With Jeff Barr back with a bang after a month long vacation and Conference going, the AWS developer chat at Secondlife is back with renewed enthusiasm.
Today’s chat was dominated by EC2 discussions with the usual rants about S3 mounting support( or the lack of it), need to have an external ‘orchestration’ servers to manage the AWS/EC2, and need for scaled down and scaled up version of EC2 instances.
There was interesting feature request tighter integration of FPS account and AWS bill payment if and when the the Amazon Flexible Payment Service(FPS) supports non-US accounts. This way, currency conversion overheads and credit card processing overheads could be reduced.
There was an interesting discussion on protection mechanisms and bandwidth metering in the context of denial of service attack(especially Dns Denial of Service(DDoS) attack):
[10:33] Jeffronius Batra: Alerque, are you concerned about our ability to handle the DDOS, or our response to it?
[10:34] Alerque Thorne: Actually my first concern was bandwidth costs. Is there any filtering done before the metering?
[10:34] Alerque Thorne: Or would all DDOS traffic be my responsibility to pay for and then filter?
[10:34] Jeffronius Batra: I don’t believe there is. We would meter it at the point where it reaches your EC2 instance.
[10:35] Jeffronius Batra: Is there a better way to do it?
[10:35] Alerque Thorne: I currently have my services on a 45mb OC3. It is not unusual for that to get filled up with traffic durring an attack. This doesn’t sound like a fun thing to have to cope with on EC2 if I use it on the forward side.
[10:36] Jeffronius Batra: That’s a lot of bits, even more than my teenagers use at home 🙂
[10:36] Jeffronius Batra: I would have to talk to the EC2 team about this. If you want, you can send me an email (jbarr at amazon dot com) with more details.
[10:36] Alerque Thorne: Yes, it’s a pile. It doen’t happen all the time, but probably monthly.
[10:37] Jeffronius Batra: Are you making the wrong people get mad at you??
[10:37] Alerque Thorne: Simple answer: yes.
[10:37] Jeffronius Batra: Ok….
Answering a question regarding Amazon SQS, Jeff Barr built a display panel with GigaVox architecture diagram.
[10:17] Jeffronius Batra: That’s the Gigavox Media architecture, build with EC2, S3, and SQS.
[10:17] Jeffronius Batra: They upload raw podcasts, store them in S3, transcode them from MP2 to MP3, and then assemble them into finished podcasts with introductions, ads, and so forth.
[10:18] Jeffronius Batra: They do a fresh “build” of the podcasts every day, scaling up EC2 instances as needed.
[10:18] Jeffronius Batra: Like I said, they monitor processing time (beginning to end) and use that to regulate the number of instances in operation.
[10:19] Jeffronius Batra: They told me that they spent less than $100 on services to build this entire architecture.
[10:19] Ciemaar Flintoff: wow, that include ther EC2 costs?
[10:19] Jeffronius Batra: I believe that it did. Doug Kaye was the architect (he’s been in these chats before).
[10:20] Jeffronius Batra: The very cool thing about these EC2-based models is that costs rise directly in line with usage.
[10:20] Jeffronius Batra: You don’t have to buy a big pile of servers just in case you need them.
[10:20] cperciva Perfferle: jeff, the costs rise in line with usage once your usage is high enough to need more than once instance 🙂
And the usual request to meet up as many developers 1:1 was there too:
[10:53] Jeffronius Batra: This goes for everyone on my team — Mike, Jinesh, and I always seek out 1:1 meetings as part of our travels. Our schedule is at http://evangelists.wetpaint.com/ and you can simpy edit the appropriate wiki page.
[10:53] Jeffronius Batra: We find that the direct contact with developers, be it through SL, face to face meetings, or forums, is really valuable.
[10:54] Chase Marellan: What do you do in those meetings? Just what we do here? Or is it more targeted?
[10:54] Chase Marellan: Or do you just have dinner. 🙂
[10:55] Jeffronius Batra: We generally discuss specific questions and issues, sometimes things that they would rather not share in public.
[10:55] Jeffronius Batra: I like to get a good understanding of how they are using our services — what works and even more importantly what doesn’t work.
[10:55] Chase Marellan: Gotcha, thanks.
[10:55] Jeffronius Batra: We write up what we learn each day and get it into the company ASAP.
[10:56] Chase Marellan: Cool.
[10:56] Jeffronius Batra: Yes, it works wonders to be able to talk about real customers when making the case for new features. No ivory tower for us.
[10:56] Jeffronius Batra: And we do have dinner (or lunch or breakfast) too!
[10:56] Chase Marellan: 🙂 Obviously it’s working. 🙂
[10:57] Jeffronius Batra: Oh yeah, definitely.
Too bad my suggestion for mashing up AWS with tiddlywiki got drowned in the interesting discussion. I’m discussing the mashup with some of the geeks at the OpenCoffeeClub Chennai group. It is simple, definitely has lot of coolness appeal if not money making opportunity.