Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

StackOverflow Update: 560M Pageviews a Month, 25 Servers

highscalability.com

Post reply on HN
Print view View original

271–278 of 278 posts

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#271

Post by edwinnathaniel »

bjourne wrote: post #270
edwinnathaniel wrote: post #257

Earlier quoted context omitted.

You do realize that most efficient DI framework only inject the dependency once. Separating Controller, Repository, and Services are good practices as well and let's be honest, we're looking at 3 methods layer at most. Here's what happened in Java: 1. When you deploy your WAR, the DI will inject necessary component _once_ (and these components only instantiated _once_ for the whole web-app so there you go, Singleton…

You probably meant to respond Marco's comment? And afaik Stackoverflow is written in ASP.NET, not Java.

C# or Java, the whole request pipeline processing should be more or less the same. Unless one platform does things less efficient than the other.

Rails and Django do things differently as to my knowledge they do it by spawning processes instead of threads. There are app-server for Rails or Django that may use threads for efficiency/performance reason but I am under the impression the whole LAMP stack is still 1 request 1 process (even though they re-use those processes from a pool of already allocated processes).

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#272

Post by BrandonM »

tjdetwiler wrote: post #252
BrandonM wrote: post #251

Earlier quoted context omitted.

Java 8 + an IDE like NetBeans (which isn't even the best Java IDE) is actually a very pleasant development experience. I am a total convert from Python in vim to Java in NetBeans+jvi. Refactoring in Python or Ruby or Javascript or whatever dynamic language is extremely painful compared to doing it in a static one. The "Software is never finished," argument can also work against dynamic languages.

Have you tried IntelliJ?

Yeah, I have. It seemed objectively better, but I started using NetBeans first, and it's what we all use at work. The workflows and keyboard shortcuts are ingrained enough at this point that switching doesn't really seem worth it right now.

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#273

Post by bhauer »

noelwelsh wrote: post #269
idlewan wrote: post #268

Earlier quoted context omitted.

Java servers get around 900K requests/s [...] while Rails squeezes out 6K Yes, but that's for the /json test, which serves a static object that the framework converts to json. I'm not sure it is really representative of the overall speed you'll get with whatever framework you'll be using: how often do you have a static page that's best served by your framework instead of leaving nginx serve it directly? I prefer watc…

The reduced variance of the /queries benchmark suggests to me that the database is the bottleneck in this setup. This would be the first thing you'd attack if you were optimising for speed. For instance checkout the section on StackOverflow's caching -- they have 5 levels of cache, including in-memory caches on the web servers. They are doing a lot of work to take the DB out of the equation. I prefer the /json test b…

This is a good point. It's true that for the high-performance frameworks and platforms, the database is the bottleneck. Or, more accurately—considering the trivial query and small payload—the overhead of the database driver and the wire protocol are significant factors. That said, there remains a fairly broad distribution over the Single-query test and Fortunes tests. For the low and medium-performance frameworks, the overhead of their ORM and other factors are more significant than the database. I find it can be illuminating that in many cases, the ORM code necessary to marshal result-sets into usable objects is more costly than the underlying queries.

Meanwhile, the 20-query test is a bit pathological as it runs into a brick wall with the database wire protocol and efficiency of the database driver. Many otherwise high-performance frameworks and platforms become bottle-necked waiting on those 20 queries per request. But you and I agree, a 20-query-per-request scenario should be the exception and not the rule. When developing a smooth-running web-application, it's common to aim for zero queries per page load. (For those who find this to be crazy talk, note that I'm saying we aim for that ideal, and may not necessarily achieve it.)

I too particularly enjoy knowing the high-water mark set by the JSON and Plaintext tests. When we add the next test type (caching-enabled multiple queries), we should see some interesting results.

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#274

Post by tim333 »

Alupis wrote: post #123
stonemetal wrote: post #37

Earlier quoted context omitted.

Joel worked at MS on office, Excel if I recall correctly. Jeff's blog is named for a feature from a book published by MS. I would be more surprised if it was MS free.

You mean to tell me Joel is partially responsible for the abomination known as Excel 2007 where I can't pull-drag the damn window from screen to screen (but all of the other Office 2007 products do!)

He seems to have left Microsoft in 1995, so no, not really.

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#275

Post by beachstartup »

csharrison wrote: post #180
beachstartup wrote: post #70

Earlier quoted context omitted.

beware that you can over-optimized this. for example, do not put every single customer of an saas solution into a separate database.

I'm curious, why not? If your saas was small enough and had few enough customers with large data needs, separating them each out into a separate database seems like a viable solution.

yeah, if you have an experienced team, you can easily refactor your one-customer-per-db model into something that will handle 1,000 new customers signing up every month. given a good enough team, you can even do this while doing other sort-of important things like customer support, bug fixes, and new development.

however, inexperienced developers and teams are not good enough, or fast enough, to do that, and end up painting themselves into a corner when they have thousands of live customers to support and need to change their entire application architecture + database schema when their backup, replication, and housekeeping tasks choke on 1,000+ databases.

and oftentimes, one-customer-per-db also means one-instance-of-application-per-customer, another anti pattern to avoid.

we see this kind of stuff ALL the time. it happens a lot - people make terrible decisions and then are stuck with them 5 years down the line and are looking at a monumental cost to redo. not everyone is experience enough to work their way out of an awful situation like that.

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#276

Post by shepik »

jacquesm wrote: post #234
shepik wrote: post #228

Earlier quoted context omitted.

> MS stack is more performant than the Linux one on the same hardware Could you please provide some evidence to support that?

Without breaking any promises to others here is an article with conclusions quite similar to mine: http://www.webperformance.com/load-testing/blog/2011/11/what... Keep in mind that it hardly ever is the webserver that is the bottle neck. My theory on why this is the case is very simple: MS can afford to throw vast amounts of money at optimizations that are next to impossible in Linux simply because the coupling betwe…

Tight coupling and money are a good arguments, thanks. However, i believe we are comparing apples to oranges here. IIS uses it's own kernel module, HTTP.SYS - http://www.microsoft.com/technet/prodtechnol/WindowsServer20... So, looks like having specific kernel api optimized for your particular usecase is an advantage, which IIS has and others don't.

It turns out there are kernel-mode webservers (or there were, at least), which seriously outperform user-mode ones. https://www.usenix.org/legacy/events/usenix01/full_papers/jo... (page 11; article is old, but anyway)

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#277

Post by bikamonki »

cruise02 wrote: post #256
bikamonki wrote: post #151

I never ever ever never use SO's search feature: I land on the result through Google. So I guess they should give Google some credit for the claimed performance?

Google isn't serving Stack Exchange's pages to you. They published stats on pages served by their own servers. Google has nothing to do with it.

You miss the point: I never use the search feature of the site, I go to the answer/subject directly from Google's results. I do not think Google is running my keywords against SO's server in 'real-time' and returning the answers as results, Google returns indexed/cached copies. I am sure that many users behave like I do, no? They might as well keep a static copy of all pages in RAM and only update when new comments/answers are posted.

Re: StackOverflow Update: 560M Pageviews a Month, 25 Servers

#278

Post by jtseng »

superuser2 wrote: post #199
bradyd wrote: post #14

Earlier quoted context omitted.

I don't believe that much RAM is uncommon for large scale database servers. 384GB RAM is only about $5000 from Dell. They also have a new server model coming out that supports up to 6TB of RAM [0]. [0] http://www.dell.com/us/business/p/poweredge-r920/pd

How common is it to buy one extremely powerful database server? Wouldn't you need the high availability properties of something like a Galera cluster? Or are most people accepting the single point of failure at huge database server? (I honestly don't know, it's just surprising.)

They do have a cluster: http://www.brentozar.com/archive/2012/09/microsoft-sql-serve...

They just run everything through one live database all the time under normal conditions. Personally, I think this is much simpler to manage than to be regularly spreading out queries over many queries.

Post reply on HN