Moore's Law, a competitive industry, or merely insanity?
In 2001 I got Verizon DSL. For $50/month I had a 768/128Kbit connection. Neato! Then prices started to come down, speeds went up and by the time I moved house in 2005 I had something like 3M/768 for $40 (memory is hazy, but it was something like that). Today I have FIOS which is 20M/5M and as part of the triple play costs roughly the same.

In 2004 I got a linode. For $20 a month (paid a year in advance) I got 64Mbyte of RAM and 4.5Gbyte of disk (and some network transfer I don't remember). Today (literally; they just increased their service) the same service is 360Mb RAM and 15Gbyte of disk and 200Gb of network transfer (lots more than before).

I don't know how competetive the VPS market is; not done any real research. I know Panix V-Colo try to remain competetive with linode. They seem to follow, so it wouldn't surprise me at all if Panix increase their service levels to match in the near future. I also have a v-colo and my websites/mail configs etc are replicated; redudancy! Some of Panix's configuration options are different to linode's and may, in some cases, work out better value for money. I'm cheap and stick with the base model from both providers.

Now I can see some of these increases coming about through changes in technology (especially Verizon switching from copper to fiber). But others like linode? Sure, some of the changes will be due to newer hardware (bigger disks, more memory)... but a factor of 6 memory increase over 3 years? Wow.

Anyway, the extra linode memory was a nice Christmas present.

Of course I built my server config back in the 64Mb days. According to "free", after 30 minutes of uptime my linode says I'm using 22Mb of RAM (the rest is cache). The v-colo (up 7 days) says 96Mb of RAM used. The v-colo is Xen based running Centos 5; linode is UML based running Centos 4; both fully patched... interestingly the v-colo memory usage always seems higher.

I know peak load is higher, especially when indexing the Panix news groups for the search engine I've built :-)

Heh, just realised the uptimes look real low. It's an anomaly. Had to reboot the linode to get the extra memory. I updated my v-colo to CentOS 5.1 last week so rebooted for sanity. Both services have been very stable.

Maybe I should look at building something more than just a plain lighttpd configuration. Perhaps a full LAMP setup!