Do you think that I can’t code an application like Varnish? Trust me, if I wanted, I could — coding is easy for me, and I can pick up a programming language in hours”

Most programmers worth their salt can pick up new programming languages in hours. My guess is that you are somewhere in your 20s and have programmed since you were a teenager. I’ve been in your position and thougth that I could make everything. I even had plans to make a Word-killer, because I thought I could do it better. When I got older, I realized that making things better is hard and was glad that I didn’t start with my Word-killer

I also took the opportunity to move all my cookies into a sub-path of the site where they were needed, instead of setting them in the default path of /.

As far as I can tell, many intermediate proxies and even client-side caches won’t cache something with cookies - confirm this by reference to http://www.mnot.net/cache_docs/ and http://www.mnot.net/cacheability/. So it’s at least arguable that varnish’s behaviour out of the box with respect to cookies is acceptable, within spec.

Having said that, I came to read this storm-in-a-teacup because I’ve noticed (I think) that ETag varnish behaviour doesn’t seem to be as I’d expect. Varnish doesn’t seem to be forwarding my ETags to the client, which may be a configuration problem of mine, or may be a lack in varnish.

If it’s true that Varnish doesn’t allow you to make decisions on server-generated fields, that’s something that’d be nice to have. It’d be nice, for example, to be able to use the ETag as varnish’s hash value, can’t see why not. Some of the other things Rudd-O suggests might also be useful.

I don’t think either side are kooks, I can see there’s a difference in perspective and priorities. Suggest that Rudd-O might have considered contributing patches … nothing speaks louder than patches.

Server is new Dual Quad-Core Intel 2.66ghz, 8gb RAM, 4 x 250gb SATA drives RAID 10.

Three runs, take the best response time for demo.html page that is 38535 bytes in size. Both Varnish & Squid cache set to 1gb in size, though that should matter with just this single page test.

ab -n 10000 -c 100 -d http://localhost/demo.html

Varnish + Lighttpd = 9470 requests per second Squid + Lighttpd = 3712 requests per second Lighttpd alone - 11244 requests per second

Squid + Lighttpd is dog slow, so I don’t see how or why I would ever use Squid. Lighttpd by itself is nearly 4x faster.

Varnish could use a better config out of the box, considering there is no caching set out of the box. I’ll keep playing with both Squid & Varnish before deciding which one to deploy in production, but so far Varnish is waaay faster than Squid in my benchmarks. If everything else turns out fine, it will be Varnish all the way.

Edward and Spenser: what’s with the ad hominems? Do you think that I can’t code an application like Varnish? Trust me, if I wanted, I could — coding is easy for me, and I can pick up a programming language in hours. However, I have other things to do, and a Varnish-type app wouldn’t put food on my table.

As for Varnish beating Squid: what’s new about that? Did I ever even suggest Squid was faster? No. However, I’m going to suggest to the Big Fool that you are (because apparently you can’t interpret what you read) to R-RTFA and STFU.

http://www.o3magazine.com/pastissues/issue6/

Thanks

I believe you need to look at it again.

And maybe your lack of technical skillz can grow some…

Sorry - but you realy suck! Just cuz you ain’t got technical foo your nagging on other ppl Thats the way to go right…

Anyways, Varnish 1.1.1 is out - and I believe your issue is fixed “out of the box”. Meaning all that you say in your post about “they don’t listen to users” is pointing you out like the Big Fool - ‘that you are’.

Or Im I wrong too?

I am not impressed.

Whereas, I am impressed by the architecture notes at the varnish site.

You seem to care more about whining than reading docs.

Hey, you must be one of those web 2-oh-oh people.

I never said Varnish was bad. If you re-read the article, you’ll see that I said it’s an amazing piece of software. It doesn’t help if you frame the discussion in a surreptitious fashion to make it look like I’m knocking the devs because Varnish is a bad piece of software. That is not true.

What is true, however, is the poor response I got from the devs, and the lousy default configuration which leads to slowdowns instead of speedups. Caches are supposed to cache static content and deliver fresh dynamic content. By default, Varnish doesn’t do that — it blindly caches everything unless cookies are present, in which case it doesn’t cache anything.

You tell me how that’s right.

First of all, the bad initial experience was a rightful one. However, the problem is not the software but the default VCL, their default/sample configuration. Just write your own VCL to bypass the if-client-has-cookies-don’t-cache state. They must have their reasons to implement this rule as a default but there is no reason you can’t override it as we have done.

Varnish is not without faults but is an extremely polished and scalable software for its age. Quite frankly, I’m amazed at the quality of the work thus far.

To knock on the developers for something that’s not a problem is beyond me.

I’m not complaining that people aren’t taking me seriously. Where do you see that? Please point to a paragraph where I complain about that.

I didn’t misunderstand what Varnish was supposed to do. It’s an HTTP accelerator, it states so on their front page, and yet it doesn’t accelerate sites in its default configuration. Fortunately I now have found (thanks to the fabulous help from ) a couple of VCL hacks that I can apply to actually improve Varnish behavior.

I contacted people through all the right channels. Bug reports and mailing lists. What other channels are there?

I gotta love how you misrepresented my entire post in just one sentence.