Server costs for video startup
I've been going over some of the posts here regarding server costs but couldn't really find any information specific to my needs. I'm planning on creating a mobile app that deals with users uploading videos. I realize this will take more server resources than a typical application that deals with just textual content. I would like to know if anyone has an idea of what kinda of server I'll need to start with and how much should I expect to spend. I do not have any external funding, so it's all coming out of my pocket. So any useful information to keep costs low would be greatly appreciated.
Thanks!
-
How is this different than Google Plus, Youtube, or any number of other services out there? I was going to suggest using their API's to build your product on top of their platforms as that would lower your barrier to entry, but it's not clear what your exact business goals are. – James Mortensen 12 years ago
-
It's some what of an exclusive video sharing application, similar to Viddy. My initial plan was to use Facebook, but their API doesn't support uploading videos to an application by users. – Zipper 12 years ago
-
Ok, what about cloud storage? Cloud storage can be quite effective, and you don't have to have space for servers or hire server admins? My suggestion is to maybe look at those options, I think you'll find that these costs are going to pale in comparison to others: (i.e people costs, marketing, etc). Amazon S3 and Google Storage are just two that come to mind. – James Mortensen 12 years ago
-
See my answer. Storage is only half of the picture here, you need a server just to process that data and you also need a database server that stores all meta info about the videos, users and other content, then you need a server for the static content. – Digital Sea 12 years ago
-
You should check out Julbul.com – Rabia Tariq 5 years ago
7 Answers
The costs depend on your traffic and code-base, based on my experience working with large-scale video websites you need quite beefy servers just to process the video and convert it into a more friendly format, then you need the bandwidth to play the video back to the end-user. I would advise against starting an online video site unless you have large amounts of cash to spend. You could use Amazon EC2 for both your converting and playback, you get charged for the bandwidth and processing power you use, at the start you'll probably want a server instance for processing video, a server instance for serving your pages running a reverse proxy like Nginx and a server for your database.
There is a lot to consider when starting up a video site, like I said I would go for Vimeo or Youtube and use their API's, if they don't meet your needs then you're going to need a bit of cash. At the beginning you'll probably be up for a few hundred per month in server costs, as your traffic grows so will your costs.
Amazon EC2 is your best bet for something like this and the number one first choice for most internet startups that want to be able to scale.
-
One advantage to the cloud is that it can be your testbed during development. It will let you test different sizes of machines. When you aren't testing you can pause them, and their cost drops to almost nothing. It would be impossible to size a system without knowing the base of users you will have, and the complexity of the processing. Cloud based lets you start small with a minimal commitment of money. – Mhoran Psprep 12 years ago
-
THat is totally different from the question, though ;) TESTING is one of those "not ins use all the time" scenarios and you can use the cloud AND Have your own machines ;) – Net Tecture 12 years ago
-
Tell Instagram, Pinterest and a lot of other well-known startups using Amazon EC2 that cloud hosting is unaffordable. The complexities in buying your own machine, backing your data up, sharding your databases onto multiple servers is eased via cloud hosting. Cloud hosting virtually pays for itself once you encounter your first scaling configuration issues. It's easy to bump up your RAM than it is to open up a machine and replace the RAM in a data centre. – Digital Sea 12 years ago
-
I would go Dwayne's route. Although NetTecture's option may be more cost effective it's probably not the best choice in your situation. Are you a server admin, do you want to be responsible for the hardware and backups etc.? Do you want to drop all that money up front on server costs with your idea un-proven? Although not convenient you can always move off of Amazon and do something more permanent later. But, Amazon is easy to get started with and until your concept is successful it just isn't that critical. – Ryan Doom 12 years ago
-
Seriously? Have a look at the prices and then you realize that you can afford 3-4 times the hardware AND the admins from the difference. I have 6 servers now, nr 7 going nline in 3 weeks and I pay a LOT less than you would for ONE of the servers in the cloud. Every time I see the prices I - sorry - vomit. I can get a 6 core / 16gb machine for 160 USD. I just ordered a 4 core (intel, so + hyper) 16gb machine with a 120gb ssd (not a lot of data there) and 10tb traffic and I pay 150 USD or so per month, with local people handling the hardware. – Net Tecture 12 years ago
-
THe counteroffer from a cloud provider was around 1 core and 4gb memory.... OUCH. Plus when you get bigger (video does) backupts etc. are similar to cloud from your work anyyway. – Net Tecture 12 years ago
Mor eprecise to Dwayne - you need a lot of CPU and a lot p f processing power.
Unless you fail, amazon is stupid - clouds are extremely cost negative when you run a lot for a long time. FOr exmample, I am odffereed a 1 rack unit hosting for about 100 usad including 10.000 gb traffic in a large chicago data center. Beat that with Amazon ;)
What I would go with is:
* A SuperMicro 4 rack unit case that holds 72 discs possibly on a dual socket motherboard and a GOOD Raid card. The case already will put you back around 2500 USD - all in all without discs you talk of likely 5000 USD.
* A SUperMicro Twin2 case. That is a u2 rack with 4 "slots" that take dual socket mobos each, allows yout to put 8 CPU sockets into 2U.
If you scale that up, I would skip the Twin and go with a Del M1000e blade server case - there are new super small dual socket blades coming out where you fit 32 of them into one center (64 sockets total) and you can get a blade center for a decent price refurbisehd (the center itself does very litle that can break).
Finding hosting is another thing - you need a LOT of power, most data centers can not handle that outside of multiople tacks (kudos for CME Group new Aurora that has 12.8kw per rack distribution IIRC - but then, this is not normal hosting). I Know of a couple of US ddata centers where modern computers DO NOT RUN (ralk of 60 watt powre allowanc eper rack unit - a joke, most decent CPU have more power requirements).
There is always the tera box concept. someone was putting a LOT of vertically mounted discs into a special rack case.
A lot depends on your growth patterns and estiamted space. THis is the wrong place to ask - serverfault.com has mre info, but your local IT guy should know. THis is why senior popeople are senios.
But beware of clouds, their cost side is RIICULOUS expnsive for long term use. I am a big fan of private clouds (i.e. you OWN the platform) running myself on the 8th (next server delivered) 144gb with 6 sockets ;) The cost side I have is a nice joke compared what it would take Amazon or AZure every month to pay for that.
-
The advantage of using cloud hosts if they do a lot of the heavy lifting for you. They take care of scaling for you (mostly). Owning your own servers is one thing, but cloud providers are great if you don't know what kind of traffic to expect. What's to say you build an amazing machine then find that the few thousands of dollars you spent wasn't enough? Most people don't have another few thousand to offload onto another server, then there's the issue of scaling your codebase and database across the multiple servers, not an easy feat for the non-technical minded. – Digital Sea 12 years ago
-
Well, the advantage of renting a car is that they do a lot of heavy lifting for you. Pay – Net Tecture 12 years ago
Your highest cost by far is going to be bandwidth.
Look for an inexepnsive colo.
I can recommend Calpop (los angeles)
Hurriance electric (by far the cheapest, in Bay Area)
Cloud offerings will cost you TOOO MUCH..
If you are looking to have a cloud, build your own using xen or hyper-v or vmware
Depending on your expected volumes, you might be just as well to consider a dedicated hosting solution such as Viddler.com. Whilst this would become more expensive as you grow, it might well be a short-to-medium-term option that would enable you to prove your concept to determine if it will fly or not.
On the other hand, if you are charging your customers to use your service, you might be able to determine an outline cost-per-xxx for a typical video. I do this on my site, and generally factor in the cost of (third-party) video storage and streaming into the price. Works well enough for me.
The cost will really depend on what you want to do, converting will cost more also different formats will be larger/smaller and will cost a different amount to store.
I suggest you run through a basic cost estimate yourself. Run through a typical use case and then estimate what you need at each stage. You then can multiply that by your estimated user base. You can then use Amazon as an estimate of what you need to fulfil those needs.
Ah - cloud vs dedicated. The debate continues. ;)
Agree this is more of a serverfault question. But, since I don't really know what the primary business model / MVP is, I can't quite determine whether the encoding / transcoding is best outsourced or not. Or what bandwidth is required. Or what encoder makes sense.
"Users uploading video" isn't extremely helpful. Are we talking about small 10 second clips or 2/5/10/15 minute clips? Are you doing any special encoding at the capture device? Once you have the video, are you multicasting it? public or private sharing?
All these things allow someone to figure out what architecture makes sense - otherwise we all are describing the elephant from our own limited vantage point.
To reduce some costs with server storage and data stream, have you considered any compression module in your mobile app? This will significantly reduce the file size before upload and also allow for a quicker upload time.
The issue now with the later model tablets and smartphones, is the size of video they create. You may want to consider some form of compression in your app to alleviate some of the costs at server side and data costs incurred by the user (if not using wifi) + nicer user experience with decreased upload time.
