Re: [load balancing] Global Load Balancing with CSS

From: Dana Quinn (danaIZZATnextbus.com)
Date: Wed Jun 25 2003 - 13:11:20 EDT

  • Next message: Dal, Choi: "RE: [load balancing] who knows about Switching capacity of Alteon 2424?"

    I implemented a partial global load balanced site a few years
    ago - late fall 2000 into 2001, and I think the limitations
    were fairly well understood then. We used the F5 solution, and I
    remember having detailed discussions with some of their
    tech people about DNS caching issues, including DNS caching
    problems in various versions of Internet Explorer.

    I said a 'partial' site because the global load balancers
    were mainly used to do a datacenter move -
    it allowed us to slowly shift traffic over a couple months.
    We saw various client systems that continued to cache our old
    DNS results one way or another for weeks, even months. :(

    Anyway, I agree that using some sort of BGP-based multi-homing
    is a better way to survive most common failures, but setting
    up and running BGP requires a strong committment in staff
    and equipment. The little content company I worked for at the
    time I worked in this realm wasn't too interested in going that
    far.

    Does anyone on list have experience with using network
    service providers that help to manage BGP-like multi-homing
    for web sites? I'm thinking of InterNAP, mainly, or other
    companies that provide similar services. I'd be interested
    in hearing how they work out for people.

    Dana

    David Taylor wrote:
    >> May I ask how you moved away from the GSLB solution? We do it to
    >> get around points of failure like the network connection/ISP,
    >> and other single points of failure at one site.
    >>
    > We did indeed go for a failover situation and built our primary site to
    > be as resilient as possible. It gave us quite a shock to discover the
    > limitations of GSLB at the time (probably 2 or 3 years ago now) as we
    > had largely implemented a hosting environment that relied on GSLB - the
    > failover approach is nowhere near as elegant as an active / active
    > approach.
    >
    > At the time, the DNS caching limitation of GSLB did not seem to be
    > widely recognised.
    >
    > Our ultimate approach was to accept that in the event of total loss of
    > the primary site, our service desk would need to anticipate a high
    > volume of customer calls and would need to be able to manage customers
    > who required assistance with refreshing DNS caches. Not ideal, but for
    > the company an acceptable solution.
    >
    > In hindsight. had we been aware of the GSLB issues, we may well have
    > designed our environment using our own physically seperate datacenters
    > and BGP with upstream ISPs. In my view, for ultimate availability, IP
    > addresses must be completely static to the customer, regardless of what
    > goes on in the hosting network.
    >
    > Regards
    >
    > David
    >
    > _________________________________________________________________
    > Surf the net and talk on the phone with Xtra Jetstream IZZAT
    > http://www.xtra.co.nz/products/0,,5803,00.html !
    >
    > ____________________
    > The Load Balancing Mailing List
    > Unsubscribe: mailto:majordomoIZZATvegan.net?body=unsubscribe%20lb-l
    > Archive: http://vegan.net/lb/archive
    > LBDigest: http://lbdigest.com
    > MRTG with SLB: http://vegan.net/MRTG
    > Hosted by: http://www.tokkisystems.com

    -- 
    

    Dana Quinn, danaIZZATnextbus.com NextBus Information Systems (510) 420-3117

    ____________________ The Load Balancing Mailing List Unsubscribe: mailto:majordomoIZZATvegan.net?body=unsubscribe%20lb-l Archive: http://vegan.net/lb/archive LBDigest: http://lbdigest.com MRTG with SLB: http://vegan.net/MRTG Hosted by: http://www.tokkisystems.com



    This archive was generated by hypermail 2.1.4 : Wed Jun 25 2003 - 13:19:07 EDT