Hosting with eApp

 

Journal

 

 Performance is a bit worse than the first day of operations with MeshCMS. The average is over 391 ms, with a low of 198 ms. Something strange is happening, and I don't really know why.

 

 

 17:26:52 up 62 days,  3:28,  1 user,  load average: 0.02, 0.13, 0.08

 

2008-08-25

Responsetime is running at 130 ms.  Not the fastest site around, but acceptable.

2008-08-25 PART 2

Swithing had a dramatic effect, the average response time dropped from 883 ms to 199 ms. There's still some sluggisj behaviour though. A test with Gomez instant test  reveled that the Validated tag was really slow, but I was also very curous about the extended times spent in DNS lookup. I'll have to look into that.

 

2008-08-25

 Decided to run the site on Tomcat by default. I don't see the huge benefit of joomla. As of today the current average response time is 883 ms running apache/joomla/mysql. Let's see how this improves in the comming days.

 

2008-08-24

I had a strange incident today, my site suddenly stopped respondin and when i logged in i discovered that my entire /opt directory was gone. I reappeared shortly again but I had to restart tomcat before everything went back to normal

Note-to-self:

1. Write a cron-script which does a route check of service quality and availability, then attempt an automated restart when things doesn't work OK. 
2. In this incident apache was ok, so it might be possible to redirect the request to a standby-tomcat?

 

2008-08-18

The eternal search for a text-editor. 

2008-08-18

I decided to take another look at MeshCMS, the benefit is speed and a light architecture making it more likely that I'll modify it when I need to. I spend a few hours during the weekend to make templates. Although I'm not kompletely done yet, it was fairly simple since the system relies on taglibs.

2008-08-02

Joomla is stable and is by most metrics the most popular open source CMS around, but I had a look around for alternative platforms anyway.

I focused my attention on Java-based open source CMS software because they can be made to be much much faster than php or python based software, while retaining flexibility. Backend UI was a key evaluation metric, because inititally writing a servlet to pump out some content isn't really that hard.

Magnolia CMS looked very promising, requiring a bit more work to get going. The huge showstopper was that the DB is easily corrupted causing the entire instance to be completely beyond rescue. There's some trule excellent stuff in that architecture, so I hope they eventually solve some of their roboustness issues and ran nicely on my eApps account.

MeshCMS,  a very nice little CMS with basic features. It can be considered an alternative to offline CMS's such as CityDesk CMS. The author says he doesn't really focus on it anymore due to other priorities, but what's there is pretty solid code.

OpenCMS, wow! this  contains a lot of features. I didn't even start to get into it because the UI is one of my key evaluation criterias. It installed nicely on my eApps account.

Apache Lenya,  lots of nice infrastructure here. Not exactly what I was looking for.

Liferay,  wow, my eApps account was underpowered for this one. Truly performed like a dog and hardly ran at all.

In progress:

jLibrary,.. under evaluation. The client looks a bit like Eclipse, which is welcome.appears to be good for metadata and files. 

 

2008-07-20

 Yesterday i discovered an issue with joomla, which prevented png images with transparency and depth of 16 to be rendered in pdf. I tried to install ImageMagick but it stopped in a scheduled status. After a while I sendt a support request, wich solved the issue.The support request was issued at 09:57 PM, acknowledged and assigned at 10:40, fixed at 11:50. I'm perfectly happy with this level of support.

 The response times on my pages improved. I think enabling cacheing in joomla has had something to do with it, because the administrator interface is mindbogglingly slow at times, I'm not sure if going with joomla was a good idea at all, but I'll stick with it anyway. Average responstime on the front page is now 523 ms, which is a bit slow. Most of this is joomla, I have apache jakarta running on port 8080 and the response time is 194 ms.

 

 

 

2008-07-18 

Average responsetime for today is currently 786 ms. I'm really not sure why, because the CPU is at 1%, but everything is running on a virtual machine which might be a factor. All in all I'm still very excited about the offering, and at $20/month which for all practical purposes is allmost like a dedicated server. It appears that the eApps access is slow.

2008-07-17

I started a test with Site24x7.com every 5 minutes to check the availability of my site. And for the past 27 hours the site has been up 100%, however the responsetime is a bit slow. The maximum time has been 4 seconds, and the average is running at 662 ms. My homepage is currently running joomla and jakarta, but the average load is only 1%.  I had hoped things to be slightly faster. (The page which is downloaded for each test is currently 12,35 Kb)

2008-07-15

Initial contact with eApps was awesome. Their web-pages are informative and provided allmost all information I wanted, but I did decide to give their callcenter a ring before making a final decision to buy. I got a call back shortly after purchasing with a followup-call to see if I needed more help. I've worked putting together call center operations before, and this is exactly the experience I'm  aiming for when designing customer care operations. A lot of companies dread the additional cost of providing a follow-up call, but it's well worth it as it protects the reputation of the company and increases the odd of a longer business relationship. In addition, if things should go wrong when getting started it being pro-active increases tolerance.