i’d been with cloud web, a division of eicomm (who also hosts anandtech) for about a year, and they’d been mostly great for that whole period of time. however, when i was updating my payment card, i realized that not only was i paying for my site hosting with cloudweb, but i was also paying for a vps (virtual private server) that i ran teamspeak and some other services on as well.  i hadn’t self-hosted or vps-hosted any of my own content since right around 2008. i got sick of feeling like i had to micromanage and tune every bit of my environment to ensure it was fast, fit in a relatively small memory space (512MB in that case) and i was also sick of paying $30/mo for the privilege of driving myself nuts. at that point, i decided to move to shared hosting, as it was significantly cheaper, reasonably secure (assuming you’re using a decent provider, of course).

this got me thinking – i’m paying for shared hosting and i’m paying for a perfectly capable vps that isn’t really even being fully utilized. why not just migrate all my content over to the VPS, configure a web server and stuff and just consolidate all that stuff in one place?

so, one night at work, i got really bored and archived my content at cloud web, wget’d the archive, set up apache, mysql and php on the VPS, hardened everything and deployed the archive. i was feeling pretty pleased with myself, but someone in IRC mentioned messing with NGINX (pronounced engine-x) on his VM and was really impressed with it. so i read up on it a bit, liked what i read and saw and decided to try it out. i’d never tried it before, so the configuration conventions were a bit different, but made more sense than the XML-style configs used by apache, which i’d had years of experience with.

apache example:

#virtual host, listening on any interface, port 80
ServerAdmin foo@bar.com #contact address for the operator
ServerName bar.com #server’s hostname
ServerAlias www.bar.com #alias of the server’s hostname
DocumentRoot /var/www/bar.com/htdocs #where apache can find the site content
ErrorLog /var/log/apache2/error.log #where apache logs errors
CustomLog /var/www/bar.com/logs/access.log combined #where/how apache logs access to the site

and the same sort of config with NGINX:

server {
server_name bar.com www.bar.com;
access_log /var/www/bar.com/logs/access.log;
error_log /var/nginx/error.log;
root /var/www/bar.com/htdocs;
listen 80; #listen on port 80 for http sessions
listen [::]:80; #if you wanna be fancy and also listen on available ipv6 interfaces

location / {
index index.html index.htm; #look for these files as the index for this path – apache does by default.
}
}

for me, NGINX’s configs just are clearer, especially when you take time to space them properly (which i’m not going to bother doing here) and annotating any odd or obscure rewrite rules (things that match stuff you specify and change it to something else. similar to mod_rewrite for apache) and couple the clarity of configuration with how much faster NGINX is than apache 2.2, i’ve just been blown away.

so, back to the hosting thing – last year, i’d migrated my DNS, service monitoring and VPS hosting to Zerigo. Zerigo was promptly acquired by a company called 8×8, who have subsequently sort of run Zerigo into the ground. support requests (for things like hideous network conditions, false positive ‘down’ alerts from watchdog, etc) are either answered really well after they’re no longer an issue or never answered at all.

it bummed me out, but i ended up having to set up a VPS with a different provider, in this case, bitcable whose servers are in ashburn, va, set up my environment and then SCP’d all the content over to the new server. a month in, and network conditions are generally much more stable and solid.

the point on the graph where it goes from peak-y and jittery to smooth is where the dns changes propagated for the change from Zerigo in Santa Clara, CA to Bitcable in Ashburn, VA as measured by a VM at my house on Comcast Business.

the black bar indicates the time after I made the changes to the DNS records for m0unds.net, to point them to superb. you’ll likely notice something…strange. yes, that says 4,093 milliseconds. that 4 seconds for a flat file with a couple kb of text and a single image to load. compare this to the current host (smaller numbers) which averaged 200ms (1/5th of one second) to load the same content.

grid/cloud stuff is interesting in part because storage is generally not actually part of the system that reads from it. so, despite the fact that the webservers may very well be extremely fast, if i/o is slow, the hosted sites will be slow. i submitted a ticket to ask tech support about the issue, and i’m anxious to see what they tell me. it’s hard to explain away something as a fluke when there are hard numbers backing it up.

*edit* – so it looks like superb’s awesome support folks strike again. after submitting my ticket with the screenshots from pingdom, including an hourly breakdown, one of their systems engineers was assigned the ticket. a few hours later, i received a maintenance email from superb’s NOC, and 5 minutes of downtime, followed by another email from the engineer indicating they isolated and corrected the problem system-wide. sure enough:

the black line indicates when the configuration change occurred.

since november, we’ve been waiting to hear anything from the legal representation retained by the title insurance company on our behalf. there was a hearing in our favor on 3/2, a court order issued on 3/16, which took effect on 3/31. so, bearing that in mind, the house is actually ours now, which is a good thing. not so good, though is that there could be an action in another court that could call the court order into question. however, this does mean that we can finally finish painting and continue about the business of making the house our home.

at any rate, i’ve decided to move all my photos over to flickr. this does a couple things- first, it moves my photos to a highly reliable and available platform separate from my hosting provider’s servers and second, i don’t have to a goddamned thing except get everything situated and organized and just let it go. a bonus is that it also affords some interesting opportunities for social interaction and discussion with like-minded individuals and some other ways for me to visualize statistics and other things that i’m really fond of, namely geotagging photos, seeing who sees photos, which ones are the most popular, etc.

it’s also been a pretty long time (again) since i used this space for anything but stupid photos simulcast from posterous, so i guess it’s giving me the opportunity to jot something down. so, uhh, yeah. work’s work, encroaching on holidays again, etc. the year’s gone by too fast, which is really starting to force me to realize how old i’m getting. i’ll be 27 next year, 3 years from 30, which is a strange feeling. slowly losing the sense of invincibility that i’d felt during my teens and early 20s. i’m sure i’ll feel even less invincible as i get closer to 40, so i should probably just shut up.

last year, in may or so, i moved from using a linode vps for my mail and domain hosting, to using hosted exchange for mail and a shared host for the sites. the shared host for the sites has been perfect. i’ve had 99.999% uptime, the only time the sites have been down is during maintenance. this is better than my own uptime with my vps. (i should say that wasn’t any fault of linode who are totally awesome and if you’re in need of a vps, you should look them up.)

why do i pay for mail? there are a few reasons.

1. i don’t trust people with my data. companies that offer enterprise-grade email and communications platforms have a lot more incentive to keep shit private and protect your stuff.

2. i like uptime, i like reliability and i like being able to contact the host to have stuff resolved if something happens, rather than getting to lose mail or something.

3. i like supporting businesses who offer competent, well-engineered, reliable services.

i started off with the hosted exchange company, sherweb, who when i signed up were offering a competitive blend of services and price that was too good to pass up. i really like exchange and i really like outlook and owa. they’re extremely well-integrated, well-tested, reliable platforms for mta use, pc clients and web clients respectively. after having service with sherweb for around 4 months, i started noticing that it seemed like they had downtime two or three times a week. of course, this isn’t a huge deal, but the loss of performance and issues with their (terrible) barracuda hardware firewall/spam appliance made me want to try something else. i read some reviews and against my own better judgement (i despise google) i signed up for google apps premier.

google does lots of stuff i don’t agree with and lots of things that make me really nervous. they don’t do any one thing terribly well. their search algorithms aren’t all that great. their maps platform hasn’t really evolved since it was first released. gmail offers free mail at the expense of letting google rifle through your personal stuff. their metrics allow them to track user site visits and stuff, which is interesting in that you can see who sees your site, but you’re also doing a big disservice to your users by allowing google to follow them around. they’re just a creepy, orwellian company who make money from your personal information and other net-metrics.

getting back to signing up for google apps, i don’t know why i did it. the thought of having what i felt was a somewhat reliable infrastructure keeping my mail flowing and my information safe (their apps privacy policy at the time was very strict and decent, all things considered.) i ran into numerous quirks and broken features and google had their instances of downtime that people seem to have absolutely no problem with. at any rate, the price was right, it was $50/year for mail, contacts, calendar access, documents and it’d sync with my pre. this in itself was significant, because it saved me approximately $60/year over my current hosted exchange provider. i purchased my service, waited for it to be provisioned, then moved all my mail over, set mx records, waited a week, then was ready to roll. i had a total of 12 incidents where i couldn’t access my mail for 4-5 hours at a time. a few of those were reported on as they affected their regular gmail services as well as the premier apps service.

for the most part, their service was alright. i hate gmail’s interface. their calendar support is alright, their docs support is actually pretty good. other than the aforementioned downtime, the system was pretty reliable. fast forward to february, this year. last month, i’d been reading some pretty screwy/scary statements made by folks at google that sort of confirmed my opinion of them re: data security/information as a commodity. it made me nervous, but i still had just a few months left until my apps acct would lapse and i could move my domain somewhere else. one night, i was bored at work and read a bunch of stuff that made me wonder about google’s apps privacy policy, so i went to read it. i wish i hadn’t.

i immediately began looking at my options as i only had a few months left on my account and decided that i’d just go back to another exchange provider. i settled on one, set up an account and was ready to go.  except, the new company didn’t process my order. i had to call the following morning, and spoke to someone who finished processing my order for me. thinking that was that, i logged into the control panel, and saw that postini had hung on provisioning and my account wasn’t ready. i called the provider again who explained that the postini provisioning api had stalled because my existing domain had postini service. it didn’t, so i thanked the guy and went to research it a bit. i learned a few really wonderful things about the google apps platform. you can’t have google apps premium and separate postini service. there’s no reason it shouldn’t work aside from google either being too fucking lazy to figure out a way or just flat-out not wanting users to be able to do it.  at any rate, i researched a bit more and found hundreds of threads on google’s forums and on other forums stating that they couldn’t move from premier to free apps without deleting their domain and having mail service downtime because google has no way to migrate between services. i also found out that when you disable service and delete your domain, it takes them 5 days to delete it. well, that’s what they say on their site, but from what i read, it can take 6-8 weeks in some cases, even with constant contact with google support, for them to finally release your domain. only then, can you add additional services or make changes to things. they also make it impossible to change your domain once your account is provisioned. also, they make it impossible to change your username without using their provisioning api. is it really that hard to add a ui element to let users change their address?

after doing some research and submitting a ticket to google to have postini deleted from my account so i could provision my new mail service, i logged in to get my support PIN and noticed i had no mail. zero mail. i checked docs, that too was empty. looked at my calendar. blank as well. well fuck. at this point, i’m livid. i have my mail backed up in preparation for moving to a new host, so it’s not the fact that i lost my mail that made me mad. it’s that there’s no recourse for it. according to their phone system, it’s not an “emergency” because i have less than 10 users.  at this point, i’m so angry i just contact the new host to see if i could get a temporary MX record to just get the fuck off google apps til i can get them to remove my services. of course, since my luck is so awesome, they won’t do it. i started scouring the interbutts for another hosted exchange provider who might give me a temporary MX record and happened upon the message center. they said they had 24/7 support, so i figured i’d give them a shot. i called their support number and spoke to someone who gave me some suggestions for removing postini/apps stuff from my domain and gave me their phone # on file for google support (which isn’t published on google’s site, btw). the dude also told me that their provisioning system would give me a temporary MX to use until i get squared away with the google bullshit. great.

so i got my mail set up, migrated over, changed my MX records and now i’m waiting til tomorrow when i can actually (hopefully) speak to someone at google and get them to remove their broken fucking apps service and other bullshit from my domain so i can get provisioned with postini service and return to ignoring their existence.

google pledges enterprise reliability, but doesn’t actually have 24/7 support, nor do they actually answer email tickets. i found numerous threads where people hadn’t heard from google in 6-8 WEEKS since submitting a ticket, their issue was still unresolved (and in some cases, totally unacceptable) and yet people still continue using their services. i don’t get it. if they treat their paying customers like shit, what do they do for the folks who use their “free” stuff? gross.