Smooth upgrasde to version 2 complete

[IPB Corporation]

Just a quick update, I completed the first automated update of my new
Bind 9 DNS server replacement code named "DNA" last night.

As designed it automatically upgraded users operating GlobalDNA
servers. Tier 2 & Tier 3 providers please note your now on Rev. 2

This was the first test of the completely automated upgrade system,
which is based on the same code I wrote for toolbar upgrades.

The only bug reported was with an institution who uses a hash code
based security system, the Chang set off a virus alarm which resulted
in them restoring the older code. At midnight tonight
That code will once again check and upgrade itself again. But this
time they will upgrade there hash table.

I am are looking into fixes for this in Rev. 3

The big change on Rev. 2 is I changed variable T1 from a fixed value
to a variable value set by the upstream DNA root.

There are two variables that control local cache expiration
T1 & T2

T1 use to be hard coded at 10 min. And T2 at 1 min.

When a local DNA server received a request for a DNS record that was
not internal it would poll the upstream DNA server.

Once it received the record it would place it in the cache. And both
T1 and T2 would begin counting down to zero, then every time a request
came in for that DNS record T2 was reset to 1 min again so as long as
the record got traffic it was retained in cache. T1 was used to keep
the cache from becoming stale
Regardless of the traffic the cache would refresh in 10 min.

Now when the upstream DNA server passes a DNS record down
It does so while passing the T1 variable. The new T1 variable
Is set by the local DNA passing the record based on its load, and Now
ranges from 5 Min to 15 Min.

This results in the upstream server being able to balance load, by
shedding redundant refreshes during peak traffic.

So far it has worked great!! In distributing the load over limited
resources.

In Rev. 3 I plan to set T2 to 5 min and add a new sub routine
Which checks the percentage of free memory in the cache anytime it
drops below 25% it will cycle through the records in the cache
reducing there T2 counters, in effect cycling the inactive stuff out
early.

I am also considering making the local Speaker Buzz while this
Is happening as an audible alarm.

Note: Currently if the cache reaches 100% it is flushed and begins
building all over again, so for the speed of your users
It pays to have plenty of RAM allocated to the cache! I recommend at
least 256 meg.

Important Note!! There are Tier 3 testers out there running DNA with
64 meg caches, I am considering limiting
The server from running if it has less then 128 meg allocated in
future releases..