It's hard to believe a company would "admit" to running all of these services on a single server. There's no way I would pay for a CRM or Ticketing App (relatively mission critical for many companies) that wasn't highly-available.
One also has to take into account how many systems that are distributed over n servers have actually n points of failure. My estimate would be "quite a lot".
Hopefully they have a staging environment running somewhere internally at least.
You can easily get near 8GB for way under $200/mo using reserved pricing. A one year reservation, heavy utilization m1.large gives you 7.5GB of RAM for $40/mo and $676 up front, or about $100/mo total.
Paying the reservation fee does just get you reduced per minute rates, but the fee is not $200/mo. The reservation fee in this example is $676 for an entire year. Once you pay that, your hourly rate for a single m1.large goes down to $0.056. My final $100/mo total is based on 24/7 uptime on that single server you've reserved.
Which will happen, when BizSpark runs out.
I've personally been in Microsoft's "let's give developers everything for basically free" programs since 2006, and am not worried that I'll suddenly face a crippling bill down the line.
Shafting little startups for a lousy grand or two really is not their business model. Think about it for a second and it'll make sense.
EDIT: To be clear, I'm not advocating gambling your business on Microsoft's goodwill here. For the same reason I don't advocate building a business that a single bill for $1,000 can kill.
The worst case result of MS pulling the rug on all their BizSpark-esque programs is an extra couple thousand dollars a year in the expense column. Here's hoping nobody here works for a shop where that would be a big deal.
But, instead, the realistic position is to understand Microsoft's tools and programs will be a decent choice for years to come, just as using FOSS would be.
Open source community supported tools are completely different. The price and license are very clear - $0 up front, pay for support if you need it. Don't want to pay for support? No problem. Need a new feature? No problem, pay someone regular rates to add it. You can't lose access to anything because you already have everything. You can't be sued into paying because the GPL specially disallows the creator from suing you.
A valid business plan can be created that involves the pricing structures that Microsoft gives. You can have a plan that involves using a program from MS that gives you their software for free now, with the idea that you will eventually be able to afford to pay for them. Any company worth existing should be able to afford it eventually.
And, hey, look, it even happens that those windows of time can be extended. No one is saying that it would be a good business plan to rely on programs like BizSpark never ending... but it's still worth mentioning that MS has been forgiving if things don't go exactly as planned, and provides a way to delay the fees kicking in.
Of course, using free software might widen your "trusted base" compared to the situation where you go full MS.
MS SQL is free since we're in BizSpark.
The rest of your post could be read as apples with oranges, Charm was expensive because they made massive architectural mistakes. End of story.
This is not to say that what you are doing/saying is incorrect. Your stuff works and is able to cope with the load thrown at it on one server.
To be honest I'm quite mystified why startups boast about using n instances to serve a few thousand hits. Node, some python frameworks and ruby are ridiculously wasteful[1]. All it tells me is that you've not though about the best way to do things, you've just used the latest tools and smooshed them together. If you had to have more than one server to handle 2[2] tickets a minute and 1 hit a second on your website I'd be questioning your programming skill
The one thing I do take issue with is your implied claim of high availability. Hourly backups, and no hot spare or active active replication is not high availability. Yes you can rebuild an instance in 5 minutes, but only in best case, and you've still lost up to an hour's worth of user data (around 60 tickets). That is a commendable recovery rate, but its still downtime.
[1] depending on usage, of course. [2] using stats from the article
Losing client data is the worst thing you could do as a SaaS. The second worst thing you could do is to not be online - e.g., as someone pointed out; because one app (or even something out of your control) took out all your services.
MS is not really the story here. Cost of HA in typical Linux vs. MS stacks, which wasn't addressed at all in the article, is what is relevant.
For how much longer? And then what?
Depending on what version of SQL Server he needs, it's either a $0.136/hr additional cost to add Web Edition or $0.610/hr additional cost to add Standard Edition. That works out to an approximate $97.92 or $439.20 additional cost per month. So, not a huge jump in cost.
I think this is just poor design and probably not an MS stack issue. But could this be an MS community thing, where they generally prefer proven technical solutions from 2001?
Until you hit the limit of what you can do on commodity-ish x86 servers, and would need to start buying Gucci kit from Sunacle or IBM or whoever, at which point your CFO is going to start crying like a baby.
Additionally, like others have mentioned, a simple memory leak from one app will affect every other app on this box. Also, it doesn't sound like there's any redundancy. What happens when the HD blows up?
If they fired these app up as separate VMs, they could at least scale each to what their needs are and only pay for what they are using.
I manage an app architecture that consists of several Heroku apps along with a couple DBs at mongoHQ for under $100/month. Heroku definitely isn't the cheapest solution, but it allows me to scale horizontally and takes the massive load of sysadmin tasks off my back so I can focus on development.
In reality most people are never going to deal with that level of scale.
Shouldn't that be more than $200 a month?
This has nothing to do with MS, but why would you set up such an architecture? If one of the apps hits a memory leak (their HeatTest project did), than all of your other critical services/apps are interrupted.
That costs us about $200 a month. By the way, that's Amazon
What Amazon configuration gets you 8GB or RAM for $200/month?? Am I missing something?
Database servers for all of the above
MS SQL licensing costs are crazy stupid. You either (a) have to fork over about $2-3k fee or (b) the cloud based cost is an extra ~$150/month on top of the standards windows vm cost.
I'm genuinely curious about all of the above, because I am working with a MS-based startup and would be good to know.
But most importantly, because the customer doesn't give a darn.
100% agree.