) just up and vanish and leave you hanging without so much as a word really stings. 2555 is scheduled for 3 more qualifying matches tomorrow and I am not sure how the refs will handle an outright withdrawal from the regional.

Many teams suffer breakdowns during the course of a regional or end up with mechanisms that don’t work quite as well in practice as they did at the machine shop, but what has struck me again and again is how FIRST teams meet adversity, come together as a team and how every other team there pitches in to help that team succeed.  There have been near mythical incidences where a team’s robot has been lost in shipping or destroyed in a freak accident and they show up at a regional and a bunch of teams join hands to build a new robot right there on the spot. For me, there is no clearer practical  demonstration of gracious professionalism in FIRST.

This evening in the pits, the TRC with help, advice, quickly volunteered tools and supplies from that other TRC (the Tahoma Robotics Club), Newport Robotics Group and Team X-Bot together transformed our robot. Our old, ineffective mechanisms were surgically sundered from our drive train and the new catapult and ball dislodging mechanism attached. A pin release mechanism for deploying the dislodging mechanism was improvised and many adjustments to the elastic tubing and winch subsystems occurred after some impromptu testing on and off the practice field. Many thanks to Dave, aka Shrek 4, the pit boss, the inspectors and other crew who were extraordinarily helpful and accommodating this evening. Also thanks to Mike Tseng who stayed late to help code and to Patricia and Chuck who did a last minute run to the hardware store for this that and the other thing. There is substantial work to be done in the morning in completing reassembly of the robot, particularly in reinserting and hooking up all the electronics boards and then we have to get reinspected prior to competing. Whether we perform brilliantly, run into or through a wall or just don’t run, we will be out on the field tomorrow, richer for the experience and the gracious teams of FIRST will be there to welcome us.

P.S. I sliced up my hand pretty good with a hacksaw, but at least my eyes were safe ;) Thank you FIRST!

It is the first day of qualifying matches here at the “Seattle” Regional (held in Tacoma). The TRC substantially completed the new mechanism last night, but there wasn’t enough time to swap it onto the existing base and pass inspection before the start of qualifying matches today. The plan is to tweak the new mechanism throughout the day and then swap it in between 3 and 6pm this evening after matches have finished for the day. As with Portland, I will try to provide updates throughout the day.

Update 1: The TRC won its first qualifying match of the day 32-0.

Update 2: The TRC lost its second qualifying match of the day 0-52. One of our partners, team 1570 did not show up for the match and our other partner scarcely moved the whole match, so it was effectively 3 on 1. Our bot autonomously crossed 3 quarter field markers during hybrid mode, for what should have been an obvious 12 point bonus (4 for each line crossed), but we only got credited for 1 line. The field announcer even commented on this, but the score was not changed! We accrued two 10 point penalties during the match, which were well deserved. Jonathan seemed to be driving in a bit of a panic and was trying to really whip the robot around the corners to do laps quickly (our only way of scoring points). In the processes, he ran into the center divider and backed up in TWO cases, garnering a penalty each time. He really needs to take it easy and go around the field at right angles. This one was a legitimate loss, but I have been distinctly unimpressed by the quality of the ref’s scoring today. In one particularly memorable incident, team 1510 was knocked over with its arm fully sprawled out across the floor, which put it WELL over the 80 inch rule. It sat there for most of the game and none of the 8 or so ref’s caught the highly obvious infraction.

Update 3: TRC lost the third match of the day 30-40 in a strategically played match. Our team had no hurdlers and the other team had two, so our alliance largely played keep-away with the opponents trackballs. At one point, I felt like we might have pinned and compressed the opponent’s ball between our robot and one of the walls, but we didn’t get called for it. Unfortunately, with two of our robots guarding the opponent’s balls, it didn’t give us much of a chance to run laps. We did manage to keep one of our trackballs on top of the overpass at the end of the game, resulting in a 12 point bonus for our alliance, but in the end the opponent’s lapping did us in. The autonomous mode which worked very well in the second match resulted in us running into a wall this match. Pity too, as if we had crossed three lines we would have won.

Update 4: I believe the last match doesn’t count against us as it was our “filler” match, similar to the one we won at Portland which didn’t count for us. TRC tied its fourth match today with an official score of 28 to 28. I’m pretty sure we actually won it by a fair margin, as our alliance crossed 3 lines during auto, had 1 trackball up at the end for a starting score of 24. Our alliance accrued one 10 point penalty during the match, which would have brought our score to 14 not counting laps. Individually we made many laps during the match and also pushed a blue ball across at the end. I think our legitimate score should have been around 40. I am trying to keep my temper in check and be as dispassionately objective as possible, but this is a first year regional and imo it is showing.

Update 5: TRC won its fifth match today 28 to 2 in a well driven match. Jonathan seemed to have really good control of the robot this time around and managed to roll trackballs over our finish line twice under heavily contested conditions. The robot’s autonomous mode seems real inconsistent, with it again slamming into the wall. This was the last match before the lunch break. I’m off to get a sandwich. More latter.

Update 6: In our first match since the lunch break, the TRC squeezed out a win by I believe the slim margin of 22 to 18. Our alliance scored a lot more, but got hit with 20 points in penalties (thankfully, we were not the offending party). In an interesting chain of events, outgoing senior Erik Wells-Thulin ended up being one of the drivers (and I presume the defacto coach), as both Ian and Alex were giving their Chairman’s Award presentation at the time of this match.

Update 7: TRC won its final match of the day, squeezing out yet another victory by a narrow margin. Final score was 20-18. TRC finishes the day with an official record of 4 wins, 1 tie and 1 loss. Now the fun begins as the team works to completely remove the existing superstructure of the robot to replace it with a new one!

As mentioned earlier, the TRC spent some time between regionals designing and fabricating a completely new mechanism to go on top of the robot’s drive train. After a great deal of assembly work, with everything screwed down to a forklift pallet instead of the robot, the mechanism is nearing completion. We were able to test it on the practice field here and it not only hurdled the ball over the goal structure, it cleared the field entirely and hit the wall in back of the field. It was mighty impressive to watch. If anyone in the world needs a good siege engineer, the TRC will consider all offers! The ball launcher is easily the most powerful one here at the competition. Unfortunately, we are down to less than 2 hours working time with less than an hour tomorrow before our first match.

A few not so minor details such as a improved release mechanism still need to be improvised. Then, comes the small matter of removing the substantial existing mechanism from the robot and mounting and wiring up the new mechanism, switching and testing the robot code to control the new motors and whatnot and then go through inspection. In short, it is extremely doubtful that this could happen without missing a few qualifying matches. Showing up with an impressive robot later in the day might garner us some support, but is not likely to make us any friends if we don’t show up for a match. My personal opinion is that they should play tomorrow with what they have and then switch the mechanism at the end of the day or wait until after the season and then mount the mechanism. It would make for a very impressive robot for public demonstration purposes. Watch as the robot picks up a 40″ diameter ball, winds up and then hurdles it 40″ through the air!

Anyways, my laptop battery is almost dead, so I will sign off for now.