Topic: Closed - page 2. (Read 8662 times)

Nah we should blame gweedo instead. I wasn't clear enough in my original post since I do think that ASP.NET is OK for scaling technically, but I just wanted to point out some gotchas and hidden costs on the business side. Granted those are probably a bit out of the range of a Bitcoin business for a while, but I can always hope for crazy volume...

Also, RAM is incredibly cheap now! I'm sure MS will change their licensing tune if they can foresee a hit to the Server and Tools business.

Looks like your right on this point.  In my day job we use the even more rare "per device" licensing structure so it wasn't an area I was that familiar with.   Still I guess this proves your larger point that Microsoft licensing makes everything overly complex. 



Agree 256 was a poor example.  Beyond 64GB of RAM (today future versions tend to increase based on Moore's law) it gets very expensive.  I think part of the problem is that Microsoft has no license specifically designed for use as a website backend.  It means their existing licenses are either too restrictive or too expensive.  There would be real value if Microsoft both simplified their licensing and offered a "web server core" version dropping most of the corporate bakend functions and just provide a high memory, high compute platform for webservers.   .... I am not holding my breath.   I guess Azure may be an option I have never done testing to see how responsive their database in the cloud can be to very high loads.


Yeah I can see why you would remember it then.


I agree there is value in open platforms and SQL Server is one of Microsofts most expensive and most restricted products.  I think this got somewhat sidetracked (partially my fault).  The original claim (not by you) was that asp.net doesn't scale and is expensive.  asp.net can use any RDBMS as a backend and is highly scalable.  For a new startup SQL Server is likely the easiest route as there is going to be some skills overlap but beyond 64GB or 16 cores I agree there is significant cost.  It is something a startup would need to consider.

As a personal note:
To avoid vendor lock in whenever possible I try to keep the database code as ANSI SQL compliant as possible to allow a "backdoor" by keeping the migration costs & complexity to a minimum.

None taken.

Instawallet is extremely slow these days, almost to the point of being unusable.

The number one reason is that the Bitcoin client is showing its limits. The wallet.dat has outgrown the blockchain data in terms of disk space usage. Now, in order to simply start up, the Bitcoin client takes an hour, an actual hour, just to load the wallet file. When creating a wallet it synchronously requests the creation of a new address from the Bitcoin client making the request block while the Bitcoin client is working, which takes a lot of time (between 5 to 10 seconds). Some sends very often timeout, that's why they're all handled asynchronously in a background queue now. Anyway, the problem is not Rails here, it's that I need to purge the wallet, keep only the active private keys and perhaps make more Bitcoin client calls asynchronous.

Bitcoin-Central is another story, it needs a hardware beef-up and some performance-oriented refactorings (more caching, more asynchronicity all over the place, etc). But that's ok, Rails is all about refactoring your apps as they grow, that's what fat automated test suites are meant for.

So it's really not about the framework, you can shoot yourself in the foot or write beautiful, maintainable and tested code in any language. Usually, the people who are the most vocal about Rails are those who wouldn't know how to "hello world" with it.

I've made several attempts to explore RoR, but each time something just turned me off about the language (syntax).  As you seem to favor the framework, how about dropping me a link to a good intro "video" tutorial? One day, I could make a lunch of it.


instawallet can be slow, but i wouldn't jump to any conclusions. in my experience, with all the cryptography, ssl and blockchaining, this could be quite normal at times. even blockchain.info will freeze my browser every now and then during certain activities.


damn, i forgot all about the OP. but much more importantly, who won the contest?

No offence davout but why are those sites so slow. I reccomend instawallet all the time to new users but it seems to take forever to load. I notice the same thing with my bitcoin-central account, very slow loading pages.

FTR Bitcoin-Central.net, Instawallet.org, Instawire.org are built with Rails.
Sucks though that the OP was simply looking for a developer and this thread turned into a my-penis-is-bigger-than-yours contest.

Sorry, this is mostly off the top of my head so it could be wrong.

This is not correct, you still need to have one CAL for every user you plan to have then. It is called 'multiplexing' and is technically not usable with a website. See:

http://serverfault.com/questions/120276/how-are-sql-server-cals-counted
From the licensing info doc I posted:

Pretty sure if they are worried about having you count users then a website on the Internet is fairly incompatible with with the Server+CAL model. Of course you would have to call MS licensing to be sure (bleh.)

Sure, but in your earlier example you posted a 256 GB RAM server with the example that 'RAM is cheap.' I agree that everything is pretty small now, but to get a server with mega RAM and find out you can't use it all because you need to pay MS more $$ is very frustrating.

4, 5 i agree with... 6 (32G RAM limit) bit me personally...

Yes, there was a reason I used only PostgreSQL in my earlier post, MySQL is pretty much dead ever since it was purchased by Oracle. PostgreSQL is completely Free, with almost all of the features of something like MSSQL Enterprise and I can see many many companies starting to migrate from MSSQL or even Oracle in the near term future.

It is wonderful not having to think about licensing restrictions, to be able to use the full capacity of your hardware at all times (buy a 256 GB RAM server? why not?), and not have arbitrary features locked out. And to not worry about things like the cost of CALs jumping 25% randomly with 2012 or BizSpark's 3 year grace period that runs out and leaves you paying MS for the rest of the company's existence.

Yes. right after i posted it occured to me that you were probably referring to more front-end management. was a little to quick to fire back. i really love the efficiency at which blockchain.info parses the blockchian in real-time. if they were in fact using RoR for this, that would have been a HUGE +1 for the platform/framework.


wow! this IS an awesome site. i'll be adding this to my bookmarks. thanks.

Oh yes I do agree about not rewriting the whole bitcoin deamon in C# when making a website. When needing to use a bitcoin deamon, it make a lot of sense to use a C++ or Java version instead, or even to use the hosted blockchain.info one (which is what I am currently using). But when it comes to things like needing to generate some bitcoin address / private key pairs, there is nothing wrong with developing this in C# and than using it in an ASP.NET website.

Builtwith.com is an awesome site for checking what frameworks etc a site uses. http://builtwith.com/coinbase.com shows it was built with Ruby on Rails. http://bitcoin.stackexchange.com/questions/5287/how-are-popular-bitcoin-wallets-implemented also backs it up. Also looks like strongcoin.com uses ruby on rails.

no worries. i've never used coinbase, but it looks like a very well developed website.  i'll take some time to properly evaluate its services and the technologies behind it.

EDIT: do you by any chance have a link to the mention of RoR?
I did find this http://news.ycombinator.com/item?id=4839772&buffer_share=7ab7b

In what world of business is taking the path of least resistance being lazy? Making informed decisions that will help minimize the foreceable issues in a project is quite "good" business sense. Selecting tools and services that are well tested and well maintained by the very same community that you are targeting falls into that category. I didn't mention RoR simply because I don't have any experience with it good or bad. I'm sure there are others that can make a valid case for its usage with Bitcoins.


Not to say that he couldn't, but WHY would Piuk would write his own bitcoin deamon in RoR, when the C++ and Java versions are so fully developed. Does that even make sense to you?  Are you suggesting that the OP take a similar route with .NET?

I stand corrected. I was sure I read somewhere it was developed in Ruby on Rails.

EDIT: Oops, it was http://coinbase.com/ which I was thinking about.

The path of least resistance is a lazy way of looking at things. When I look at make a website, I don't look at how fast I can pump it out, I look at how good of a job can be done with it. And I believe that is where languages like C#, Java, Ruby on Rails excel. Test Driven Development works well in these languages.

There is no language of Bitcoin. Blockchain.info is the most well done bitcoin related website. And it is programmed in Ruby on Rails. Using your logic, I guess Piuk didn't make a good choice of language for blockchain.info

I was strickly speaking from my "personal" experience.  Obvisously there are many devs who can make perfect sense of the Bitcoin protocol.  My point was that having a good programmer (even the best programmer) does not guarantee that THEY will also be able to make sense of it (the cyptography stuff is far from what ~99% of programmers have ever been exposed to).  And why would you want to lock yourself into using unmaintained libraries that are 18 months?

If the OP was a .NET programmer, or had some preference towards that technology, then I would have encouraged him to "go with what you know best". However, that is not the case and it seems to make sense to go with the path of "least" resistance (I hope you would agree that .NET is NOT that path). PHP, Java, Python, C/C++ are the languages of Bitcoin.


And there's the catch!  Programs like Bizspark may actually allow startups to get in with limited upfront costs, but then where to you go from there?  Hopefully it up.  And up comes with shit-loads of licensing costs. If the value added is acceptable to you then fine, but I'm arguing that you can get the same results with alternatives that WON'T lock you in to expensive licensing agreements.

I am admittendly biased against MSFT technologies, but that doesn't make my points any less valid.

paybitcoin ... not sure where you got your info

1) Startups can get free licenses using bizspark program for 3 years plus heavy discounts on conversion at "graduation".  Any company older than 3 years is hardly a startup.

2) Your pricing but it is way off.  Standard can be licensed per user (CAL) instead of per core which is much lower cost for applications like web backend datastore.

3) SQL Server 2012 Standard is limited to "only" 64 GB of RAM but it is hardly a deal killer for startups.  To my knowledge no Bitcoin venture uses a larger database than that (startup or note).  Very few companies in the world do and they tend to be very large established enterprises spending seven figures on IT costs.

4) If the 64GB limit is a concern, SQL Server 2008 R2 is an option and it has no memory limit.

5) SQL Server standard has always supported mirroring (Safety Full).  There is no additional license for a failover (passive) cluster.  Additional licenses are only required for a load balancing (active) cluster.

6) While Windows 2008 is limited to 32GB of RAM.  Windows 2012 standard has no such limit (well technically it is an OS limit of 4TB but that is regardless of license).


Which is why bizspark was created.  Three years of essentially unlimited licenses. Then a graduation with conversion of licenses at 70% to 90% off retail.   Any company older than three years is hardly a startup.Can it be done on other platforms for less?  Sure but there is no reason for FUD.  

Still asp.net can be use MySql (or any RDBMS) as the datastore.  As you pointed out the windows license is downright cheap compared to SQL Server so that is a lower cost option.  Microsoft's big competitor for SQL Server is Oracle and there pricing makes SQL Server look downright cheap.

This is exactly the point I was trying to make with scaling up on the Microsoft stack. If you were to go to Dell and build your $8k DB server, you would still need a $3000 license for Server 2008 R2 Enterprise edition (arbitrary 32 GB RAM limit with Standard; note this is better if you use Server 2012 Standard, then its just $800), and a $10,500 MS SQL Server Standard license, and even that can only use 4 of your 32 cores!!! Want to mirror your DB for redundancy? Sorry, that's Enterprise edition, (4 cores or 1CPU only even!) = $29,339.95 for each server!!!

Granted I don't keep up with MS licensing details since they change with every edition and are very confusing. But using Linux and PostgreSQL you end up with an entire redundant copy of your beefy DB server for the cost of 1 MSSQL license. This is especially bad since the MS cost doesn't hit you at the beginning on a small server when you are using their nice tools to develop it (or bizspark), but only when you need to scale to a beefy server and are already locked in.

Anyone proposing using the MS stack for a startup should be familiar with the licensing (PDF: MSSQL licensing, Windows Server licensing) up-front. It's a total mess. Even worse with VMs in the mix. That's pretty much the root of my problem with scaling with the MS stack; having just a bit of knowledge of Linux & related solutions you can run circles around an MS solution at a far lower cost. This is critically important for a startup without a reliable revenue stream.

BTW, we use the ASP.NET and MSSQL server stack where I work, and I have used it quite a bit (admittedly, for Intranet-style apps, but they get a reasonable amount of traffic.) It's great as a technical solution, but has a bad value proposition as a web startup solution.
---

Sorry, this counts as a troll thread. Granted it was a reply to a troll thread but I don't think that's the proper counter-troll response.

gweedo,

Do you have to angrily troll every thread you participate in??

I noticed you advertise "Java, PHP, HTML/CSS Programmer for Hire!"

You're really not doing yourself or your business much good with these kind of post.

Based on your posts, If I was 12 and met you on 4chan and wanted to create a justin beiber fan site I might hire you.

But for anything related to btc, probably now.

I'm just saying...

 

Encryption / Hashing is decided on a case by case basis. Usually when dealing with passwords I would use hashing / salting of the password. However depending on what I want to do with a bitcoin private key would depend on how I would store it.

And yup, you nailed the head on that, I use an ORM since I don't take time to learn about security.

If you read the article and look at how there setup is, they obviously are locked into a scenario where they can't even change databases, so the only way they can scale is with hardware. Now you can get more performance out of MySQL but either changing the database engine, or even using a mysql build that has beter performance and is tested. Kinda like twitter. SO the point I am trying to get across is that with ASP.NET and C# the only way to scale is thru hardware, and with other options you can just switch out some software and then you can do hardware scaling. So yea what would you want to do spend cash as a startup on hardware? Or go with this proven software that is free?

Actually no. I'm the kind of guy that uses a ORM (specifically http://www.mindscapehq.com/products/lightspeed) so I don't have to concentrate on making sure queries are escaped. I'm also the kind of guy that uses encryption / hashing when storing sensitive data. But thanks for sharing your concern.