So, you have all the web2.0 ingredients: Silly name: check, Hosted service: check, Big blog arses kissed: check, working website: doh!
It’s a bit of a shame for poor CrazyEgg because it sounds like a pretty fun service, and certainly Brian Benzinger thinks it’s cool. Let’s hope they get their hosting problems sorted out ASAP as it’s almost certainly something a lot of the Performancing community would be interested in aswell. Keep an eye on it, and tell us what you think when it’s back up.
Note: If the link above works for you, at the time of writing it was giving either a blank page or a 500 server error.
powered by performancing firefox
22 thoughts on “Launch Rule #1: Make Sure it Works”
@hnshah: OK, take your time 🙂 I am really interested to read about the issue and the solution.
When Zooomr launched their update they (also?) had a multiple day delay because of some DoS attack.
@Chris: I know that it is another dimension
I would also be interested about the PMetrics setup as the service is really incredible smooth and fast responding.
There are service providers that do shared web farm hosting too Markus, but something on this scale really needs a dedicated multiple server setup.
We will tell everyone what the problem was, once we have it fixed.
I have to throw in a hand for shared hosting … Sitting on a machine with another 99 domains can give you the power of the big server if the load is shared dynamically. You only get the 1/100 guaranteed but in a shared environment you can expect to have more power then needed and a better machine then the normally offered cheap dedicated server.
I know that we are talking other dimensions here …
I’ll email you guys.
Heh, count me in for noseyparker too. Brian do you really want to host the video yourself?
No begging required, i’ll help just to be nosey…
I’ve just spent a week digging through hosting options to make sure that a video site could handle this type of success. I’d really like to know if there’s something I’ve failed to consider.
Or I may just have to beg Nick for help.
Yeah, getting the 5% for sign-ups is not too bad. It’s the rate for ‘paying’ sign ups that’s more interesting! Actually, it’s not commercially sensitive, so I can say… It’s just over 1%. Which, from this article: http://www.thinkvitamin.com/features/webapps/how-to-measure-the-success-of-your-web-app looks like it’s about average.
Hiten, as you’ve not told me what the issue is, i’ll have to take your word for it that you couldnt have forseen it. Though i’d urge you to let us know what the problem was once you’ve fixed it so we all might learn from it. Good luck with it anyway..
spx, 5% sounds like a good rate. Congrats on that!
I’d like to take a look at your service, and wiil.
Yeah, good point! (We’re getting about a 5% average from all sources at the moment, so I’d used that as an estimate).
Nice use of ‘pfff’ too.
I did a blog post about this just yesterday, there is nothing like a real world load test from Digg, and the “blogosphere”. We actually have planned for scaling and could not have foreseen / tested for what is happening right now. Anyway we are working on getting things stablized as fast as we can.
By the way, just to clarify, we are NOT on a shared hosting plan, or just one dedicated box, we have a large infrastructure in place, and that is not the issue that we are tackling at the current momment.
5-10% sign up rate? From Digg? Pfff… more lke .5 if you’re very lucky.
We run Metrics with over 13,000 blogs in the system sending us data continuously. So we know a little bit about this aswell….
We handle incoming data on just 2 front end servers. We have a more complex setup for processing that data, but if we expect our site, and service to be usable when we get digg’ed, then as you say, it’s our duty to provide a proper service.
Ta for the response.
Now come on, we don’t need any straw man arguments here!
Speaking from experience, a single dedicated server with 4 fast CPUs, a ton of memory, and ultra fast disks could still struggle – for this type of site, with this type of load. Now, ideally of course you want a server farm for this type of service, but how do you afford that without huge initial investment? Is there maybe a gap in the market for someone who can offer the resource, for perhaps a cut of the revenue? Would seem to make sense to me, with so many people releasing great Web 2.0 style services these days… Anyone up for offering that?!?
Anyway, if you look at what the service does (assuming the service is being run from the same server as the site); it collects data from other people’s sites.
So, for every 10,000 visitors you get, about 5-10% of those will sign-up to the service, and give you their traffic as well. It’s not just about handling the straight website traffic – it’s also the cumulative re-directed data that’s building all the time (e.g. if someone signs up with another site that gets 10,000 visitors an hour, you get some of that ‘traffic’ too).
But you’re absolutely right – I can make as many ‘excuses’ as I want; at the end of the day, if people like us launch these services, we have a duty (as a service provider) to provide a service, no matter what the demand.
I guess I was just trying to spread a little understanding! Good luck to them.
At least now they have a prettier page.. no more the nasty 500 page.. but a “unscrambling some eggs” page 😀
It’s a good test, to be reviewd here at perforancing. You guys should think about doing it as a pay service 😛
Ricardo
…and?
Of course there is. If a service crashes or becomes inoperable due to being digged, it’s time to move that service from the gaffa taped rubberbanded mickey mouse hosting services it’s currently on to something a bit more suitable.
The reality is that digg traffic just isn’t that bad. What brings sites down is the fact that few sites ever get that amount of traffic so hosting services dont have to tune their servers to handle it.
One dedicated server with adequate bandwidth does not cost the earth, and can easily handle the traffic a site like digg will send it.
I don’t wish them any ill will, but saying that it’s not their fault is just not right. It is, clearly.
They wanted to be able to stay up with all those people actually using the service though right? It wasn’t a proof of concept but a working service?
To be fair to crazyegg (we run a similar service, clickdensity), sometimes you can’t put the resources in place.
If you look at what’s happened – they’ve got front page digg, mashable, and a bunch of other major websites pushing them traffic at the same time. One of these would normally make a dynamic service, running on a dedicated server, struggle.
Hosting costs a lot of money, and you can’t easily get ‘temporary’ contracts (e.g. for a server farm) for launches. Without a huge chunk of external investment (and surely one of the philosophies of Web 2.0 is that you don’t need huge investment to launch), there’s no possible way to adequately prepare for this level of traffic. I’ve seen much bigger sites/infrastructures get taken down by less.
@Nick: Your CE link has an ” (%22) at the end.
It shouldnt be a danger though — you should just have thought it through and have the required resources in place.
If they’re on some crummy shared server, it’s poor planning. I can’t see how Brian can call crazyegg “highly anticipated” at all, but surely they didnt think it would be a no show when they launched heh..
This is the danger when you generate too much buzz : a crowd of webmasters try to sign-up when the service is available…
With that link, there are gonna be another 500-1000 people in the next hour dying to get a shot at that site, and they are gonna come back tut-tutting CrazyEgg 😀
I had been to CrazyEgg when it was still in beta. I liked the concept of the Heatmap, although I did realize even then that this would involve a lot of client-side scripting => bulky pages.
But, I think I will reserve my opinions till I get a shot at that.
Regards,
Shri.
Comments are closed.