Man, Sometimes This Stuff Is Just Weird...

[John Dyson]

15 hours ago, mrvco said:

 

I know local DNS caching used to be a thing when you were more dependent upon your ISP for DNS.  I can imagine it still speeding things up a bit for people with sub-par ISPs and/or high-latency Internet connections.

20-50msec here or there can add up to a perceptable delay.  Back when I was developing OS code, using a local copy of BIND was the norm, and frankly, if I had multiple boxes, I'd be running BIND (or more recent caching equivalent) on one of them.  WRT Windows -- I don't know what would be best.  Having a local caching server is especially nice when more than one computer is active online.  Also, my BIND would be feeding information about my machines within my domain, but nowadays without incoming connections -- the need for supplying nameservice information to the net is specious.

 

If the name isn't in the cache, then the delay is  pretty much the same as if it was the dip directly into a non-local nameserver.  Some programs do a local cache also -- and that does help mitigate the advantage of a local nameserver.

 

Do many people have less than 20msec latency nowadays? I remember that real T1s/T3s had short delays, but these hyper fancy complex modulation schemes tend to have slightly longer latencies.

 

So, it seems like a local cache is nothing but a benefit (assuming that the cache software is written correctly, and can deal with changes -- whenever they occur.)

 

John

[John Dyson]

1 hour ago, mrvco said:

 

I ping 1.1.1.1 (8 hops, Cloudflare) at 12-15ms from my home network connection via Wifi (Comcast).  8.8.8.8 (10 hops, Google) and 75.75.75.75 (7 hops, Comcast) are each a few ms higher on average.  I'm not sure how that translates to actual DNS performance, but as long as the DNS server in question is up and working, the difference between each of these isn't perceptible of course.

If you get that fast a ping time (much faster than my 20-30msec), then it makes little (but still SOME -- transient network issues) difference to use a caching server .  Just checked dig (the interactive dns lookup program) and get 22msec, so it is a little faster than I thought.  I was thinking closer to 25-30msec.

 

Since caching servers are so simple to enable, then why not?  I'd expect that Windows probably has its own local caching, and on Linux derived systems - which is an indirect descendant of the traditional internet machines for over a decade -- all kinds of internet tools are easily available and simple ones are easy to use.

 

When doing one of the crazy WWW pages with nonsense from all over the place WRT advertisements, the delays can add up with a 20msec ping time.  I also use Comcast, but only an 'advertised 60Mbps', but 90 in reality.  Nowadays faster connections are often available, but super fast data rates dont automatically mean fast ping times.

 

So, if you get 10-20msec, then it probably isn't too bad, but at 20msec or more, I'd definitely do local caching just because it is so cheap and easy.  If it was brain surgery to put together a caching server, I'd slough it off -- but caching servers are infinitely more simple to use than Bind, even when using Bind in caching mode (well, if your vendor didn't give you a simple caching config file.)  Of course, nowadays Bind might be easier to config.  The simple caching servers are TRIVIAL to set up -- zero cost, only benefit.

 

 

John

[John Dyson]

24 minutes ago, crenca said:

 

I think you and @Jud's experience answered your own question.  With typical bandwidth and latency now an order of magnitude better than "real T1's" of your day (and mine in corp IT land), why another point of failure (even though yes, DNS and DNS caching is a relatively simple, perhaps even "trivial" IT exercise) and task being given to rock bottom quality equipment and software?  Since home networking equipment is the equivalent of those cheap earbuds and DAC/amps (made in China with parts that cost mere pennies) that most listen to their 128kbs mp3's on, why are we even discussing the equivalent of asking them to play 24/96 through them and "hear" the difference? 

I consider needing an extra external network transaction as being a point of failure or congestion.  If a caching server fails (so easy to fire-up on machines nowadays, and Windows already does the caching -- must be simple), then there are problems that would probably make using a WWW browser impractical :-).  It is just SO easy to put together a caching server if you already have some kind of server on the network, and under Linux or whatever -- it is triival.  Under WIndows, I think that one needs to disable the local caching server if you don't want it (AFAIR.)   A caching server can easily run on the same computer -- in fact, I had historically done that also -- no reason not to run it, like I said.   So, I use a local caching server on the local machine (AFAIR default on windows, and make sure that the version of Linux uses the local caching server, and then an external machine in case you reboot alot.)

* maybe there is a problem with usage of the term 'caching server', because that means any local software/hardware that can cache the DNS.  It is very common to use caching servers on Windows boxes locally?  Is that the problem?  I write servers all of the time -- but I don't need a separate box for them.

 

I truly don't understand the difficulty of running the server...  IMO, it is kind of silly to do DNS lookups for every darned WWW with advertisements.  Of course, I used to run an entire internet server on each one of my machines -- it just isnt a challenge.

 

Also, a local DNS server shouldn't be a local point of failure -- the system (properly designed) goes off and grabs another source if there is a problem.  The major time where there is a problem is when there are problems with the root servers and then problems propagate down, but I seem to remember that (roots having connectivity problems) has bene mitigated.   I'd' suspect that organizations like Comcast might even fake themselves as root servers?  The internet is still the wild west in a lot of ways.

 

John

[John Dyson]

32 minutes ago, crenca said:

 

It's the opposite.  The transaction is trivial given normal bandwidth/latency, thus making the whole WWW thing even more practical than it was 20 years ago when local DNS caching was actually needful in many situations.  

 

 

Easy does not make it necessary or even wise.  Besides, @Judlow powered, poorly programed, cheap as China home networking gear was trying to play this role - not a desktop OS.

 

 

Right, so why would it be wise for the "average" home network and non-technical user want to wade into this - with equipment that is far worse/reliable than the upstream ISP equipment, to say nothing of Google's DNS infrastructure (or the other choices available)?  

 

Barring the oddity in ISP service, 99.99999999% of home users don't need to be caching DNS for any reason, no matter how easy it is.  Sure, if they want to play around, poke, hack, learn, etc.  Still, I hope no "average" home user reads this thread and gets the idea that local caching will be anything other than fun and games - or worse, that it will somehow improve their audio experience.... 😋

 

On the other hand, you folks with teenager who think your "safe" browsing app is anything more than a DNS service redirect, and that your teenager did not figure out how to get around it in 2.3 seconds....perhaps you need to play around a bit and become aware of some basic Internet wonkery...

But, your windows box has a caching server built in (AFAIR.)  It is very likely that anyone running a recent version of Windows hasn't really been without a caching server.  On Linux (AFAIR), it all depends on the distribution, but I would make sure that one is running.

Run with the caching servers disabled -- then do the comparison.  I have always liked one running.

 

There just is ZERO downside -- really.

 

Now, web proxy caches do tend to be less useful -- content is to variable nowadays, and there ARE invalidation issues with those.  (You have to program timeouts and depend on some things that aren't always supported in WWW sites.)  I used to use a squid proxy also -- but the WWW browsers already do good enough.

 

 

 

John

[John Dyson]

18 minutes ago, crenca said:

 

Right.  Back in the olden days with Windows NT (both server and desktop), "ipconfig /flushdns" was basic network fix stuff.  I have long forgotten what the persistence of this cache was (and is).  

 

Still, this is not the same as setting up a DNS "caching server" with your cheap-ass chi-fi home networking crap gear in the modern situation of ISP DNS service (or Goggles, or...) which is fast as hell and has an uptime several orders of magnitude greater than whatever you can pull off with your China Net gear like Jud sort of stumbled into 😋 🤣

 

They are essentially the same thing (machine local and on your main server.)  The original reason (still valid) for the caching servers is that they would be used when there were multiiple machines on the network, and since I am likly  to have a server anyway when I have multiple machines, a caching server is a freebee.  (DNS caching already machine-locally active.)

 

Using a caching server on a garbage box -- the person is already dealing with garbage of some kind anyway...  I am speaking of a real DNS server -- machine local or a server that also does DNS.

 

All of the 'disadvantages' of a caching server are already manifest on the existant machine local caching servers -- so there is no downside.  A nice thing about DNS (from day one) is that if it doens't get a UDP packet back right away -- if a zone request hasn't been done, then it goes on to the next source -- likedy split...  If you are running on a primitive (kind of 'slave') unix system, then failure is even handled there with the list of 3 servers -- FAILURE HAS ALREADY BEEN CONSIDERED

 

DNS was designed to be 'caching' for many years (since at least the early 1990s'.)  Again -- ZERO disadvantage to having one.  It only provides advantages -- your server for other things is already online.  No biggie, right?

 

Isn't this an # angels on the head of a pin?  I say essentially zero cost, then why not (the disadvantages are already manifest, if ANY - which are essentially none, and failure modes are already handled.)

 

The arguments that I have seen against DNS servers ARE valid for WWW proxy servers, but not DNS.

 

John