Boston Linux & Unix (BLU) Home | Calendar | Mail Lists | List Archives | Desktop SIG | Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings
Linux Cafe | Meeting Notes | Linux Links | Bling | About BLU

BLU Discuss list archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

named



On Nov 30, 2010, at 2:30 PM, Derek Martin wrote:
> 
> False.  Typically, ISP's DNS servers get a lot more traffic than one
> you run yourself at home, so cached answers for sites that you visit
> infrequently tend to stay cached, whereas your home server may need to
> do a full look-up every time you visit.  As a result, for infrequently
> used sites, your ISP's DNS servers will very often be faster.  It's

Faster by a whole half a second, give or take, for the first lookup only.  Then the local lookup will be about an order of magnitude faster than hitting the ISP due to network latency.  And that's the counter to your points.  The first lookup will be slower by about a half second, give or take, assuming that your ISP has a cached RR and you do not, and then every lookup thereafter until the RR expires will be an order of magnitude faster because your local network has lower latency than going out to your ISP.  Going to the ISP for every lookup is a big, fat lose.

Going through a dial-up pipe for the first lookup won't matter much since you still have the same -- effectively; it's milliseconds, maybe microseconds, on the other side of the dial-up link -- latency be it home to ISP or home to root.  Same bottleneck.  Once you have the RR cached locally then it is that much better since you will not have to make that expensive lookup until the RR expires.  You'll use the local cache instead.  Again, going to the ISP is a big, fat lose.

Your example of DNS with low TTL as an excuse for HA or load balancing is actually an argument in my favor.  ISPs ignore the TTLs on RRs.  I've seen some of them rewrite 24-hour TTLs with week-long and even month-long TTLs before handing them down to clients.  You end up losing any possibility that the failover/balancing will actually work if your ISP is like that.  Yet again, going to the ISP is a big, fat lose.  Then again, relying on DNS for your HA and load balancing is also a big, fat lose.

Which brings me back around to the point: the only good reason not to run your own local resolver cache is because you are unwilling or unable to do it.

--Rich P.









BLU is a member of BostonUserGroups
BLU is a member of BostonUserGroups
We also thank MIT for the use of their facilities.

Valid HTML 4.01! Valid CSS!



Boston Linux & Unix / webmaster@blu.org