Results 1 to 6 of 6
  1. #1
    Join Date
    Oct 2005
    Posts
    73

    How to use 2 Windows VPSs for Higher reliability ?

    I would like your expert opinions regarding this scenario:

    In order to increase the reliability of a audio streaming service I am thinking to take the action I describe below.

    1. Buy two Windows VPSs with WMS installed.
    2. Register a domain name (i.e audiostream.com) with 4 nameservers: ns1.ip1_vps1, ns2.ip2_vps1, ns3.ip1_vps2, ns4.ip2_vps2.
    3. Create all publishing points (streams) on both servers.

    Normaly VPS1 should serve all clients. In case VPS1 goes down, VPS2 should jump in and serve all clients-connections. As soon as VPS1 becomes available (ns1 & ns2 start responding again), VPS1 starts serving all new connections.

    Load balancing or any other kind of advanced load, traffic, etc management is not important.


    Then we have the following cases:

    1. VPS1 is down , which means that VPS2 should take charge.
    2. VPS1 is up (ns, http, ftp services), but WMS1 is down (crashed). Means that the playlist (asx) file should be built so that it also includes the IP, port & publishing points of VPS2. This should be done because ns1 & ns2 will answer without any problem, but WMS1 will be crashed and won't serve any media connections, thus ASX will look for the next available stream in the playlist.


    Do you think that the above is possible. Is this gonna work? If not have you got any other suggestions?

  2. #2
    you can use something like DNSmadeeasy for Failback service between VPS1 and VPS2. I believe you can configure it for services also.

    I have used it, not all ISP will follow the low TTL DNS value, but most do. You can even do Round Robin service via DNSMadeeasy.

    There are other companies that do the same service, I have only used DNSMadeeasy.

  3. #3
    Quote Originally Posted by tonino View Post
    Normaly VPS1 should serve all clients. In case VPS1 goes down, VPS2 should jump in and serve all clients-connections. As soon as VPS1 becomes available (ns1 & ns2 start responding again), VPS1 starts serving all new connections.
    This is a very common mistake. If we can find who the original person that stated this "Internet urban legend" we would take them out back and shoot them since many of our support tickets are based on this misconception.
    But DNS has _never_ worked this way. Actually when I say never I mean that if the resolving name server follows the RFCs then it should never work that way. Since most resolving name servers use Bind, something like djbcache, or Microsoft DNS then they all follow the RFCs here and would not work the way you mentioned either.

    When you give your registrar 4 name servers it's not the case that three lay idle until the first one goes down.
    If that was the case then there would zero resason to have 13 root name servers.

    When you give your registrar 4 name servers then ALL 4 are hit 25% the first time. If that one is down then there is 3 left. So there is a 33% chance of one being chosen. If the second one is down then there would be 2 left and a 50% chance.
    Usually a resolving name server can make it up to 3 to 5 checks before it times out and the query fails.

  4. #4
    Quote Originally Posted by tiggee View Post
    This is a very common mistake. If we can find who the original person that stated this "Internet urban legend" we would take them out back and shoot them since many of our support tickets are based on this misconception.
    But DNS has _never_ worked this way. Actually when I say never I mean that if the resolving name server follows the RFCs then it should never work that way. Since most resolving name servers use Bind, something like djbcache, or Microsoft DNS then they all follow the RFCs here and would not work the way you mentioned either.

    When you give your registrar 4 name servers it's not the case that three lay idle until the first one goes down.
    If that was the case then there would zero resason to have 13 root name servers.

    When you give your registrar 4 name servers then ALL 4 are hit 25% the first time. If that one is down then there is 3 left. So there is a 33% chance of one being chosen. If the second one is down then there would be 2 left and a 50% chance.
    Usually a resolving name server can make it up to 3 to 5 checks before it times out and the query fails.
    I must say I have been very happy with DNSMadeeasy, makes it very fast for me to move my sites between hosts, and the failover service has saved my bacon a bunch of times. Cheap insurance and I like to not have bind running, one less thing to patch and secure.

  5. #5
    Join Date
    Dec 2006
    Posts
    4,151
    The 4 nameservers do not work in a certain order, where if the first is down, it goes to next, etc.

    Instead, it's more of a percentage chance.

  6. #6
    Join Date
    Sep 2006
    Location
    Ohio
    Posts
    63
    I also use DNSMadeEasy and like the service a lot. Currently I use it with HostingZoom and Lunarpages. If the HostingZoom account goes down, within a few minutes my sites are pointed to the backup sites on LunarPages.

    You can also set up DNSMadeEasy to point at both so that both share the load. If one goes down, then (through failover) you can have them all redirected to just live server very quickly. (As a test I made my sites unavailable on the live server and it took less than 10 minutes for DNSMadeEasy to find the problem (max is 2 to 3 minutes) AND for the Change to propagate so that I could view it from various locations in proxy browsers and my own browser. Very nice!).

    Normaly VPS1 should serve all clients. In case VPS1 goes down, VPS2 should jump in and serve all clients-connections. As soon as VPS1 becomes available (ns1 & ns2 start responding again), VPS1 starts serving all new connections.
    DNSMadeEasy (and I'm sure that other companies that provide the same service will do the same) checks the server every couple of minutes to make sure its live (you can also set it so that if you have a database driven site, it will look for specific works/ text, etc). If it doesn't find it/the site is down, then it will failover to the backup site. They will automatically keep checking to see when the main site comes back online. Once it comes back online, the DNS is again pointed to the main site (Some folks that might be logging transactions to a database might not want that to happen until the databases can be synched... that can also be set so that after failover it stays pointed to the backup host).

    Matt
    Last edited by Certifiedwebpro; 10-05-2007 at 12:07 PM.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •