Re: [load balancing] software vs appliances.

From: tc lewis (
Date: Thu Feb 22 2001 - 15:42:09 EST

  • Next message: tc lewis: "RE: [load balancing] software vs appliances."

    > I like doing my work on the network level, having a seperate box doing the
    > work. It makes things simpler that way, just my preference.

    i still use separate boxes with lvs. but it's not required.

    > 1. Maturaty. From the looks of it that Linux stuff is still in
    > development, not quite geared towards a product release that the
    > commercial stuff is. When things go apeshit it's hard to explain to a
    > customer that the reason is some bug you havn't figured out on a Linux
    > server. It doesn't seem to be nearly as far along as the commercial
    > products as far as features and optimization.

    hmm, i'm not sure. from a linux standpoint, the biggest linux companies
    in the game support lvs. redhat sells a separate balancing/clusting
    package of which the balancing is lvs-based. turbolinux's balancing is
    based on lvs. valinux as well. and so on...

    i guess "linux" isn't a name like "alteon". but the corporate-view of
    big-name branding was never an argument that held a lot of water with me.

    lvs is still being developed, but the last number of 2.2 releases have
    been quite stable. only enhancements have been added (maybe some minor
    format issues or something, but nothing major). the 2.4 series of lvs is
    a bit less mature, i'll grant you that.

    isn't an f5 bigip just a stripped down bsd?

    at least if a bug is discovered with lvs, i can fix it. if a bug is
    discovered in my bigip, i have to spend a week on the phone with support
    trying to get some sort of firmware upgrade i guess. ok i'm exagerrating.

    > 2. Support. Having a commercial support structure is a big plus, even if
    > you have to pay for it. There is more accountability in the vendors part,
    > so they are more apt to make their releases stable. This is an argument
    > used in alot of pro-microsoft dialog, but I think this situation is
    > different. The user base for Linux Virtual Server is alot lower compared
    > to Linux in general, and help on news groups isn't really there like there
    > would be for a network card driver problem.

    the lvs-users list is nothing short of absolutely spectacular in terms of
    support. i've never been more impressed with some of the minds on there.

    commercial support is also available. redhat, turbolinux, va, etc...

    it's the same old argument of open vs closed software and being able to
    speak with the developers directly vs being able to motivate a company
    with money. there are pros and cons on both sides. if you lean on the
    other side, that's perfectly understood.

    > 3. Features. There are several features that Linux Virtual Server
    > doesn't support yet from what I can see. Critically, it doesn't do
    > cookie-based persistence. Most web-store application sites require this
    > to get around the AOL-megaproxy issue.

    can this be explained more? the balancer should open http requests, take
    out the cookies, parse them, and then send certain flags to a persistent
    real server?

    i thought half the point of cookie or url sessioning was so that
    persistent connections to real servers isn't necessary.

    lvs is strictly a layer 4 switch operating at the ip level. it doesn't
    look at http/other protocol information. i'll grant you that
    protocol-level could give functionality that lvs doesn't have. i've
    never found that functionality to be useful. i imagine others have. has
    anyone found this to be necessary? just more convenient?

    > 4. Speed. F5 has spent a considerable amount of resources streamlining
    > their code from the looks of it, and the switch-based vendors have their
    > ASICs. All in all, the commerical products probably would be able to
    > handle quite a bit more traffic than the Linux stuff.

    there are plenty of reports of lvs outperforming f5.
    i could dig up a few...
    they might be on the web page somewhere too.

    lvs allows for network balancing via nat, ip tunneling, or direct routing.
    it also has a variety of scheduling algorithms. dr is very efficient.
    it's the same sort of algorithm that ibm's netdispatcher first provided (i
    think someone just mentioned that on this list). balancer redirects
    packets to real servers without even rewriting the ip headers at all --
    return traffic bypasses the balancer completely and goes back out the next
    hop. basically infinitely scalable [up to the bandwidth of the network
    interfaces]. very fast. i've been told that f5 has been providing this
    sort of algorithm now as well, but i've never used it -- only its nat.

    > So yeah. You can't beat the Linux Virtual Server's price in some cases,
    > however when there is a ton of money on the line for a site or a hosting
    > provider, I think the commercial stuff is the best bet. I suspect that
    > will change at some point on the future, but for now the commercial
    > products have a head start.

    i think that has a lot to do with branding. ibm is a big name, cisco is a
    trustworthy name, so they must be good enough for us.

    lvs is good enough for, sourceforge,, valinux,

    but hey, lots of people don't like linux either. chuckle.


    This archive was generated by hypermail 2b30 : Thu Feb 22 2001 - 15:43:27 EST