djbdns and multiple IP's

As off currently, i am looking for a way to switch from bind to djbdns, from exim to qMail, but one thing that is holding me back is this:

I have 5 resellers on my box, and they all have 2 ips each for their own nameservers.

As i know DJBDNS needs one process per IP to be run, so this would mean that there are 5 x 2 = 10 processess running that are just for resellers, and then 2 more just for my main dns servers for shared hosting.

Wouldnt this be a little bit innefficient? Wouldnt it make it slower? Or is this a good thing?

The good thing i can see from it is that there are 5 different data files, meaning you can query the DNS server, but only get replies for the domains that, the reseller has.

Thats all i could see as being good, for the rest wouldnt it be resource wasting to have 10 different processes running instead of 2 that support all?

X-Istence

There's no real harm in this. Just make sure you have enough RAM spare; that way, your operating system will store the data.cdb files in memory if they're accessed frequently enough, thus eliminating the problem of having to answer DNS queries from 5 separate files on disk (if all your resellers had to constantly answer queries at the same time).

Yep.

It depends on what kind of resources your server has (CPU, RAM, etc). It would be cleaner to have 10 processes rather than 2 with some kludgery to make it only answer queries for some domains based on other factors. Linux can support hundreds, maybe thousands, of processes without any problems, so you're not likely to run into issues here.

Use top, free and ps to measure resource usage on your system if you have performance problems.

The target boxes are:

700 Mhz AMD Duron
256 MB Sdram
Running:
Apache, MySQL, qMail, and djbdns.
Low resource usage websites, with 2 resellers

2 Ghz
2 GB ram
Running:
Apache, MySQL, qMail, djbdns
High resource usage sites. This is the box with 5 resellers

2.4 Ghz dual pentium
4 GB ram
Apache, MySQL, PostGreSQL, qMail, and djbdns
High resource usage sites, with 10 resellers, and around 50 shared hosting sites.

Currently all these servers use BIND and Exim, taking Exim off the 2.4 Dual really helped the load, as it went down instantly. Ram usage was also less.

So do you think this would be okay?

The machine specifications are more than adequate, though it depends on the size of the data.cdb file and what else is competing for RAM on the machines as to whether the caching will be effective or not.

The best way to find out is just to monitor the boxes with top and sysstat; look for excessive (>75%) CPU usage, and disk I/O bottlenecks (using sysstat).

Depending on how much mail/HTTP traffic you handle, you can realise significant performance increases by mounting your log partition on a separate physical drive. This means that log files can be written at the same time as other files are read/written - else you may find that a lot of I/O is tied up writing log file entries (which will happen a lot on a busy system, since each HTTP request results in one log file line being written, and each mail message that is received and processed results in around 15 lines being written).

Log files are allready being written to a different drive, just for that reason.

I monitored one box like you said, the one with the 10 resellers, load went down from when i was using bind.

Per reseller, there are about 10 - 15 accounts avg.

So far i am liking it, more resources are open for use on that box. And load is down.

Thanks alexgreg.