In IE8 and Chrome, Processes Are the New Threads

SenFo writes "To many of the people who downloaded Google Chrome last week, it was a surprise to observe that each opened tab runs in a separate process rather than a separate thread. Scott Hanselman, Lead Program Manager at Microsoft, discusses some of the benefits of running in separate processes as opposed to separate threads. A quote: 'Ah! But they're slow! They're slow to start up, and they are slow to communicate between, right? Well, kind of, not really anymore.'"

Comments

Have to watch what I say

[bigtallmofo]

I may be inadvertently responsible for Internet Explorer 8's use of separate processes for each tab. Months ago, when they invited me to install the beta of their latest web browser, I told them to do something that sounds very similar to "Go fork yourself!"

I think they took that as architectural advice.

Re:Have to watch what I say

[dwywit]

What, something like hawking fell ?

Processes

[Ethanol-fueled]

Running each instance in a seperate process is NOT new technology, hell, any n00b who knows what JCreator is has seen that option before(see this comment I posted awhile back).

Increases in computing power have made insignificant the perceived sluggishness of running multiple processes -- if Chrome won't run smoothly on that Pentium 2 of yours, then perhaps you should install command-line linux anyway! :)

Regarding Chrome, check out this response to my comment I linked to above, posted on June 30. At the time, I thought it was just an extension of a good idea but since his comment was posted earlier than Chrome was released I'm beginning to wonder if that fellow had any inside knowledge...

[/tinfoil hat]

Re:Processes

[ergo98]

Running each instance in a seperate process is NOT new technology

Well of course. It isn't even new in the browser world. In fact it's where we started.

The earliest browsers required you to run a new instance for each concurrently opened site. This presented onerous resource demands, so they made it more efficient by having multiple window instances run under one process, and then with tabs that obviously carried over to tabs running under one shared process.

This is so much ado about nothing. I can count on one hand the number of times I've had a problem with Firefox that would have been solved by it being in its own process.

Every bandwagoner, technical lightweight is now stomping their feet that Firefox needs to get on this yesterday, but really this is pretty low on the list of things that make a real improvement in people's lives. In fact I would go so far as to call it a gimmick. Presuming that the sandbox of a browser automatically stops sites from doing stupid stuff (unlike IE that will let a site kill just by going into a perpetual loop in JavaScript), and plug-ins are created by an idiot, this is completely unnecessary.

Chrome's great JavaScript is a real story, one upped by Firefox's ThreadMonkey doing one better. Those are real improvements that really do matter.

Re:Processes

[ThePhilips]

Running each instance in a separate process is NOT new technology [...]

True, *nix does that for last 3 decades.

The point here (and of RTFA) is that finally on Windows processes are becoming cheaper, making them usable for all the nasty stuff *nix was indulging itself for all the time.

On NT 3.5, creation of new process was taking sometime as long as 300ms. Imaging Chrome on NT: if you open three tabs and start three new processes, on NT in the times, that alone would have taken about one second.

Unix never had the problem. It's Windows specific. And they are improving.

Re:Processes

[Firehed]

javascript application

Well there's your problem...

Re:Processes

[FishWithAHammer]

It's primarily improvements in computer speed. Threads are very cheap in Windows (and this is why .NET in particular is so heavily dependent on spawning tons of threads for many types of tasks) and processes remain fairly expensive, but that expense is somewhat minimized by being about to throw ten kajillion bogomips at the problem.

Re:Processes

[piquadratCH]

Not to rain on your parade, but exactly how are you intending to use browsers in cluster computing?

Didn't you get the memo? The browser is the new operating system! So, naturally, it belongs on a cluster.

Re:Processes

[firefly4f4]

I can count on one hand the number of times I've had a problem with Firefox that would have been solved by it being in its own process.

Possibly you can, but the biggest one I can count is having a flash plug-in (or similar) crash the entire browser when there's only a problem on one tab. That happens more frequently than I'd care for, so if there was a change that only brought down one tab, that would be great.

Re:Processes

[ThePhilips]

Threads historically were expensive because historically nobody used them. E.g. on Windows IO multiplexing had some history of bugs and didn't worked in the beginning - you had to use threads (and people used to use threads and even now use threads instead of NT's IO multiplexing). On *nix IO multiplexing worked more like always thus threads were (and are) used rarely.

Now, since number of CPUs increased dramatically in recent years, threads were optimized to be fast: developers now throw them as panacea at any task at hand. (Most stupid use of threads seen in the month: start new thread for every child application to wait for its termination; due to stupid code, it still might miss its termination).

As a system developer, I have went trhu user space parts of Linux 2.6 and Solaris 10 threads implementations (in disassembler; x64 and SPARC64 respectively) and can say that they are implemented well. (I was looking for example of atomic ops implementation.) Kernel parts of both Linux and Solaris are well known to perform extremely well, since they were tuned to support extremely large Java applications (and Java only now got IO multiplexing - before that threads were only option to perform multiple IO tasks simultaneously). HP-UX 11.x also didn't showed any abnormalities during internal benchmarks and generally its implementation is faster than that of Solaris 10 (after leveling results using speed of CPUs; SPARC64 vs. Itanic 2).

But I guess "slow *nix threads" is now the myth of the same kind as "slow Windows process creation." (Problem is of course that process creation in Windows would always remains expensive compared to *nix. But not that those millisecons of difference matter much for desktop applications.)

Re:Processes

[smartdreamer]

Actually, threads are heavier in Windows than in every other OS including Linux, MacOSX and Singularity. Still they are way cheaper than a process and that is the whole point of their existence. As for .NET, it uses its own built-in threads model; .NET threads are different than OS threads.

Re:Processes

[apoc.famine]

Currently that's spelled "half of the internet". The other half is flash.

Re:Processes

[jc42]

Running each instance in a seperate process is NOT new technology

Well of course. It isn't even new in the browser world. In fact it's where we started.

And, as us old-timers know, this architecture was the basis of the original Bell Labs unix system, back in the 1970s. Lots of little, single-task processes communicating via that newfangled sort of file called a "pipe". That was before the advent of our fancy graphical displays, of course.

Somewhat later, in the mid-1980s when we had graphical displays, the tcl language's tk graphics package encouraged the same design by adding to the usual unix-style pipes and forked processes. The language had a simple, elegant "send" command that would let a sub-process run arbitrary commands (typically function calls) inside the parent process using the sub-process's data. The idea was that you have a main process that is responsible for maintaining the on-screen window(s), but the work is done by the sub-processes. This design prevented blocking in the GUI, because the actions that would block were done in the other processes. The result was a language in which GUI tools could be developed very quickly, and would take maximal advantage of any parallelism supplied by the OS.

But that was all decades ago, before most of our current programmers had ever touched a computer. Imagine if we only knew how to design things that way today. Is it possible that current software developers are rediscovering this decades-old sort of design?

Re:Processes

[Johnno74]

Ahh .net threads are regular win32 threads. They are scheduled by the OS, not the runtime. .Net code is (well, can be...) much, much more robust than c++ code as its managed, and type conversion errors, null references and other things can be caught and recovered from more gracefully than C++. This applies to Java too, of course.

Re:Processes

[Chrisje]

The funny thing is that I'm working on a shiny new HP 6910p laptop. The kind with ~6 hrs battery life, a good deal of memory, a fast CPU and even a decent GPU. Everyone goes on and on about how the "cost" to start different processes all the time is no longer significant, but I really noticed the difference. I run FireFox 3 with a whole bunch of plug-ins and a nice skin. That contraption, in spite of the plug-ins, feels quite a bit faster than the Chrome browser does out of the box. I've tested Chrome yesterday, and at the end of 6 hours of work (in which everything did work off the bat, even starting my Citrix apps from a web portal, kudos to that) I concluded that firefox feels leaner and meaner, and went back to it.

One of the major gripes I have is that it feels like FireFox is much quicker with regards to my proxy. We run a proxy configuration script which gives us different settings depending on which office we're in, and in Firefox I never notice the damn thing running. Now in Chrome, whenever I open a new tab, I see the damn thing executing. New process, sandbagging (or whatever you call it)... bah, humbug. I agree with the parent poster. I can count the times when that would have come in handy on the fingers of one hand after 3 versions of firefox on my system, and it makes the experience noticeably slower.

Cut a long story short: I appreciate it's a beta. Come release time I'll give it another whirr. But right now, I don't see what the big hubbub is about save the fact there's another Open Source competitor on the market, which is always good. What is funny is that when you de-install Chrome, a dialogue pops up asking "Are you sure you want to uninstall Google Chrome? - Was it something we said?"

So...

[Anonymous Coward]

...his argument that processes aren't really slower than threads anymore is because your processor is faster?

Oblig

[plopez]

"The 70's called...." I can't bring myself to say the rest....

Re:Oblig

[Anonymous Coward]

1990 called, they'd like their joke back.

Deja vu

[overshoot]

I remember the "processes vs. threads" argument, but last time around wasn't it Microsoft arguing that a threaded process model was superior to an isolated task model like Linux had? Weren't the Linux camp blowing the horn for the superior robustness and security of full task isolation?

My head hurts, I'm confused.

Re:Deja vu

[Anonymous Coward]

Windows people never really understood processes, they cannot distinguish them from programs (look at CreateProcess). They traditionally don't have cheap processes and abuse threads.

In Linux we have NPTL now so there is a robust threads implementation if you need it. I don't thing processes are "superior" to threads (processes sharing their address space) or the other way round. They are for different purposes. If you need different operations sharing a lot of data go for threads I would say.

Re:Deja vu

[FishWithAHammer]

Both have pluses and minuses, as with anything. (I won't speak to the Unix model as I am not terribly conversant with it, but I know a good bit about the Windows model of threaded processes.)

A threaded process model has one enormous advantage: you stay within the same address space. Inter-process communication is annoying at best and painful at worse; you have to do some very ugly things like pipes, shared memory, or DBus (on Linux, that is). Using the threaded process model, I can do something like the following (it's C#-ish and off the cuff, so it probably won't compile, but it should be easy to follow):

class Foo
{
    Object o = new Object(); // mutex lock, functionality built into C#
    SomeClass c = new SomeClass();
 
    static void Main(String[] args)
    {
        Thread t = new Thread(ThreadFunc1);
        Thread t2 = new Thread(ThreadFunc2);
        t.Start();
        t2.Start();
        while (t.IsRunning || t2.IsRunning) { Thread.Sleep(0); } // cede time
    }
 
    static void ThreadFunc1()
    {
        while (true)
        {
            lock (o)
            {
                c.DoFunc1();
            }
        }
    }
 
    static void ThreadFunc2()
    {
        while (true)
        {
            lock (o)
            {
                c.DoFunc2();
            }
        }
    }
}

In an isolated task model, this is nowhere near as simple. The problem, though, is that one thread can, at least in C++, take down the whole damn process if something goes sour. (You can get around that in .NET with stuff like catching NullPointerExceptions, but you'll almost certainly be left in an unrecoverable state and have to either kill the thread or kill the program.) The Loosely Coupled Internet Explorer (LCIE) model is forced to use processes to avoid taking everything down when one tab barfs up its lunch.

Re:Deja vu

[shird]

For the majority of local applications, a threaded model is superior. This is because local applications can be "trusted" in the sense they don't need to run each child thread sandboxed etc, so they gain the benefits of greater efficiency without worrying about reduced security. A browser is quite a different beast - it is effectively an OS to run remote "applications" (read: web 2.0 style web sites). So it kind of makes sense to run each as a seperate process.

Windows the OS still runs each application in its own process. So it's not right to compare it to Chrome and argue that it doesn't use seperate processes, because it does - where it counts.

Re:Deja vu

[ratboy666]

Yes, you are correct. Unix started with a process model based on fork() and explicit IPC. Threads were "grafted on" later. It tends to result in more robust software (good multi-threading is HARD).

In Linux a "thread" is a "process", just with more sharing. Thread creation is cheaper in Windows; process creation is cheaper in Linux. I tend to like the isolation that processes offer (multithreading brings with it the joy of variables that can appear to just change by themselves).

There was never any good reason to NOT use multiple processes in a browser, except one. The GUI was "unified" amongst the browser Windows, and it has always been presumed that it would be too difficult to co-ordinate the drawing of multiple browsers. Also, the menu bars and controls would have to assigned to a separate process for each of the browsers. This can be done with an IPC channel, but that code would not have been portable between Unix and Windows at all.

Since process creation was SO expensive in Windows (in days of old), the "thread" or "lightweight thread" approach was used instead (to maximize portability).

It is an amazing testament to Google that they have achieved the multi-process, single UI model (I just don't know how they did the portability part).

Re:Deja vu

[The Raven]

This is a misunderstanding of the application.

Microsoft said 'Threads are better than Processes for a web server', where you're wasting a ton of resources creating a new process for every CGI script that's run. They were right! Now every major web server supports in-process applications that are created once per server (perhaps with a pool of shared app space) rather than once per request.

Microsoft has never said that all the applications on your computer should run in one thread... that's just crazy talk.

This is simply a decision by Microsoft and Google to treat a browser tab as an application, rather than as a document. Now that web pages do a lot more processing (and crashing), this makes more sense than the old way. There's nothing particularly bad about using threads instead... Firefox is just fine with threads, I see no reason for them to undertake a massive change due to misplaced hype.

It really has to do with how much the processes share. If most of the memory per process could be shared, threads are probably more efficient. If not, processes. I'm no browser architect though, so I'll leave it up to Google, Mozilla, and Microsoft to make their own decisions.

Re:Deja vu

[aspx]

You just defended Microsoft on Slashdot and you got modded +5 Insightful. I will follow you anywhere, O great one!

Cheap and Dirty Crash Isolation...

[nweaver]

The real reason for processes instead of threads is cheap & dirty crash isolation. Who cares about RPC time, you don't do THAT much of it in a web browser.

But with more and more apps being composed IN the browser, you need isolation to get at least some crash isolation between "apps"

Limitations

[truthsearch]

There are some details to Chrome's sandboxing implementation that limit its security benefits:

- The process limit is 20. Anything requiring an additional process once this limit is reached, such as opening another tab, will be assigned randomly to any of the existing 20 processes.

- Frames are run within the same process as the parent window, regardless of domain. Hyperlinking from one frame to another does not change processes.

There are also some problems where valid cross-site JavaScript doesn't work. Of course it's still only a beta. Some specific details are documented by Google.

Processes in Vista

[hardburn]

I remember a story from a long time ago, during Longhorn's early development, where Microsoft did a study of the cpu cycles needed for various tasks between WinXP and Linux. I've never been able to track the study down again since, but I remember that creating a new process took about an order of magnitude more cycles on Windows than Linux. Linux processes are also lighterweight in general; Linux admins think nothing of having 100 processes running, while Windows admins panic when it hits 50.

(The basic reasoning goes that Linux has an awesome processes model because its thread model sucks, and Windows has an awesome thread model because its process model sucks. That's why Apache2 has pluggable modules to make it run with either forking or threading.)

A lot of development from the early Longhorn was scrapped, so how does Vista fare? Does its process model still suck?

Re:Processes in Vista

[bluefoxlucid]

The basic reasoning goes that Linux has an awesome processes model because its thread model sucks,

NPTL has scaled to hundreds of millions of threads created in under 10 seconds on common hardware....

Slow to start a process!?

[Sloppy]

They [processes] 're slow to start up

It's hilarious anyone would think that. We're talking about a web browser, not a web server. Even on platforms where process creation is "slow", it's still going to be instantaneous from a single human's point of view. It's not like the user is opening 100 tabs per second.

Re:Only one benefit discussed: isolation

[ceoyoyo]

From the other perspective, having used IE in the past, I know how easy it is for a page to open lots of popups. In fact, you could open so many popups that it would crash the browser.

Now that the browser likes opening new processes, an out of control web page can crash my whole OS instead?

Requirements/Trade-offs

[bill_mcgonigle]

There are at least three problems here.

One is efficiency. Nobody will argue that a properly implemented multi-threaded software project is going to be less efficient than a new process per job. If you're writing a server to handle 100,000 connections simultaneously you probably want to use threads.

One is necessity. If you're only going to have at most a couple hundred threads you don't need to think in terms of 100,000 processes - orders of magnitude change things.

The last is correctness. Most multi-threaded browsers aren't actually implemented correctly. So they grow in resource consumption over time and you have to do horrendous things like kill the main process and start over, which loses at least some state with current implementations.

So theory vs. reality vs. scale. There's no "one true" answer.

Re:Requirements/Trade-offs

[FooBarWidget]

"If you're writing a server to handle 100,000 connections simultaneously you probably want to use threads."

Actually, if you want to scale to 100000 connections then you will *not* want to use threads. Google "C10K problem".

Re:Chrome = slow as hell

[B3ryllium]

"to load 8 sites" ... 8 sites that you visit frequently and thus have cached on your Firefox installation, perchance?

Don't be so rash to judge. Chrome has many other areas where it lacks compared to Firefox, speed isn't generally one of them. I've heard many users say that it loads pages more than twice as fast as Firefox, and also scrolls much faster on graphical/data-intensive pages.

The lack of extensions (such as adblock, firebug/firephp, flash block, noscript, coloured tabs) is the main reason why I've barely used it since I installed it.

quite ironic

[speedtux]

UNIX didn't have threads for many years because its developers thought that processes were a better choice. Then, a whole bunch of people coming from other systems pushed for threads to be added to UNIX, and they did. Now, 30 years later, people are moving back to the processes-are-better view that UNIX originally was pushing.

Microsoft and Apple have moved to X11-like window systems, Microsoft and Google are moving from threads to processes, ... Maybe it's time to switch to V7 (or Plan 9)? :-)

Re:quite ironic

[the_B0fh]

And debugging threads is easy? Oh boy, talk about a crack pipe.

Re:Could someone give the gory details

[Shin-LaC]

I went looking for the same information earlier today. Surprisingly, the design document titled "How Chromium Displays Web Pages" doesn't shed any light on that, at least at this time. You have to dive into the source to find out.

Basically, a single process (the one main browser process) owns the window and draws to it. Renderer processes draw their web content into shared memory; the browser process then transfers the data into a backing store, which it uses to paint the window. The process is coordinated via inter-process message-passing (using pipes, it seems), but the rendering output travels via shared memory.