Results 1 to 25 of 25
  1. #1

    I need 100% uptime... Can I do that with 2-3 servers with the same info?

    Hello there...

    I have a website that receives a lot of traffic and one crucial thing is to have 100% uptime.

    I think I can do that with 2-3 dedicated servers mirroring each other right and with multiple connections?

    Please help me with how can I do that and which hosts would be willing to set up such thing for me.

    Best,
    Carl

  2. #2
    Join Date
    Mar 2004
    Location
    Los Angeles, California
    Posts
    979
    I think the simple response is no. But I'm sure someone knows of a way to do this.
    ••• Edrick Smith •••
    ••• C47 Hosting LLC | Media Centric High Demand Web Services •••

  3. #3
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    Are you hosting a database also?
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  4. #4
    Join Date
    Dec 2005
    Location
    United States
    Posts
    503

  5. #5
    Quote Originally Posted by FrozenHost
    If LT Databank has a network issue and goes down, he goes down, so that wouldn't be 100% uptime.


    Alex

  6. #6
    Join Date
    Dec 2005
    Location
    United States
    Posts
    503
    Which is really not going to happen.

  7. #7
    Join Date
    Jul 2004
    Location
    Texas
    Posts
    688
    Quote Originally Posted by FrozenHost
    Which is really not going to happen.
    If you search, Databank has had quite a few problems.

  8. #8
    100% is not possible, even if you could get five 9's with in your own setup, your upstream providers (colo host, networking) will be your greatest cause of failure.

  9. #9
    Join Date
    Feb 2005
    Location
    United Kingdom
    Posts
    3,476
    100% uptime impossible due to network packets loss...
    Explore our outstanding VPS choices tailored to your budget, and we welcome reasonable offers.
    VPS Price Match Guarantee on: All our range of DDOS protected XEN HVM VPS
    == Contact us for any online solution development or managed / unmanaged vps hosting ==

  10. #10
    Join Date
    Dec 2006
    Location
    Missoula,MT
    Posts
    46

    How do you define 100% uptime?

    I'm not being silly here, but it's simply not possible. If your server runs a service, let's say Apache (your web server), and that server fails, some people will be directed to the down server prior to being redirected to the failover server. You will be effectively "down" but again, you must define what you mean by uptime.

    For example, I host and manage a MySQL cluster that has a 6-15 second lag in failover components engaging and taking over IP and database duties. While most will find this time frame reasonable, some will not.

    The time frame for failover depends on what you are failing over and the underlying technology that service runs on. For example, some SANs can have hardware failure in any one disk of the SAN and you will see relatively short (a few MS) lag in takeover.

    While DNS records are often cached, and if the visitor of a website has a cached IP for a website, ie. the A record for www.domain.com is cached by AOL, it will not matter that you have failover DNS, it will continue to fail until AOL refreshes it's DNS cache of that IP. Since AOL is not wanting to check DNS for every domain upon each request, it caches, and so doing actually prevents failover of services based on DNS, which is every website. In order to avoid this you would require the ability to failover IP services among multiple Data Centers and again, you will always find some downtime associated with that and it's not practical to do this.

    The shorter the "downtime" of a service that you want, the higher the price will be to accomplish it and the technical requirements will also grow, probably exponentially.

    So, realistically define "downtime" and then you will be able build a solution that you can consider 100% failover and will provide you with the most reliable services you desire. No $10/month host will give you 100% uptime and no $100/month dedicated server can do it either. The claims of 100% uptime are myth, even considering the awesome power that these technologies give us, there is no perfect solution and problems/issues will happen and downtime is something to be minimized, but cannot be eliminated.
    SupportLayer - Enterprise Linux Server Management

  11. #11
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    If you have a network that is up 100% of the time, what you can do is:

    Redundant Load Balancers
    --
    Redundant Web Servers; Database Servers; E-mail Servers, etc. [whatever you require]
    --
    Redundant NAS

    This is a very brief example but basically all of the storage is done on the two redundant NAS' keeping all of the other servers in-line. Obviously everything here comes at a price.

    If you're looking for a control panel that can allow a similar setup to the above, you're going to want to look at H-Sphere. www.psoft.net

  12. #12
    Hello carlsp. Consider this solution (I was making use of it for two years): geographically distributed dedicated servers (California, Chicago, Atlanta) with mirrored (by mod_proxy, reverse caching) versions of the website, and round robin DNS by DNS Made Easy.
    Guille

  13. #13
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    Quote Originally Posted by Guille
    Hello carlsp. Consider this solution (I was making use of it for two years): geographically distributed dedicated servers (California, Chicago, Atlanta) with mirrored (by mod_proxy, reverse caching) versions of the website, and round robin DNS by DNS Made Easy.
    That's fine if it is a static site but when you start looking at requiring MySQL things get very complicated.

  14. #14
    Quote Originally Posted by layer0
    That's fine if it is a static site but when you start looking at requiring MySQL things get very complicated.
    It's also fine when most of the dynamic pages can be converted to static.
    Guille

  15. #15
    Join Date
    Dec 2001
    Location
    127.0.0.1
    Posts
    3,642
    The nearest uptime that's currently truely attainable is 99.999% which is provided by IBM, at a cost of above $1M USD per year.
    Simpli Networks, LLC :: http://www.simplinetworks.com :: Proudly 100% Owned.
    Providing Affordable Managed Cloud/VPS Servers & Server Management Solutions.
    We offer REAL 24x7x365 in-house support - proudly serving our customers since 2005!
    Want to learn more? Give us a call - +1 (844) 4SIMPLI or email sales[@]simplinetworks.com today!

  16. #16
    Join Date
    Feb 2004
    Posts
    633
    Quote Originally Posted by layer0
    If you have a network that is up 100% of the time, what you can do is:

    Redundant Load Balancers
    --
    Redundant Web Servers; Database Servers; E-mail Servers, etc. [whatever you require]
    --
    Redundant NAS

    This is a very brief example but basically all of the storage is done on the two redundant NAS' keeping all of the other servers in-line. Obviously everything here comes at a price.

    If you're looking for a control panel that can allow a similar setup to the above, you're going to want to look at H-Sphere. www.psoft.net
    Shared storage (NAS or SAN), redundant or otherwise, is not necessarily required for high availability. A shared nothing architecture is as equally valid as a shared storage architecture for high availability.

  17. #17
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    A redundent nas would make creating a redundent email system much easier.
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  18. #18
    Join Date
    Nov 2003
    Location
    San Francisco, CA
    Posts
    921
    100% uptime IS possible regardless of what you hear. Visit RagingWire.com for a facility without a single planned or UN-planned network cooling or power outage for the last 5 years straight. They also offer managed services and will lease machines if asked. They are a colo facility but will do much more when requested.

    I would offer our facility except that (a) it is against forum rules to do so and (b) we typically do not heavily pursue managed hosting of this nature so we would be a useless option in this case.

    Good luck and for the record: www.ragingwire.com
    Last edited by thecloudguy; 01-08-2007 at 01:41 AM.
    Adam - AQORN
    Official OpenStack Foundation Member, Corporate Sponsor and AI Research Team
    Our Clients: Facebook | Red Hat | Starbucks | AT&T | HPE | NVIDIA | CrowdStrike | Cisco | Juniper | SAP | Autodesk | SUSE | Ubuntu

  19. #19
    If they had 100% power uptime (that is what you mean i think) it doesnt mean they will have it in the future!

  20. #20
    Join Date
    Dec 2006
    Location
    Missoula,MT
    Posts
    46
    "We provide "Five Nines" (99.999%) availability with power and cooling....."

    They don't even advertise 100% availability on the homepage.

    Just because a host has a great record, it doesn't affect chaos theory and the Heisenberg principle of uncertainty, which limits the concept of 100% uptime/availability to the realm of Marketing Hype!!

    This is why I asked the OP to clarify what "uptime" was, as we can have quite a confused conversation using such a vague term.
    SupportLayer - Enterprise Linux Server Management

  21. #21
    Quote Originally Posted by layer0
    If you have a network that is up 100% of the time, what you can do is:
    I would be interested in knowing where/who you think can give a network with 100% uptime?

    In my experience even major backbone suppliers make routing mistakes, flash the firmwire in routers which require reboots, and make other dumb misconfiguration mistakes which cause outages. No ones perfect which means no network is 100% .

  22. #22
    Join Date
    Apr 2004
    Location
    SF Bay Area
    Posts
    879
    Quote Originally Posted by CALHOST-Adam
    100% uptime IS possible regardless of what you hear. Visit RagingWire.com for a facility without a single planned or UN-planned network cooling or power outage for the last 5 years straight. They also offer managed services and will lease machines if asked. They are a colo facility but will do much more when requested.
    Actually, I'm fairly sure Ragingwire finally had its first ever pretty major power incident in the last year. I'm sure it wasn't widely publicized but I know they had power issues that killed hardware of several customers. I'd ask around from some known RW customers like Alibris, eQuest, Adteractive etc. to verify this if you ever want to go into this facility, which I agree is a very nice DC (had equipment in there once myself) and shouldn't be discounted for having 1 power issue in 5+ years.

    In any case, 100% uptime is always an illusion. You can architect pretty darn reliable (99.999%) solutions but even then there can be problems. Even Google's had issues and they far and away have the world's largest global cluster of servers - well, if cluster is quite the right way to think of what google has :>

  23. #23
    does the original poster actually need 100% uptime?

    First off, no one can guarantee 100% uptime nor can they offer any sort of reasonable SLA against this value.

    I think the best available right now is 6 to 7 9's - and as Jonathan pointed out, the costs for this are enormous...

    From my experience, most people searching for 100% uptime would be happy with 4 9's (as those seeking a higher uptime performance the 4 9's would not be posting on wht nor would they be contacting the average company even mentioned here). This is certainly attainable within a reasonable budget - but, it is important to be done correctly. layer0 is correct in saying that redundant LB'ers, server nodes with as many redundant componants as possible as well as redundant shared storage arrays will greatly improve an environments capabilities of providing superior, reliable uptime of all services...
    www.cartika.com
    www.clusterlogics.com - You simply cannot run a hosting company without this software. Backups, Disaster Recovery, Big Data, Virtualization. 20 years of building software that solves your problems

  24. #24
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    Quote Originally Posted by prickett233
    I would be interested in knowing where/who you think can give a network with 100% uptime?

    In my experience even major backbone suppliers make routing mistakes, flash the firmwire in routers which require reboots, and make other dumb misconfiguration mistakes which cause outages. No ones perfect which means no network is 100% .
    I never said it was available, but instead I said *IF*. Basically I described a high availability solution that is as reliable as the network.

  25. #25
    Join Date
    Jun 2005
    Location
    CA, USA
    Posts
    145
    Quote Originally Posted by CALHOST-Adam
    100% uptime IS possible regardless of what you hear. Visit RagingWire.com for a facility without a single planned or UN-planned network cooling or power outage for the last 5 years straight. They also offer managed services and will lease machines if asked. They are a colo facility but will do much more when requested.

    I would offer our facility except that (a) it is against forum rules to do so and (b) we typically do not heavily pursue managed hosting of this nature so we would be a useless option in this case.

    Good luck and for the record: www.ragingwire.com
    This is for the data center yes? One of the datacenter's I use offered 4 9's but also offers 10,000%% money back on any outages which makes it worth more 9's to me. But you still need to have a good server as they are the most unreliable, especially the OS. I recommend getting a fallow server oe even faster one big server VMed so if hardware in the server or the OS failes the mirroring node is switched to. But even routers take a while to fallover or are often not set up right.

    Robert Davis

Posting Permissions

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