Websites and Database Report - June 09

Computerbank's Plone Website

Shortly after our last meeting I noticed worrying signs from our Plone website hosts High Speed Rails. HSR for many years an excellent company began not answering emails or support tickets and their payment system is faulty.

Googling the HSR owners blog, he seems to have joined an anti voting, anti establishment, pro gun movement and moved from Seattle to New Hampshire. Although Computerbank's website was and still is (June) working on the HSR server I decided after reading this blog and getting no replies to my emails it was time to look for a new host.

On 21st April I joined Webfaction a UK based Plone hosting organisation with servers in Texas and began to progressively copy and test my 9 websites from HSR to the new hosts.  Then I started pointing 15 domain names to the new servers.  All went well at first.

Plone on our new hosts Webfaction has a problem with growing Ram.  In 12 hours it goes from 118Mb to 180Mb, our allocation is 200Mb.  So I have been packing the database and restarting Zope once or twice a day.  Still investigating the causes.  The difference in configuration is Webfaction has a older version of Zope 2.10.5 but a newer version of Plone 3.1.7. compared to Zope 2.10.6, Plone 3.1.5.1 on HSR. 

Yesterday after a request from Kylie for a read only login to an Intranet (for an overseas group wanting to set up a similar organistion to us) I discovered that there is some inconsistency in the permissions area.

Using a test instance to check these problems.

Glom Database

I had a discussion about our slow Glom database on 28/5 with Murray Cumming, the German based developer of Glom.  Murray was very surprised about the slow speed of our searches. 

After this I decided to take a second look at Glom using our fastest box  - 3Ghz dual core duo with 2Gg Ram (thanks Michael G). Our test instance with 60 names in the database now takes 4 seconds from the time the 'find' button is pressed until the record is retrieved.  Although this is slow this is three times as fast as earlier experiments.  I suspect it is the 'interface' not the 'database' that is at fault.  If this is true it means with more records it shouldn't run any slower.

I've sent Patrick a backup copy of our existing Rhadam database and he is trialling importing this into an uncustomised Glom.  We should have some feedback about this shortly.

~ Jan

With Glom you can design the database user interface the backend is PostgreSQL.