Posted on 07/07/2011 8:55:49 PM PDT by TenthAmendmentChampion
So you think Oracle would handle Facebook on a single instance? Of course you shard when you have 750 million active users!
Of course, eBay's pretty much killed the business - we're just on liquidation mode now with the most recent ‘changes to better service the customers’ - IE: more fees raised and more demands to make shipping free, while shipping becomes ever more expensive.
Though the bright spot was a couple days ago - Economics graduate who does shipping part time finally figured out that the business is unsustainable between eBay fees and eBay (Paypal) fees. Then again, got a call from a competitor asking if we wanted to buy them out, offered to sell out to them.
Long and short of it, eBay's wanting to only cater to small sellers in the US, and giving away the store in Asia. Our eBay/paypal fees would be a quarter of what they are now if we were based in Hong Kong selling the same products.
Just don’t say normalize.
Or cleaning woman.
A query walks into a bar, sees two tables, says “Mind if I join you?”
Oh, well, the 300,000 are the base items. Atop that are 700,000 orders, 300,000 customer records, 800,000 current sales records, and 4 million archived sales records. Call it 6 million for simplicity.
Don't laugh. Just a couple days ago, a dental person asked about the programming business. I told her that pretty soon, American programmers will be helping people run their businesses on Excel spreadsheets, powered by off-the-grid mini-solar panels, without any internet to be found. She didn't laugh, and I didn't mean it to be funny.
LOL
I thought it was Dr. Pepper.
I haven't seen a design phase included in any project since Y2K.
Design functions happen at the individual developer level, and get discussed over lunch (if at all).
So long as the economy is booming, with no debt and no unemployment, lunar landings every month, business and industry running like well-oiled machinery, we don't need no steenkin design.
Sign on the wall at my work: WYGIWYG
Ooooh, what a stinker!
[I am SO stealing that...]
IMHO...
They started off with an application / system design that was not scalable.
Once you go live making drastic design changes to the live system becomes a non-trivial task which carries signficant operational and budgetary impact.
To be successfully scalable to such throughput volumes systems don’t employ architectures like the simple garden-variety database application, for example, a typcial sales invoice system that has a front end user interface that directly updates tables with a layout that mirrors the users screen in a relational database when the user clicks Save. While that approach is fine for most normal applications, it is limited in it’s throughput to the transaction performance of the given database running on the given platform, which is sometimes not satisfactory under super-huge loads such as hundreds of thousands of simultaneous unique users all updating data. The stop-gap hosehead method of getting the required scalability in such a situation (like fbook) is to do what they did, go beserk using every db adminisration technique possible, which sometimes will suffice if you find solid db professionals accustomed to the throughput that is needed. However if throughput requirements simply continue to climb and even your db administration experts start coming to the end of what’s possible, then the garden-variety design paradigm must be abandoned.
Once the requirement to design in that manner is dropped and any and all solutions that a system programmer can create are acceptable, such as writing special purpose server software, etc., then with today’s hardware phenomenal performance is possible. When you consider that today’s computers are executing billions of instructions per second and drives can transfer gigabytes per second, the horsepower is there to do quite a bit if elegant software designs are employed. In fact, those who write the database engine software are the kind of programmer who solves these problems, they are just developing the db, which is a generic solution instead of one targeted to a specific application like fbook.
Of course, very good performance is available using fairly standard approaches as long as the developers are used to developing software in the performance range that is desired. Think of trading platforms today that process huge amounts of transactions and have very stringent requirements for both integrity and real-time performance. But if standard approaches won’t do, there are all sorts of options. Consider google: they knew that file system performance would not scale sufficiently for them so they created their own google file system, or gfs. Chances are most people don’t even understand, in truth, exactly what that means, to write their own file system, or how one would even know if one could benefit from doing that. Many people any more don’t know or like to acknowledge that programming is still generally broken down into the fields of application programming and systems programming, so in cases like fbook they do not go out and hire the right types of systems progammers from day one.
There are qualified people out there, but in cases like idiotbook - I mean facebook - where one harvard brat with no heavy (meaning high-performance) database programming experience is viewed as some kind of genius by investors and is at the helm of the corporation, an inmate is in charge of the asylum. Like bill gates, his personality is very well suited towards building an empire, i.e., product positioning and retaining management control for himself, but not towards technical prowess. He was very much a one-man-band in terms of being the top technical decision-maker in the early days of fbook. Rather than have a real architect come in who would undoubtedly seek significant ownership at that point, in good-old-American IPO style, he maintained his unquestioned position as top-dog and planned on bringing in people to handle the real big-time scalability issues after his top-dog status was eminently secure and the company was producing a lot of revenue. It’s no great tragedy, of course, it’s just an FYI to those who consider investing in tech IPOs - hyped yet murky valuations, business models, management and underlying technology can make it a wild ride.
True technicians understand that if you have a team of guys working on a project, each has their own job, and you work together: when the xyz expert comes in to xyz, you lettem have at it. And amongst true technicians, it quickly becomes apparent who overreaches and who can always deliver when they say they will deliver. Good teams tend to drop off everyone else. Bad teams result from promoting the wrong people. The challenge then is for management to know when they are only hacking in a given technical subject, and they need to go get a new person who is a real kick a$$ expert in that technical subject if no one can or will put in the effort needed to get really knowledgable of the subject in question. They must be able to deliver the industrial-strength code needed and also have the vision to come up with not only passable design ideas, but excellent ideas verging on optimal solutions given the project constraints. Most college graduates and entry-level programmers today are assumed to have far more capabilities than they do. We often forget that like any other trade or profession, programmers benefit from years of deep and broad experience as well as pure study and a serious interest in their field.
It all comes down to, as Herman Cain says, having the right people in the right job and just letting them do what they do. Once a team has the right people and is firing on all cylinders, it’s good.
Just be aware the use of “query” might make some ansi.
Those definitely are numbers where you need to start paying attention to MySQL.. Still, that shouldn’t be too bad as long as the hardware is ok. Unless you are table scanning the 4 million archived records..
Seems to me, FB has the resources to start or fund development of its own database, as you say. That's probably what it will eventually have to do.
My understanding is that Google’s Eric Schmidt was shown the door because he didn’t secure a proper agreement with the “Borg”; - and now his former company has been exposed to all sorts of potential liability.
Now you’re projecting your own fears. In my view, you should really restrict that.
Gah ZEEEEEEE Bo.
The poor man’s Flying Buttress!.....
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.