An old Macintosh IIci 25Mhz running Apache under NetBSD

http://littledork.err0neous.org

by emersonrsantos • 12 years ago

80 74 12 years ago

littledork.err0neous.org

nsxwolf• 12 years ago

Static page performance appears very high.

sliverstorm2• 12 years ago

... and it still loads faster than most of the web today.

dubcanada• 12 years ago

All 986 bytes of it.

owenversteeg• 12 years ago

You can do a lot with that many bytes you know :) http://minfwk.com is only 9 bytes more (I'm the creator)

nodesocket• 12 years ago

Proof that web servers serve static content insanely fast and efficiently. Would be a smaller memory footprint using nginx.

UPDATE: And it appears to have gone down. Foot in mouth!

_delirium• 12 years ago

A single page load with no AJAX roundtrips, and no client-side scripts running helps a lot. Was reminded of that when I clicked on a search result to a mailing list archive today that happened to be hosted at Google Groups (vs the usual Mailman-style archives), and it took something absurd like ~4-5 seconds to load.

sliverstorm2• 12 years ago

Basically I see it as an example of how we spend performance on complexity and overall stay at a similar net level of performance.

vidarh14• 12 years ago

It's quite amusing that this seems interesting today. Apache works on really old, slow hardware quite well. E.g. here's Apache, MySQL and PHP for AmigaOS: http://www.amigaos.net/software/77/aamp-apache-mysql-php

My first job was an ISP I co-founded. In retrospect, the biggest mistake we did was over-investing in hardware, by buying two 120MHz Pentium servers when we could have gotten away with 486's and a leased Cisco access server or two.

But those two 120MHz Pentium servers with 128MB a piece, between them served up dozens of corporate websites, handled 32 dialup connections (we had a horrific setup with Cyclades serial port multipliers wired up onto a large wooden board with US Robotics Sportster modems wall mounted - the Sportster were cheap consumer grade modems prone to overheating, so stacking them in any way was generally a bad idea), ran a NNTP server with 30,000 USENET groups, mail for all our customers, and shell access.

We got 5-6 25Mhz or 33MHz 486's with 16MB RAM to use as X servers, and ran all the clients (Emacs, Netscape and a bunch of terminals was a typical session) on the two Pentiums.

nacs2• 12 years ago

We got 5-6 25Mhz or 33MHz 486's with 16MB RAM to use as X servers

But did those have Turbo buttons? If so then you were clearly cheating.

sebcat• 12 years ago

IIRC, "Turbo" was actually the normal clock frequency. Not having turbo-button enabled was mostly for things like games where the game speed was tied to the number of CPU ticks instead of absolute time. Turbo button sounds better than "slow-down-my-PC button".

enneff• 12 years ago

When Apache was written that wasn't old, slow hardware.

vidarh14• 12 years ago

Exactly. The 120 MHz Pentium servers were state of the art when we bought them. Some of the fastest we could get our hands on. And way too expensive (I seem to remember we spent the equivalent of about $16k/piece on them - driven up by the absurdly large 128MB RAM, and the huge diskspace of a few GB - but I may be mistaken).

I was primarily an Amiga user still at that time - happily using an Amiga 3000 quite similar in hw specs to the Mac in the linked article as my primary desktop.

budu3• 12 years ago

So what eventual became of the ISP, if I may ask?

vidarh14• 12 years ago

We sold of the dialup part to the Norwegian subsidiary of International Data Group for a pittance after we in our naivety (we were mainly CS students with no business experience at the time) started a price war with the largest ISPs.

Biggest lesson apart from being careful about large capital investments: If you think you can do things _far_ cheaper than an incumbent in an underdeveloped market, chances are they too can do things far cheaper, but have so far opted not to; forcing them into a price war without large backing is not a good idea.

Then we continued for a few more years doing consulting and business hosting as well as handling the backend support for IDG's Norwegian ISP, before the lawyer of our original investor hooked us up with much better paying jobs for the Norwegian office of a California based startup.

It was an interesting experience. Learned a lot. Didn't make a lot of money. But it defined my career - 6 out of my 8 subsequent jobs have a "lineage" going back to that company.

bane13• 12 years ago

Funny, my 2nd job was as employee #1 for a local ISP. We had a very similar arrangement (USR modems and all) for dial-in until we financed out and bought proper hardware, I wish I could remember the details. We hosted something like 3000 users, all their websites, shell access, Usenet, everything off of just 2 or 3 Pentiums with even fewer specs than what you describe. Everything ran quite well. We even ran a couple very high volume corporate websites for some rather famous television shows of the time off of this arrangement.

I'd point out that it appears even a high volume site like HN hasn't knocked this ancient system offline.

It drives me nuts when I read about figuring out how to handle some couple hundred requests a second on racks full of modern hardware with Gigabytes of RAM. Any clearance bin laptop you can buy at Costco should be able to handle thousands to tens-of-thousands of requests per second. We're clearly doing something wrong these days.

meifun• 12 years ago

I owned an ISP in upstate NY for a few years. Basically hosting websites for clients. I bought a used APC NetShelter cabinet and put 2 large UPS's in the bottom along with and Air conditioner cooling fan in the top. I have a Sun Fire V 1 u server that was running apache. I have a few Sun Blade 100 Workstations handling mail and a few more handing DNS, NTP, etc. All I remember spending at least 25K on all of this.

One interesting thing that I remember was trying to build a drive array back then. Trying to figure out how many IDE drives I could stuff into a 4U case to use as what we call NAS now a days.....

I did this from my house and had to upgrade to 200amp electrical service and have dedicated circuits run to my house. In order to do this the phone company had to dig up part of the road and it took them a week.. Not convenient for all that traffic down my street.

vidarh14• 12 years ago

The thing is, so much was static. It's easy to shuffle bytes.

There are a few problems: People don't know or care about profiling the basic stuff like context switches and data copies; people don't have a baseline idea of what should be possible; abstractions upon abstractions upon abstractions even when it complicates the code.

One of my biggest pet peeves in that respect is that people tend to not even know how many objects they create in many dynamic languages. Many people don't even know of any simple ways of finding out.

My most successful application speedup (EDIT: measured in speedup per hour expended - plenty of examples of much greater speedups, but they are rarely as quick) was spending an afternoon eliminating string copies in a late 90's CMS written in C++, and cutting page generation time by 30% in a system that was already heavily optimized (it provided a statically typed C++ like scripting language that was bytecode compiled at a time when most competing solutions were using horribly inefficient interpreters, and were often themselves written in slow intepreted languages).

My second biggest pet peeve is when people think system calls are cheap because they look like function calls. Sometimes you can increase throughput by an order of magnitude just by eliminating small read()'s (non-blocking filling of userspace buffer instead, and "read" from that). When I see slow throughput, the first thing I do is break out "strace" to look for unnecessary system calls...

But these problems are often a direct result of abstractions that means huge numbers of developers don't know or understand the cost of a lot of the functionality they depend on any more, and can get away with that "too often" because they work on hardware that is so fast it is extremely forgiving (e.g. my new "weird unknown Chinese brand" 8-core phone does about 10 times as many instructions per second as my ISPs total computing capacity in 1995, and it's nowhere near the peak performance of todays flagship phones), ignoring the fact that at "web scale", things are not as forgiving any more: server costs quickly start adding up.

But of course, most of us most of the time work on stuff that needs to handle little enough traffic that paying an extra $20k for hardware is cheaper than spending the developer time to optimize this stuff. Ironically this makes the optimization more expensive: Fewer developers ever get to obtain the experience to optimize this stuff at scale.

userbinator• 12 years ago

Is this a "load test"? It's certainly handling it quite well! Performance of the CPU is ~9 Dhrystone MIPS - comparable to a 33MHz 386 (9.9 DMIPS). Serving static content is really not CPU-intensive; most of the computation would probably be in TCP/IP processing. I think this, and other examples of old hardware commonly thought to be too slow for anything but still doing something useful, shows that much of the software we use today is really vastly less efficient than what's theoretically possible. I.e. how to make use of limited computing power efficiently can be more important than how much of it there is.

For comparison, here is a static site served by a machine with a 4MHz custom CPU: http://www.magic-1.org/ (more details at http://www.homebrewcpu.com/ )

adfm• 12 years ago

The Macintosh IIci came out in 1989. Pretty good for a 24 year old machine! I pull mine out every once in a while to remind myself that we haven't advanced that much in a quarter of a century.

api• 12 years ago

I had a Commodore 64 with GEOS as a kid. It always amazes me what you can do without umpteen gazillioplex levels of abstraction.

vidarh14• 12 years ago

Including surfing the web: http://www.techrepublic.com/blog/classics-rock/surf-the-web-...

And some people have run web servers on C64's too.

Contiki OS runs on C64 and provides a full TCP/IP stack for the C64 as well as a simple browser: http://www.contiki-os.org/ (This site will apparently create a customized image: http://contiki.cbm8bit.com/)

boomlinde• 12 years ago

There used to be a quite persistent website hosted from a C64 running Contiki. The pages were probably cached in RAM, but I still like to imagine that a beige 1541 disk drive spun up every time someone made a request.

rburhum• 12 years ago

Too many of you trying to see it. My requests are timing out every single time :(

jorgis• 12 years ago

They're still working on the Z80-based load balancer so they can bring up another server.

rbanffy2• 12 years ago

The first Unix machine I used had 2 megabytes of RAM and was connected to two dozen terminals serial. It ran on a 68020. It probably wouldn't be able to run Apache, but it ran all our accounting and managed production for a large-ish corrugated paper factory.

A Mac IIci is ridiculously powerful in comparison.

vidarh14• 12 years ago

A 68020 can run Apache just fine. A 68030 is only about 20% faster per cycle on a lot of stuff than a 68020.

Of course, with two dozen terminals active, things does change a bit...

whoopdedo• 12 years ago

The advantage the 68030 has is the PMMU. I don't think you can use virtual memory on a 68020.

My first Linux machine was a IIsi, the little-brother to the IIci. But I had to put NetBSD on it because the Linux keyboard driver was broken.

rbanffy2• 12 years ago

You forgot about the memory. Also, I don't think that specific machine had TCP/IP networking support.

ja27• 12 years ago

The first Unix machine I used had 768k or less and was a 6 or 8 MHz 68000. It's still bizarre to think that we had 12 users on a TRS-80 at once, via serial terminals. Even more bizarre, my first *nix was Microsoft Xenix.

http://en.wikipedia.org/wiki/TRS-80#Model_16.2C_Model_16B.2C...

A few years later, we had a lab running Apple A/UX on either the Mac II or Mac IIx. They were barely usable as small XTerminals at 640×480 8-bit color.

zak_mc_kracken4• 12 years ago

This thread is going to contain a lot people complaining that even though today's hardware is orders of magnitude more powerful than twenty years ago, our computers seem to be just as slow.

thejosh2• 12 years ago

Well it is slow because programs have been built to use all of our processing power. Like flash.

bane13• 12 years ago

Well, we got rid of flash and replaced it with Javascript and HTML5 and whatever, so your theory holds.

vidarh14• 12 years ago

Unnecessary abstractions and expensive context switches are my favourite pet peeves.

bane13• 12 years ago

And we don't really seem to be doing all that much more with them computationally.

flomo3• 12 years ago

I worked in a college lab full of Mac IIcis, and they felt ridiculously slow even back then. :) I think people's memories get funny and they forget that it took minutes to launch programs and you could even watch the windows redraw.

kibibu• 12 years ago

Having worked in audio and video editing on Pentium-era machines, I don't think modern machines are slow at all.

wiredfool2• 12 years ago

I remember setting up a IIci with a fast scsi drive and system 6. It booted in about 5 seconds. I don't think anything I've seen has come close to that fast until the era of SSDs.

Later, when I got my first DSL line, I had a different cast off IIci running netbsd doing NAT for my little network. That would have been 1997 or so, 640k symmetric, I think the IIci had 24 megs of memory or so. Old and slow at the time, but less powerful than this one.

Bud• 12 years ago

In 1997 or so, I was running Apache under NetBSD on an SE/30. Ran quite well and was very stable.

mixmastamyk• 12 years ago

Interesting!

I ran my first webserver on a Mac IIci at work, circa '94. We didn't have a firewall and were right on the internet. I soon was able to apple? script up a page with a live QuickCam image, text-box, cgi-bin, and the speech synth and let my friends have at it!

I received nothing but innuendo and sheep noises, but it was a blast. This was a year or so before I started administering the definitive Quake server on the corporate network, and dare I say... internet. Up to twenty players at a time, some on dialup.

Ah the good ol' days, hehe. ;)

An old Macintosh IIci 25Mhz running Apache under NetBSD

Related Stories