14.3 million pageviews per day comes to 163 per second, on average. Reddit's traffic, like most sites of its kind, probably peaks during lunch hour during the workday and probably peaks at 5x that number. So we're looking at 815 pageviews per second. Given 80 servers total, we're looking at 10 pageviews per second per server.
Let's compare this with Facebook. In October 2009, Facebook announced they had 30k servers[0]. In September 2009, there was a rough estimate that Facebook served 200 billion pageviews per month[1]. That implies 73k pageviews per second, or 2 pageviews per second per server. Clearly the pageviews are a rough estimate, but even if facebook served 1 trillion pageviews per month they still wouldn't be beating reddit for efficiency.
I have a feeling if you run the numbers for any other highly dynamic site at scale, you'll find that amortized over every server in use, you won't get a lot better than 10 pageviews per second.
Before I found this overview and your comments here today, I started two other similar overviews. One on Quora and the other on Bestvendor. Here they are:
I think you've mischaracterised things a bit in your post - not sure if this is just because you're repeating what you read elsewhere? Anyhow, regarding the bullet points:
1. Google Wave was released an experimental consumer product, designed to combine the best of IM and email. According to the announcement, it was finally sunset due to lack of interest from the public. Personally, I was super excited about Wave - I even pestered an ex-Googler at the time to get me in on the preview. (I was in uni at the time). However, I had a lot of trouble getting other people to use it.
Did you personally use Wave yourself? Were you able to convert anybody else to using it?
Firstly - this is hardly apples to oranges. You're comparing a experimental consumer-oriented preview product, to an enterprise offering with SLAs. It would be like if we released a Star Wars game showcasing awesome new WebGL features - then decided to sunset it 12 months later.
Secondly - the project was open-sourced and then moved to under the Apache banner - an effort that in itself took significant engineering effort.
In the case of Reader - this is somewhat personal, as I was a big lover of Google Reader as well. From memory, it was deprecated with around 12 months of notice, there were provisions for migrating/exporting all your data etc. It wasn't open-sourced like Wave, but there were reasons behind that (I can't comment). However, it was hardly the worst handled deprecation - and to be honest, the writing was on the wall for RSS for a long time before that.
2. You didn't provide any specifics, so I'm not sure what event you're referring to.
AFAIK, the only major price change was when Google AppEngine left beta/preview, and became an actual commercial product, and Google added SLAs, and other things etc:
The changes were announced at I/O that year (May), I believe, and the actual changes took effect in September.
Is that what you're referring to? Because even taken aside my Google hat, that seems like a fairly reasonable thing to do.
3. All of Google's enterprise offerings have support. I know, because I actually work in such a support team. For GSuite, support comes free as part of the product - you can contact us via email/chat or over the phone.
> They've done it again and again. I'd say its the default way Google operates.
This is a pretty bold claim - can you cite any examples of this?
If this is something that personally affected you - please feel free to reach out to me. I can't promise I can fix everything, but I can definitely try to find out more for you, or see what I can do internally.
Disclaimer: I work for Google, but views expressed above are my own.
This will sound crazy, but I recently deployed a content management workflow for a rails app that mounts a WebDAV drive inside of a staging rails application that's running http://sitepress.cc/ (think Middleman in Rails, without all the dependencies)
If the marketing team wants to edit something, they mount the content drive, make changes using their text editor of choice, and are able to preview when they hit Save.
After they make a bunch of changes and add whatever media an engineer copies the changes off of the drive into a commit and deploys to production.
So far it's been working pretty well. The marketing and CS team are all able to edit markdown copy and run it easier through their content editing processes since it's just files. They don't need to learn or care about git. Designers can throw images in the volume without thinking too much about it.
Engineering is happy because they control their git history and commits. When it's time to optimize PNGs, compress assets, etc. the engineer can deal with and deploy to product with confidence.
We also run rspecs against every single page of content to enforce SEO, branding, etc.
Happy to answer questions or write something more detailed about it if others are interested.
We are working on an editing interface for static websites that can compete with the ease of use and flexibility of wordpress. Our alpha users love it and this is an open source project.
We've been working in this space for quite a while, and are extremely excited about the potential for static web-tech.
We started out with simple drag and drop deploys at BitBalloon (you don't even need to zip your site first, just drag a folder unto BitBalloon.com).
Later we launched our premium solution called netlify that's by far the most feature rich on the market (continuous deployment, form processing, API proxying, redirects, rewrite rules, SSL, etc, etc) and offers the best performance you'll get today.
It's spot on that the really big deal will be solving the CMS need, without making developers give up on all the advantages of static site generators.
Some of our larger clients are using netlify + roots/middleman/metalsmith/etc + Contentful/Prismic/etc to build large CMS drives sites that are built up front and hosted directly on our CDN, but this is still a bit too advances a setup for the millions and millions of normal CMS driven sites out there.
Our solution is building a completely open-source CMS that works with all common static site generators. It's in private beta right now, but getting real close to opening up the repo to the public. It’s completely free of any lock-in, and you can take it and host it anywhere with ease.
Happy to send an invite to anyone here. Just ping me at [email protected]
Some feedback to the product of this thread:
Love the initiative.
Drag and drop uploads are nice, but unless you handle CDN configuration and cache invalidation, it doesn't seem like a big step up from just FTPing files to an S3 bucket.
Otherwise you might risk mainly appealing to beginners or people looking for a cheap way to publish a personal website, and those can't pay much. With a monthly price of $1/site, the life-time value of a client will be very low, so you will barely be able to spend anything on customer acquisition. One.com, GoDaddy or S3 can offer extremely low prices because they have huge scale, but as a tiny startup, you have no way of reaching the amount of users you would need to get on board with a model like that. Especially since this is way more technical than Wix/Weebly/SquareSpace,etc, while not really offering real value to professional developers or agencies.
If you want to build in the static hosting space, you should start by figuring out if there's something you could offer beyond what the existing players like netlify, BitBalloon, CloudCannon, Divshot, etc, already have, and make sure you're not just trying to compete on price.
Eat less calories than you use each day. (fullstop!)
While true, that statement is useless. It doesn't take into account the effect insulin has on the ability one has to actually take in those fewer calories. It also doesn't take into account metabolic pathways that allow one to eat more of certain kinds of foods than others.
The scientific literature is pretty clear (once you filter out the government propaganda that caused 300 million people to be unwitting participants in a dietetic experiment):
If you[1] want to loose weight and be healthier, screw calories and worry about insulin (full stop!)
There is nothing that more closely predicts insulin reaction than the number of carbs you are eating. Cut out all the simple carbs (which includes the bags of sugar we've cultivated as most fruit) and focus on fats, protein and very complex carbs and you will lose weight. More importantly, nothing has been shown to affect metabolic disorder[2] as positively.
I look at it like this: what do we use to fatten cattle up before we slaughter them? It's corn. Now look at every nutrition label to see what is in it. Is it surprising we are fat, too?
1. The body is incredibly complex. There are people who will respond better on other diets. So take any general study with a grain of salt and experiment on yourself to find what works!
2. High blood pressure, high cholesterol, heat disease, hypoglycemia, type 2 diabetes, gout, and others have all been lumped together into a disorder that strongly appears to be the result of hyperinsulenimia (eg too much insulin in the blood stream).
It depends on the value you get. Some SaaS are free or cheap, they probably cost little to the company, or the company is running cheap (single founder in ramen).
Now for how much a single person spends, it's a different story. Let's assume your target audience is software developers. An average software developer should net more than $5,000 a month. Spending $150, is like 3% of his salary. Not a lot, so he could even subscribe to more services and more expensive services.
If your target audience is Egypt for example, where the developer net $200-$250/month, subscribing to only one service may cost 10% of his salary. So it's a luxury he can't afford.
Bottom of the line, your target audience and market readiness is the one that decide the price tag for your SaaS.
I'm quite likely in the minority, but I strongly gravitate to services that 1) use a cost-based micro-pricing scheme; and 2) that I trust to not mark-up unreasonably much. Generally I want to offload the "how much should this kind of thing cost?" calculation to someone else who I can trust to calculate it accurately, add on a reasonable profit margin, and charge me that amount. I definitely avoid companies that I feel are instead trying to figure out how much they can get me to pay, and then pricing it at that.
One example, which I use for hosting, is nearlyfreespeech.net. Amazon's and Google's cloud services have a similar utility-pricing model, though I'm not sure how much I personally trust them to price in line with costs, especially whether they're going to pass on future cost decreases. Prgmr.com is a flat fee rather than a combination of micro-prices, but it has a little of the same vibe of, "I calculated how much it should cost to operate these servers, divided by the number of VPSs, added on a reasonable profit margin, and that's what I'm charging", which is one reason I went with them.
It's not even necessarily the overall price that worries me so much as liking the general warm-and-fuzzy feeling that I'm not getting a hugely marked up service. Same as when shopping for, say, a mechanic: I feel most comfortable with a mechanic I can trust to charge me what the repair costs, plus a reasonable profit, not one who's trying to size me up to figure out how much the repair is "worth to me".
Kiln from Fogcreek. Slow on a good day, it just mysteriously stopped noticing my commits in my activity stream some four years ago, you can't have reviews spanning multiple repos, they've blown off pull-requests as a feature for years.
As I said down thread, we (Google) and some of our customers are routinely the target of attacks (DDoS and otherwise). We don't talk about it (sorry? yet?), but we have mechanisms to protect you as well as ourselves and other customers. My hope is we'll be talking about a particularly high profile one soon, but if you've got an account and want to talk to us privately, we're here to help.
Disclosure: I work on Cloud, but I'm not a networking expert.
I completely agree. For years now I've been working on using the mouse less and less and the keyboard more and more up to the point now that I hardly ever use the mouse (a couple of months ago I sat down and used a public terminal for 25 minutes before I discovered that somebody had stolen the mouse).
Up to now I've been using conkeror for my browsing need (and w3 in the terminal when following links in email from wanderlust in emacs) since emacs is my OS of choice. However I'm now in the process of converting to chrome. I found out that with the help of some extensions and a little bit of scripting I will be able to have a chrome almost as keyboard driven as conkeror and finally be able to use some extensions I have sorely missed in conkeror (biowalled2browser, readability, etc) and that most google products work fine without having to write modes for them (nice since many of googles products are rather keyboard driven by nature)
Why do I do this? For one I hate wasting time, and I hate wasting screen real estate. My computer should be a tool that helps me get tings done as fast as possible, not something beautiful to look at. The more use of GUI and/or mouse the slower one works. If you can to everything from the keyboard, and _know how_ to do it from the keyboard, you can work much faster than when jumping back and forth between the keyboard and mouse or other pointing device.
But I take it one step further. Being all about efficiency I don't want to have to move my hands much either. I therefor strive for using and/or creating setups where ones hands never have to leave the asdf-jkl; position of the keyboard. Thus I don't use enter (C-m or C-j), I don't use arrow keys, and avoid at all cost using any keys that are not reachable without moving my hands from the home location on the keyboard.
To sum up: Basically I'm a keyboard and efficiency geek ;)
A few bits I've noticed over 25+ years in the industry:
• Tell me what your product is. What it does, where it works, how it does it, what it requires. Is it a physical product (or is it shipped in one), an interactive application, a Web service, a programming language / tool?
• Tell me what the fuck it is EVERY GODDAMNED TIME YOU COMMUNICATE ABOUT THE PRODUCT. It doesn't have to be long or detailed, you can link to your detailed description in the communication. But your press releases, emails, Tweets, blog posts, marketing collateral, etc., are going to get passed around, word-of-mouthed, and/or pulled out of drawers (or browser history / searches) for weeks, months, and years to come. Make them work for you.
• The Economist's practice of briefly introducing any individual, no matter how famous or obscure, is a wonderful practice of microcontent contextualization. "Using the Economist house style offers an elegant alternative, wherein virtually all people and organizations are identified explicitly, no matter how prominent. For example, you might see 'Google, a search giant', 'GE, an American conglomerate', or 'Tim Cook, boss of Apple'." http://redd.it/1x8yky
• Tell me how to try it out. Preferably for 60-90 days (a 30 day cycle can go far too fast. I've been very, very impressed with New Relic's "use it for free, convert upmarket for additional features" model, and it's apparently worked well for them. For small accounts, their cost of sales is effectively nil (and for large accounts, COS is always a PITA). But for those large accounts, you've got a proven track record with the prospect, and they really know what they're getting.
• Put your tech docs front and center. As a technical lead / director, my questions are "how the fuck do I make this thing work", and if you can't tell me
• It's been observed many times that those who have the best appreciation for how a product works are those who use it directly, and secondly, those who either service it or support those using it. John Sealy Brown's The Social Life of Information addresses this with both Xerox copier repairmen and support staff. Use this to your advantage two ways: let these people share and collaborate, even if informally For the repairmen, this was a morning coffee break turned out to be a hugely valuable cross-training and troubleshooting feature. For phone support, after an "expert system" and changes in technology separating phone reps from technicians, researchers noted two reps who consistently provided good advice: one was a veteran from the earlier stage, the other a recent hire who sat across from the other and learned from her. Similarly, user support groups (mailing lists, Web forums, Usenet groups), in which users interact and share knowledge with one another directly (Hacker News would be an instance) are often (though not always) far more useful than direct tech support.
• Provide clear pricing information. This has been noted from Jacob Nielsen on forward as the information people are most interested in.
• Make damned sure that whatever process or workflow you've created online works, and for as many possible end-user environments as possible. Keeping interfaces as simple and legible as possible is a huge bonus.
• Remove distractions from your transactional webpages. Once someone's homed in on a product, focus on that, though you may mention alternatives or (truly useful) related products. Every additional piece of information on the screen is an opportunity to confuse and lose the sale. I've been restyling many websites simply for my own use (1000+), and simply removing distracting elements produces a far more productive environment.
• Ensure your pages are legible. Backgrounds should be light, foregrounds light (and where, with extreme reluctance, you invert these, separation should be clear). DO NOT SCALE FONTS IN PX. On far, far too many devices this renders as unreadable, particularly from older (e.g., more senior w/in the organization) readers. Grey-on-grey is just cause to fire whomever suggested or required it. See ContrastRebellion: http://www.contrastrebellion.com/
• Don't organize your website according to internal corporate structures. Your website is an outward facing tool, and should address the needs of users, not of internal departments. Lenovo's laptop site organization would be highly typical of this: I want a Linux-capable, large-display, full-keyboard, trackpoint device. The rest I generally don't give a shit about, and its product line confuses me every fucking goddamned time I try to buy something there (usually every 2-3 years). I'm not a sufficiently frequent customer that I keep up with every last change, but I've spent thousands of dollars on IBM/Lenovo products, as an individual (hundreds of thousands to millions as an enterprise customer).
And of course: test all of this, don't simply take my word for it. But yes, I've walked from far, far, far too many product pages, from free software projects to Fortune 10 companies to edgy app devs.
Dropbox is targeting a B2C market and started with poor twenty-somethings.
Github, and virtually every other thing that costs more than $20 a month, targets primarily a B2B market. It might be popular with some local poor 20-somethings, but honestly, you're just an infection vector to get your day job on board.
The pricing is designed to extract maximum value out of business customers. If they have 125 simultaneous projects, they officially have More Money Than God. "The price of a residential Internet connection" is not a pricing anchor to them. (Should they need one, they're probably going to be persuaded by "We have 500 man-years of labor in our projects, one man-month costs us $15k, lemme break out Excel for a minute, oh it seems all my options cost pigeon poop.")
I strongly, strongly encourage you to listen to the Mixergy video about Wufoo or talk to anyone who runs a SaaS business if you do not understand where most of the money is likely getting made. That topmost plan which costs $$$$$ prints money, primarily from people who don't need all that it offers and couldn't care less because it costs less than pigeon poop on their scales.
If you don't use Github for your projects because $100 is a lot of money for you that's perfectly fine for Github because it does not make them meaningfully worse off.
"Remembering that I'll be dead soon is the most important tool I've ever encountered to help me make the big choices in life. Because almost everything — all external expectations, all pride, all fear of embarrassment or failure - these things just fall away in the face of death, leaving only what is truly important. Remembering that you are going to die is the best way I know to avoid the trap of thinking you have something to lose. You are already naked. There is no reason not to follow your heart."
A lot of people think following all best practices and offering a unit test sacrifice to Uncle Bob will lead to success
Shipping will put food on the table. Of course, if your system can't stand 5 simultaneous users it won't, still it's difficult to get that bad. (but yeah, some people manage to get to that level)
And some "online services" have very bad code, and still they sell millions. Get rich first, then you can improve your code.
Of the three startups I've worked with, two of the three were ridiculously over-engineered monstrosities that were way over time budget. It was clear that the CIO/CTO wanted to do cool fun stuff and not build a marketable product.
The other was cobbled together with completely shit code, was constantly breaking on releases, and was glued together with perl scripts. They're now publicly traded.
I work on PHP at my day job (in a public company), before this, I came from Ruby, and .NET before that.
I'm convinced the reason so many successful projects use PHP, is not because of any inherent nature of the language. I think it's the people who use it. They just don't care. A successful project needs to be started by someone that cares just enough, but not too much.
If you're programming in PHP, you're not running around talking about "convention over configuration" giving talks, or trying to make your code beautiful. It's a garbage language, and you know it. But it allows you to get something up and running, so dang quick. You're failing while the other guy is still updating his gem file. You're advertising while the other guy is trying out some fancy new deploy script. You're incorporating feedback while the other guy is just starting to get to work. When he finally fails, he's used up half his runway, whereas you, the guy who didn't give a fuck about your code has gotten past that first failure, and are finally getting some traction.
Hopefully, the next guy to join the company will clean up your shit. The other guys code may not look like shit, but it doesn't solve any useful problems... so they never got the chance to hire that next guy.
I do use OVH and recommend them from time to time.
The main issue is that I lost a bit of faith in their support and reliability. vracks going down for hours with no updates. Connectivity issues. Servers disappearing.
Besides that, their DDoS protection works well for l3 attacks, except that they force a TCP reset on every connection. So if you are picky about extra connect times and having your clients re-establish their connections, they are great.
Whether or not I decide to use this service, this is one of the most useful announcement blog posts I've read in a while. The tone is not just "ok we released this" but instead "we released this, and here are some practical use cases if you are already doing X, Y, or Z" -- nice job.
Here's an option you might not have considered, but might be applicable, depending on the read/write speeds you need from those hundreds of gigabytes of data. Amazon has cloud drive, which costs something like 50 bucks a year, and has unlimited storage.
The neat thing is someone created a FUSE file system called acd_cli that allows you to use the cloud drive as if it were a normal hard drive. The speeds aren't too shabby either; I easily get 150MB/s up and down on my dedicated server, and response times are snappy. Additionally, you can create a unionfs mount so that writes are instant, and you sync new files on a regular schedule.
Of course, that might not be applicable to your use case, but I use mine for many things. I have a plex server running using that with over 13TB of videos, and it works flawlessly. It allows me to run a full plex server with unlimited storage for 20 bucks a month.
I have a Dell Chromebook 13. Basically it is a Macbook Pro that happens to run Linux really well. I have used Thinkpads X (200, 201, 220) in the past, then switched to Chromebook C720, then this.
After you install John Lewis' ROM (basically a Coreboot build), it is an absolute dream, and everything worked out of the box, wifi, bt, 3d acceleration, suspend/resume, audio jack detection, you name it. Trackpad is really smooth, 1080p resolution IPS, the battery lasts 10 hours, never gets hot, fancy back-lit keyboard, has 8 philips screws in total to take it apart. Hell, it even has a TPM module so I can do those fancy ssh password-less logins (https://github.com/ThomasHabets/simple-tpm-pk11). If you really like the clit mouse, then this machine wouldn't cut it though. The speaker is kinda iffy and doesnt sound very nice, but nowhere bad.
I upvoted you for the link to the old thread, but why is it self-evidently crazy? Just because most humans do something by default doesn't mean it's an optimal solution. Many things that people do by default are a result of environmental training but are not necessarily optimal behavior (here in China people juggle walnuts between their fingers in an effort to ward of arthritis, but as far as I can tell that behavior is just as likely to encourage arthritis as protect against it).
I also attended Dr. Dement's Sleep and Dreams class while I was at Stanford. His class was basically a propaganda program for students to get more sleep: He even gave us extra credit for sleeping during class. The man is a true believer in getting tons of sleep and that does indeed say a lot towards discrediting people who vouch for polyphasic sleep schedules. I tend to trust what he says, even if he was rather old and eccentric.
On the other hand, what's going on with all of these people who write such glowing reviews of the process? Are they just delusional?
Let's compare this with Facebook. In October 2009, Facebook announced they had 30k servers[0]. In September 2009, there was a rough estimate that Facebook served 200 billion pageviews per month[1]. That implies 73k pageviews per second, or 2 pageviews per second per server. Clearly the pageviews are a rough estimate, but even if facebook served 1 trillion pageviews per month they still wouldn't be beating reddit for efficiency.
I have a feeling if you run the numbers for any other highly dynamic site at scale, you'll find that amortized over every server in use, you won't get a lot better than 10 pageviews per second.
[0] http://www.datacenterknowledge.com/archives/2009/10/13/faceb... [1] http://www.businessinsider.com/googles-estimates-for-faceboo...