Apache 2.4 Comes Out As Web Server On Steroids

[phoronix]

Phoronix: Apache 2.4 Comes Out As Web Server On Steroids

The Apache Software Foundation officially released the Apache 2.4 today as the first major update to this leading open-source web-server in more than a half-decade. Apache 2.4 is slated to deliver superior performance to its 2.2 predecessor and better compete with the growingly-popular NGINX web-server...

http://www.phoronix.com/vr.php?view=MTA2MDQ

[Azpegath]

Release early, release often? ;-)

[oibaf]

Besides much faster performance...

Where do you read that? There is very little about performance in the release notes.

[jabl]

Where do you read that? There is very little about performance in the release notes.

Here, for instance: https://people.apache.org/~jim/preso...ttpd_cloud.pdf

[cwells]

Apache Foundation avoids real benchmarks in favor of marketing

Here, for instance: https://people.apache.org/~jim/preso...ttpd_cloud.pdf

That benchmark is as uninteresting as it is misleading. Problems I note right away:

1) No tuning was done. If this is absolutely true, then we must assume Nginx was performing with only 1 worker (the default). That means Nginx only utilized a single core and kept up with Apache using multiple cores. If it's not true, then we need to know what tuning was done for each server.

2) Without a graph of CPU and memory utilization, this is a very one-dimensional benchmark. It *does* matter if Apache is using 1GB of RAM and 90% of CPU and Nginx is using only 20MB and 20% CPU (making up numbers, since the benchmark provided none).

3) Under "Caveats", it's noted that "Apache is never resource-starved". Of course, this is the exact scenario that people considering Apache vs Nginx should be considering and that the benchmark skips over altogether (providing graphs for (1) would have given some clue about this). Apache (like most software) works fine when it has tons of resources to consume. The reason people have been moving to Nginx isn't because Apache is slow, it's because Apache tends to consume lots of RAM, and once it's done that, "thundering herd syndrome" sets in and brings the entire thing to its knees in short order. The fact that *this* issue is the primary drive to Nginx (and other lightweight HTTP servers) and the benchmark specifically avoids it, I can't help but think that the biggest change in Apache 2.4 is the number.

I'd really like to see an independent 3rd party (hint-hint, Phoronix) provide more useful metrics. What the Apache Foundation has been spreading appears to me to be mostly marketing materials. I'd be shocked if they didn't have real benchmarks that arose during the development of 2.4, but I've yet to see them. That's doesn't seem reassuring.