Poor Dreamhost performance

I really wish I’d read this a couple of days ago… I’ve been getting the dreaded CPU minutes mails too over the past week.

I tried just about everything I could think off, but even with all plugins turned off and using the default template a single post page takes between one and two seconds to display. No idea to how much CPU time that translates to, but if I do the math (divide CPU time use by number of pages served, I end up with about the same numbers).

So, long story short:

my CPU minute usage for the last few days was as follows:
110.98, 115.55,127.01,135.10,112.77,156.55 — and now I’ve been shunted off to an even slower server, with no guarantees whatsoever.

I’m basically in the dog house for one or two weeks (no specifics) after which they’ll re-evaluate.

If anyone has any feedback from Dreamhost (basically if they admit to anyone that there servers can’t cope) — please share. ??

And if anyone has any pointers to PEAR benchmarking for dummies or performance optimizing tips…

If those of you who are having problems don’t mind, what kind of traffic are you getting a day? I am about to switch hosts and DH seemed to be the answer. But not if I am to run into CPU cycle BS.

So, I too am now concerned. While I’m still within my MBG period, and BEFORE I migrate, are there folks who’d care to comment on the downtime/issues of some of the other successful hosts (thinking midphase, bluehost, site5, to name a few)? I trust that ASO does well, but is also more pricey than I want at the moment. Yahoo is out, as they are single-domain oriented (at least from everything I read). I need at least a handful of domains, plus reasonable subdomain support (not for extra sites, but for ‘category’ SEO).

-d

I have around 2000 unique visitors a day.

I run 6 WP installs at DH and between them all get somewhere around 10,000 page views a day.

Yahoo is only one domain? Yuck! Guess that’s not an option for me.

I’d suggest lots of blog posts about how appalling DH are and how they seem to be exploiting the link here in their pursuit of profit.

It sounds to me like all shared hosting services use CPU cycles in order to ensure everyone uses resources appropriatly, but some (like DH) seem to rake you over the coals with it — if you cross a line, you must upgrade. Ideally, I would like a host that would work with me on streamlining code first. Thats my 2 cents.

Based on this thread and others, I am shying away from DH in favour of a smaller company.

I saw a post on WHT that someone had contacted DH, and the response basically sounded like “yeah, we turned something on, and lots more people got emailed than we thought would, so ignore it unless you hear from us again…”. Which is annoying.

Something else I hit upon: try disabling gzipping in WP if you can — that’ll increase bandwidth for your HTML, but significantly reduce the server CPU load… Hadn’t thought of that until I saw site5 doesn’t use mod_gzip, and that was one of the few gripes I saw about them (still considering sticking it out with DH vs site5 or midphase or going to a VPS somewhere… but need to do in next week!).

-d

I haven’t received one of their warning emails in almost 2 weeks, but according to the resource log I’m still using 60 minutes/day. This is quite interesting considering that my traffic doubles in the week around and after Christmas each year, but my CPU usage has stayed the same. How could this possibly be a problem with WP? Seems like DH needs to get their shit together!

I haven’t received their warning emails recently too. It is fun.

Yesterday, I tried to install MediaWiki on my dreamhost host. The configuration script gives some informations. One of them is:

PHP’s memory_limit is 50M.

Is it reasonable value?

All hosting companies limit CPU usage somehow. The bigger issue is expectations of the client. Most clients are looking for the cheapest host they can find.. The cheaper the packages then typically the more clients they will have to load on the servers to make even a minimal profit. https://www.Blogs-About.com hosting does the opposite. We do not enter the pricing war for hosting clients because quite frankly an active blog will use a fair amount of the cpu at different times which is something we understand. So instead we charge a higher price per month with the expectation that even on a dual processor server with 2gbs of ram that you can only host upto around 80 blogs.
So smaller hosting companies with higher prices tend to be far more forgiving and simply stack less clients on a box (most of the time).
I mean lets face it.. 100Gbs of bandwidth for 3.95$.. The cpu usage you will need to actually transfer 100GBs is way over what these cheap hosts are willing to give you. Its effectively a way to force an upsell. Hosts force you to upgrade your packages based on CPU usage or face termination of service or force you to move which takes days etc.

We found on our own servers that due to anti spam plugins people just left the spam in their tables which forced mysql to spike tremendously because mysql doesen’t seem to calculate necessary space for tmp tables very well and resorted to tables on disk simply because those tables had more than 50K rows of comment spam or unmoderated comments. So as a hosting company that understands blogs we wrote our own scripts that go through nightly and delete spam or unmoderated comments that are older than 21 days. There are some solutions to high CPU usage that are not always readily available to clients and whether its us or another hosting company you should always expect to get what you pay for and for 3.95$ I wouldnt expect too much.

how about to setup a MovableType on DH?
It creates static html files only.
Maybe it works, maybe.

I asked the same question to one of the DH techs (about moving to MT), it made sense: the only hit to CPU would be in creating posts, anything in the admin section, and generating the static HTML files from the content. The DH Techs said that MT actually has doubled and tripled CPU minutes for people who tried the same thing, mainly because the MT code is not written as efficiently as WP.

I’m getting the same problems with DH. I receive around 12-15k hits a day (3-6k uniques) (spiking to 27k) and was around 30 CPU minutes. That was 30 CPU minutes without posting. On days I posted I was hitting 97 minutes. That was down from 250 minutes when I had a few bad plugins running and without WP-Cache. But I also have Mint installed.

As an experiment, I did not post and did not check my Mint stats over the weekend. CPU usage dropped to 30m. I then started to post and check Mint every once in a while (it is all dynamic), and my CPU spiked to 97. I also had increased my WP-Cache timeout to 7 days (since none of my archives need to be dynamic). Something in either WP or Mint is causing a spike, and due to DH’s bad resource checking, all I know is whatever is happening is going through PHP.

I was told they are working on some tools to give better reporting on what exactly is causing the CPU drain, but that is 6-12 months away, and more like 12.

So now I am posting and not hitting Mint (and switching off) to see if I can isolate one or the other. It’s getting kind of bad, tho. I either give up Mint (if that is it), or remove more plugins from WP (which I’ve already removed a lot of, removing more would start to be a drain on the site’s features).

So, yeah, it kind of sucks.

I’m over at InMotion.net and have just started getting nasty-grams about server resources usage, and have already been bumped to Level 2 help. This really came across as an upsell to me. Considering my blog access is rather small (9000 hits a day), this really didn’t make sense. While moving to another ISP sounds like a real drag, the upsell approach really doesn’t sit well with me.

I see “asmallorange.com” mentioned above – has anyone tried this, and if so, do they handle the CPU usage thing differently?

Blogs-about,
Nice site design, but no contact information. I am interested in talking to you about future hosting, but none of your packages will work for my needs.