Basecamp-Maker 37Signals Says Its 'Cloud Exit' Will Save It $10 Million Over 5 Years (arstechnica.com) 83
An anonymous reader quotes a report from Ars Technica: 37Signals is not a company that makes its policy or management decisions quietly. The productivity software company was an avowedly Mac-centric shop until Apple's move to kill home screen web apps (or Progressive Web Apps, or PWAs) led the firm and its very-public-facing co-founder, David Heinemeier Hansson, to declare a "Return to Windows," followed by a stew of Windows/Mac/Linux. The company waged a public battle with Apple over its App Store subscription policies, and the resulting outcry helped nudge Apple a bit. 37Signals has maintained an active blog for years, its co-founders and employees have written numerous business advice books, and its blog and social media posts regularly hit the front pages of Hacker News.
So when 37Signals decided to pull its seven cloud-based apps off Amazon Web Services in the fall of 2022, it didn't do so quietly or without details. Back then, Hansson described his firm as paying "an at times almost absurd premium" for defense against "wild swings or towering peaks in usage." In early 2023, Hansson wrote that 37Signals expected to save $7 million over five years by buying more than $600,000 worth of Dell server gear and hosting its own apps.
Late last week, Hansson had an update: it's more like $10 million (and, he told the BBC, more like $800,000 in gear). By squeezing more hardware into existing racks and power allowances, estimating seven years' life for that hardware, and eventually transferring its 10 petabytes of S3 storage into a dual-DC Pure Storage flash array, 37Signals expects to save money, run faster, and have more storage available. "The motto of the 2010s and early 2020s -- all-cloud, everything, all the time -- seems to finally have peaked," Hansson writes. "And thank heavens for that!" He adds the caveat that companies with "enormous fluctuations in load," and those in early or uncertain stages, still have a place in the cloud.
So when 37Signals decided to pull its seven cloud-based apps off Amazon Web Services in the fall of 2022, it didn't do so quietly or without details. Back then, Hansson described his firm as paying "an at times almost absurd premium" for defense against "wild swings or towering peaks in usage." In early 2023, Hansson wrote that 37Signals expected to save $7 million over five years by buying more than $600,000 worth of Dell server gear and hosting its own apps.
Late last week, Hansson had an update: it's more like $10 million (and, he told the BBC, more like $800,000 in gear). By squeezing more hardware into existing racks and power allowances, estimating seven years' life for that hardware, and eventually transferring its 10 petabytes of S3 storage into a dual-DC Pure Storage flash array, 37Signals expects to save money, run faster, and have more storage available. "The motto of the 2010s and early 2020s -- all-cloud, everything, all the time -- seems to finally have peaked," Hansson writes. "And thank heavens for that!" He adds the caveat that companies with "enormous fluctuations in load," and those in early or uncertain stages, still have a place in the cloud.
Why did this take them 10 years to figure out? (Score:5, Interesting)
We already knew the cloud was just an easy but expensive solution for startups or an expensive way to temporarily scale out one's infra.
Re:Why did this take them 10 years to figure out? (Score:5, Insightful)
It's one of those things that "everyone knows" but no-one really knows. In our case (small obscure company) all it took was some bored geek sitting down and pricing out what a cloud version would cost once you went past the kiddie-level joke stuff (1 CPU, 2GB RAM, almost no disk space) and how rapidly the graph climbed after that, vs. the cost of a second-hand Dell server or two and half a day to set it up, and it was a complete no-brainer.
But most companies who aren't run by geeks won't know that, and in particular anyone who's used to spraying around hundreds or even thousands of VMs (most of them obsolete, idle, or forgotten) and duct-taping the services on them together to achieve the aggregate performance of a 386 on the equivalent of a farm of Cray supercomputers, just thinks that's the way things are done, and you can't convince them otherwise.
Re:Why did this take them 10 years to figure out? (Score:4, Insightful)
The cloud makes sense if you don't have a mature IT department. You're basically just outsourcing your IT work to the cloud provider until you can grow your own IT department, which doesn't happen overnight. And yeah, you'll spend a lot more on cloud services, period.
The same can be said for applications as well. It's even harder to build an internal application development team than an IT department, but if you do you'll save a crapload on licensing costs in the long run. The reason it's usually harder is developers are either chronic under-performers and/or impossible to work with, or they're both hard to find AND expensive. What makes things even harder is the shittier the developer, the more they have themselves convinced that they're really good at it.
Re: (Score:3)
It was a small company (not really a start up), and it did well. This was pre-cloud
Post-cloud, guess what.... we still have servers on prem. We still run clusters, and data warehouses, and high speed computes.
They looked at cloud solution over VMWare, 1 million a year for cloud on our then, current config. It doesnt even compare, and I think thats why broadcom raised their price on VMWare, to make c
no (Score:2, Insightful)
Broadcom raised their prices because they are basically milking the existent customer base into oblivion. I.e., extracting all profits in the short term and leaving the business as a smoking crater, just like a private equity firm does. They don't want the customers that have an easy time jumping ship; the "long tail" is not profitable for VMware because those customers need new features, support, marketing etc. There are other customers that spend literally millions if not tens of millions a year on VMware
Re: Why did this take them 10 years to figure out? (Score:2)
That makes no sense. I cut my teeth on in house server builds, setups, configurations, and management.
It was a small company (not really a start up), and it did well. This was pre-cloud
That's nice and all, but most businesses are run by people who have no idea how to do all of that, nor do they have any idea how to manage a department that does
Re: (Score:2)
This. If you do not have at least 3 FTE sysadmin, its usually a good idea.
Re: (Score:3)
Until cloud raised in price to close to or exceeding on-prem consistent loads. They keep getting sales because big companies are averse to CAPEX these days.
Re: (Score:2)
I'm not talking about consistent loads either. I'm talking about pricing that comes close to it even if you're only paying for bursts.
Re:Why did this take them 10 years to figure out? (Score:4, Informative)
It's one of those things that "everyone knows" but no-one really knows. In our case (small obscure company) all it took was some bored geek sitting down and pricing out what a cloud version would cost once you went past the kiddie-level joke stuff (1 CPU, 2GB RAM, almost no disk space) and how rapidly the graph climbed after that, vs. the cost of a second-hand Dell server or two and half a day to set it up, and it was a complete no-brainer.
It wasn't such a no-brainer back in 2015 when the "migrate everything to the cloud" started. I was working for a company that did stuff like resell server hardware and we really *could* sell customers cloud migration projects where it was relatively easy - no fudging the numbers - to show that if we just migrate everything you have running in your local datacenter to Azure as VMs (+ AD/Exchange to Azure AD and O365), it really is cheaper than buying new set of servers, maintenance contracts for them and then replace them in 10 years time at the supposed EoL.
The key factor here was that those numbers worked out *only* if the prices were not increased. It was quite obvious that this is all going to be temporary - the cloud prices were only fixed for a few years, tops, and price hikes were more or less inevitable.
Smart customers started migrating all those VMs to SaaS while the VMs were still cheap to run on the cloud. The ones who had taken it as a datacenter replacement and just running VMs were pissed after a few years...but it's not like there was any guarantee on pricing (and frankly, anyone with a brain could see that once there's enough captive audience, the prices were bound to go up).
Re: (Score:2)
True. In our case we did the maths only a few years back when it came time to replace some of our servers, and the question was go cloud or keep it in-house. If we'd done it in the time frame you mention, before the cloud providers had had time to optimise their setup to squeeze very possible cent out of their customers, we may well have gone to the cloud and now be stuck there.
As an aside, it's amazing the level of performance you can get for knock-down prices from servers only a few years old that some
Re: (Score:2)
Re:Why did this take them 10 years to figure out? (Score:4, Insightful)
Meanwhile in the public sector, we're still ripping shit out and putting it in 'the cloud'. I'm not a manager. I don't know what the reasoning is, I suspect it's so they can not hire and tell insurance they're in cybersecurity compliance. We have one in particular who does not understand that it doesn't matter how secure your shit is if no one can use it.
Re:Why did this take them 10 years to figure out? (Score:4, Insightful)
Considerations for moving to SAAS or cloud mostly had to do with management, not with bottom line cost:
- Capex vs Opex (this is what business school taught us)
- Don't want to run a large IT dept (we're lazy and incompetent)
- Blame for any failure falls on the provider (we're cowards)
Re: (Score:2)
It wasn't such a no-brainer back in 2015 when the "migrate everything to the cloud" started.
It WAS a no-brainer back in 2015. Many of us here predicted exactly these circumstances would come to pass. Once the server and bandwidth renters got you in their claws, prices would steadily rise. While the term surge-pricing had yet to be born, it was obvious that the practice was going to apply.
It was quite obvious that this is all going to be temporary....
Yes it was, which made it a no-brainer to stay away. Prices were low because, shocker, they were trying to get people locked into services that would be too expensive to extract once prices started to dramatically
Re: (Score:2)
Right it did not take a lot of insight to realize that Azure/AWS were priced aggressively to eat the enterprise datacenter market. That really should have been obvious, and I think most IT people understood it; but got told no you're doing this by the bean counters who did not realize that those price points persisting were not realistic especially in light of only two *real* players.
The could simply trades capex for op-ex in the same way leasing equipment does, but with more of the maintenance outsourced
Re: Why did this take them 10 years to figure out? (Score:2)
The cloud is considerably more than just VMs and organisations that only use AWS or Azure in the same way as they use a datacentre are obviously going to run up a huge bill. The costs of that aren't hidden. Anyone however that takes full advantage of the resources available is going to have a much better experience.
Re: (Score:2)
For many years experts said "don't worry about it" because costs were only getting lower, and presumably your company was only making more money as time went on. For unsure bets, yeah cloud it up to the 9's with an escape strategy of some sort, even if its: we absorb the cost of the cloud when until we figure out a plan to get out once we've reached this level of
Re: (Score:2)
Even just switching to a smaller cloud provider could save most shops 50+% of their hyperscaler bill*. Those who went all-in on the cloud-native thing might have more of a challenge on their hands, but in fairness, "cloud native" *should* be more cost effective than "just run some EC2 instances".
If you've got instances running 24x7, you can absolutely move that workload somewhere else (and leave any non-instance workload behind). On prem if you want to air condition some servers, colo if you really want to
Re: (Score:2)
Re: (Score:2)
Wait wait wait... ...doing basic due diligence cost analysis isn't something everyone does?
Re: (Score:2)
No company I've ever seen would run anything on a second hand server. The cost in maintaining that, in downtime, and in finding parts even needed, would be far higher than buying a new server.
Re: (Score:1)
Agree @1s44c.
A used server could work well for small businesses, non-profits or organizations do not need much IT support. The original poster could clarify.
Suppose you have 20k servers in two data centers. You need and have real-time disaster recovery. Those servers are mission critical. Would you not
Re: (Score:2)
I think many people don't punch in the numbers. They just look and see how much they're paying in hardware and employees (especially employees!) and figure that they can save money by outsourcing it. But you get what you pay for, and if you're cheap you're getting a bad product, and if you pay more you may as well just just pay your own workers to do it. The cloud providers suck you in too; it's easy and relatively cheap to get started. But once you start migrating in earnest so that it is difficult to
Re: (Score:2)
Hosting your own stuff requires an entire department of bearded grumpy people that speak a language that no one understands. If your C-Level is tech adverse, it's hard to swallow AND to manage correctly. And if you outsource everything, it's probably a bad idea. So there is still a place in the cloud for companies that don't want to touch tech with a 10 foot pole.
Re: (Score:2)
So there is still a place in the cloud for companies that don't want to touch tech with a 10 foot pole.
And that place is bankruptcy court.
Re: (Score:2)
If you make enough profits, not necessarily.
Re: (Score:3)
... If your C-Level is tech adverse, it's hard (for them) to swallow ...
Yes exactly, 100%. I've found that no matter how eager they seem, nor how sincere they present their questions. Never, ever, give them an objective honest answer when the honest answer is going to point out their repetitive failures to consider numerous pitfalls and/or their inability to apply simple logic.
Unfortunately, I'm still at the trial and error stage of navigating this super fun game. So I'm not exactly sure what the right thing to do is, but it's definitely not that.
Re: (Score:3)
We already knew the cloud was just an easy but expensive solution for startups or an expensive way to temporarily scale out one's infra.
There is another case, which is what's mentioned in the article. "companies with 'enormous fluctuations in load,'" and this kind of generalizes as "if you can avoid running things completely for large amounts of time". That's true of normal load scaling applications and it's especially true of AWS's Lambda system or other similar FaaS systems in the right case. If you have your own servers, you end up paying at least for the servers and the space that you need for peak capacity. If you switch them all off t
Re: (Score:2)
If your company sells, for example, tickets on the internet and a typical pattern is that you have to scale up a particular chosen application in response to a TV News segment that mentions something (think Taylor Swift's next concert) then running your application on the right cloud system can let you scale up to 100,000 times your typical application needs within minutes or even seconds, stay at that scale for 1/2 an hour, make massive massive sales and basically justify the entire existence of your company and then scale down again immediately the load goes away and perhaps never get to that scale ever again.
TICKETMASTER! They are the WORST!
They'll internally price-out the costs to scale; they'll only half-ass the scaling of some services (usually administrative side) without actually supporting the client-side breadth of services; they'll charge end-users exorbitant service fees; and they'll eventually wind-down those half-ass scaled services.
So much so that they'll literally never scale again for the duration of a performer's ticket selling efforts. And the best part? They'll also charge the performer fo
Re: (Score:2)
It's pretty tough to avoid once Ticketmaster also owned all the venues and require using their platform.
Re: (Score:2)
Yes, and famously they have had their systems blow up completely with big concerts. It looks like they roll their own kubernetes inside AWS in some ways, so probably getting the costs of the cloud together with the problems of self managed infrastructure.
Re: (Score:2)
If you abstract your software enough and you're in the right business, you can make your software multi-cloud capable and have a private cloud in a datacenter for your base load. And spin up AWS or whatever during peaks. For things like tickets, you have some obvious bottlenecks to make sure you don't oversell. But even internal to AWS you might have to deal with propagation delay, so it might not be that different.
Re: (Score:2)
I am a psychic (Score:4, Funny)
Progressive Web Apps (Score:5, Insightful)
Also known to those of my generation as "the mainframe model":
- You don't own what you pay for
- The software vendor changes whatever they want willy-nilly and if you don't like it, tough cookie
- When the software vendor fucks up and introduces bugs or vulnerabilities, you can't hold off the upgrade until it's fixed
- If you have a problem and you're a small company, customer service is nonexistent or insulting at best because you're not worth the software vendor's time
- The software vendor can charge you repeatedly whatever they want in perpetuity because you don't buy a license, you buy a subscription
When the personal computer came about, we couldn't get enough of it: finally computers would be yanked out of the greedy hands of IBM and the likes and we'd be free to do what we want on our own term. Finally!
And now we're right back where we started. The more things change, the more they stay the same...
Re: Progressive Web Apps (Score:3)
Re: (Score:2, Troll)
Trolls need to eat too!
Re: (Score:2)
And their favorite food is shit.
Re: (Score:2)
The other day, I suddenly realized how much they're even moving the web browser into the cloud (as much as they can)!
My work-from-home desktop is a beefy workstation. I run multiple browsers and browser profiles to keep some things separate (personal, research, play, work-normal, work-admin). In particular, the work-normal and work-admin browsers come into play - I have a separate work login to do Azure and AWS administrative tasks, so I keep that in a separate browser instance, which prevents MS from loggi
I was CTO of a company when it got sold. (Score:2)
The new board wanted everything in the cloud. We went from 10k a year to 800k. And still had the same staffing.
Re: (Score:2)
I see this happen a lot because people just move to the cloud. They don't consider how to use a cloud and think "It's just another datacenter". Designing your solution to be efficient in the cloud is way different than at DC. Most on prem systems will not be efficient when moved to the cloud.
As a cloud engineering consultant, I often advised against lifting and shifting. Those who did ended up upset at the cost.
Meanwhile, in my current role, I have systems processing TBs of data per month, costing around $3
Re: (Score:2)
This is exactly the problem. 'We moved to the cloud' == probably not taking advantage of cloud features aka lift-and-shift. Of course it is going to be more expensive as you haven't reduced your footprint but are paying the vendor premium. If you migrate into the cloud, making use of cloud-native features to shrink your environment to what you actually need right now, that is a different story.
The first part (L&S) happens. Then, since nobody wants to rebuild-what-already-exists-and-works, the rework t
I'd be the first one to tell you (Score:2)
When my latest client wanted the SAAS I was writing to be hosted on Amazon or Microsoft cloud, I wasn't too happy. It would have been cheaper for me to co-locate a couple of servers in data centers, but they were adamant about it. So I doubled my hosting cost and hosted a virtual instance on Amazon. Yes, it costs more than any server I would have put together, but ...
1) I made the client pay for it
2) I don't have to worry about hardware issues
3) I don't have to worry about unauthorized, physical access t
Re: (Score:3)
7) they don't need to worry about training their staff in your environment if they ever decide to take it over
8) they get a whole bunch of security stuff such as a repository for recording and validating software artefacts in to protect against supply chain attacks more or less for free
9) whatever security auditing work they have already done for their other applications also applies to your application..
And on and on. At the same time, doing your own hosting also has a bunch of benefits beyond cost
1) there
Cloud is for startups (Score:2)
Re: (Score:2)
I've worked for an undisclosed world-wide food processing machines company for years, and it seems a perfect example on how to make bad decisions in IT.
The good thing was connecting all the different brand companies with one international network. After that there was migrating to terminal servers for branch offfices (at my local company) which only had a meager few MB bandwidth ADSL connection, made worse by giving everyone who whined enough a laptop instead of a terminal client.
Then came the centralisatio
Re: (Score:2)
And now they're busy moving all ERP to one instance of SAP (aka Sandlaufer Anschau Programm), complete with missing functionalities, more slowdown, and even higher costs.
Actually, it's Sanduhr Anzeige Programm (Sandglass Display Program).
Re: (Score:2)
Ahh, thanks.
I heard the term from a co-worker who had frequent contact with the German IT side. Seems his language dictionary wasn't that good after all (I did see the term I used on a couple of Dutch pages).
Re: (Score:2)
The real news is... (Score:2)
... that AWS is even more expensive than Pure Storage? Wow!
Clexit? (Score:1, Offtopic)
remove the "ex" and shorten it to four letters.
Company plays lift-and-shift (Score:3)
Surprised to discover that it's more expensive.
This is pretty much EVERY failed cloud journey in a nutshell and provides significant evidence 37Signals shouldn't be viewed as techxperts on cloud. Ever.
I am permanently surprised that Azure remains cheaper than on-prem despite all the accounts of how expensive cloud is. Then again, proper prep and planning meant we didn't just build big, stupidly expensive, VMs in the cloud. SaaS offers significant value. Storage is a damned site cheaper and requires zero maintenance . Visibility and service offering to business is signficantly improved with no need to force business through gateways or, even worse, VPNs.
That being said, pretty sure one of the internal AWS teams announced a year or so ago that they dropped AWS micro-services because of the cost. They were not wrong. The value propisition there is negative all the way.
Re: (Score:2)
> I am permanently surprised that Azure remains cheaper than on-prem despite all the accounts of how expensive cloud is.
Are you talking about Azure VMs? They are not cheaper than on-prem even for very reasonably sized systems. They are quite a bit more expensive. I know, I used to run our servers on Azure.
My company was renting our servers from an ISP in Arizona. It was stupid expensive. We move to Azure and our costs dropped to 1/3rd. We did it again and moved to OVH and our cost went down to 1/3r
What was the point of moving to the Cloud again (Score:3)
Re: (Score:2)
Re:What was the point of moving to the Cloud again (Score:5, Informative)
What was the point of moving to the Cloud again
Versatility. Deployment. Scalability. Security. Visibility. Access. Manageability. Automation. Redundancy. Disaster Recovery.
The list of benefits is long.
Not having to manage hardware rotation, connection redundancy, maintenance, energy and staffing is a pretty big selling point as well.
But then, you already knew that. Right?
Re: (Score:2)
Re: (Score:2)
I keep hearing people say that moving to the cloud will be cheaper for them
That would be an error.
The point isn't to be cheaper. The point is to achieve more at the same cost. Cloud does that, really well. If "cloud has failed" to do that, whoever was involved should be fired.
Re: (Score:2)
Which means cheaper, which the Cloud proponents promised and never delivered on.
Re: (Score:2)
Which means cheaper, which the Cloud proponents promised and never delivered on.
Anecdotal failures do not the truth make.
The cloud is the proverbial basket (Score:2)
Re: (Score:2)
If you only have one customer or supplier, they own you.
Nobody's putting their infrastructure into a vague 'cloud' that pits different providers against each other - they get locked in to a solution and then they're vulnerable to being squeezed. And they've just handed over their data to that provider and whatever back doors they provide to government agencies (both wittingly and un).
If you are big enough to be able to afford it, you should have your own servers and the only 'cloud' should be an offsite ba
People who maintain on-premise hardware. (Score:2)
I hope they put that money to good use (Score:2)
That $10million in savings can get expensive if you don't pay for a competent and well staffed IT department to keep your new servers running and secure.
There's more to the cloud than hardware costs. The cloud still makes sense for idiots who don't fund their IT departments.
Cloud isn't the problem. (Score:5, Insightful)
The cloud delivers something no managed onprem datacenter could, ephemeral environments. That's the true power of the cloud and its abstractions. But what ends up happening is developers use the cloud like they would use an onprem datacenter. They treat their infra as pets. Maybe they use IaC, but still will spin up infrastructure manually in the console and track it with an excel spreadsheet. And if they do use IaC it's old, not updated and they will complain their state is corrupted when they do aforementioned manual things in the console.
No the problem is like most things, if not all. The user of the tool and not the tool itself.
Re: (Score:1)
The pet VS cattle thing is the single most important concept in moving from on-prem to the cloud. Real shift of paradigm. So weird to do it the other way around!
Re: (Score:2)
The cloud delivers something no managed onprem datacenter could, ephemeral environments.
Isn't that what Open Stack is meant to do on private piles of servers? What do you even call a private cloud? A steam room?
Leasing stuff costs more than owning it (Score:3)
Leasing stuff costs more than owning it: film at 11.
0.0001% of revenu ? (Score:2)
A 100billion dollar company (https://medium.com/the-mission/how-basecamp-built-a-100-billion-business-by-doing-less-on-purpose-5f978ce6478c) saved 2m/year by moving of the cloud....Big story.....
Re: (Score:2)
Saving pennies scales to dollars. It all matters, and as long as you're including all your costs (including the cost of evaluating the potential savings), then it absolutely makes sense to pursue these things.
Re: (Score:2)
While I agree with you in principle, I take issue with how these stories are reported. They rarely take into account the size of the company, its maturity, etc. There are size where this move makes senses, some not. It is also pretty hard to quantify the lost/gain of agility with the move, so these (already low) $ values are not very informative imho.
Tick tock, tick tock (Score:3)
Outsource to save money
CTO leaves, new CTO joins
Insource to save money
CTO leaves, new CTO joins
Move to cloud to save money
CTO leaves, new CTO joins
Move back on prem to save money ... and on it goes
Re: (Score:3)
The cloud is a trap (Score:3)
Run Away!
Cloud only makes sense in handful of cases (Score:3)
There are only a few reasons that makes cloud deployment valid. Though these reasons cover lots of use cases. The main 3 reasons I can think of are:
1/ You are too small. If you are only going to need a handful of servers, it probably does not make sense to keep the expertise to run your own machines. You probably want 99% uptime. You need power contingencies, you need network contingency, you need people to swap hard drives when they go bad. That's a lot of expertise to keep in-house if you only need 10 machines. Cloud will probably be cheaper.
2/ your peak load is MUCH higher than your typical load. Especially if you can't afford slower service during peak time. If you are going to need 50x the computing resources you typically need but only during a one month window, then it probably does not make sense to run and operate a system for the 50x peak internally and let the machines idle the rest of the year.
3/ You need low latency to your end customer which are world wide distributed. Geographically distributed cloud does give you decently low latency to about anywhere in the world, provided you have a close by machine. If probably does not make sense to run your own data center in every country in the world. Cloud regions can give you high geographical availability at a reasonnably low cost.
Seems like basecamp is exiting 1/. Because I don't think 2/ and 3/ ever applied to them.
And that does make sense, I almost didn't see anyone use basecamp 10 years ago. I see a lot more users today. So maybe it is the right point to switch for them.
Quelle surprise (Score:3)
Another win for EU regulations (Score:2)
"AWS made data transfer out of AWS free for customers who were moving off their servers in March, spurred in part by European regulations."
The EU has been on the forefront of consumer/public oriented rights and protections. I hope one day, we can learn from there successes in the U.S.
We have some of the most extreme laws and punishments against our own citizens (see our ridiculous incarceration rates in the world [wikipedia.org]). So, I don't understand why people object to laws that protect citizens from corporations and
37 Signals use ruby (Score:2)
The last I heard 37 Signals use ruby for everything. Ruby is a crazy CPU hog. If you watch a ruby server and a java server with the same load you are going to get 20 times higher CPU load on the ruby process. So if you need 10 servers for a java server you need 200 servers for ruby.