Mailing List Archive

On Friday, November 2, 2001, at 06:36 AM, Geoffrey Young wrote:
> I caught Theo's presentation way back in ApacheCon Orlando - cool
> stuff.
> anyway, just curious if anyone knows about using backhand with recent
> versions of BigIP - the network guys here are saying that BigIP 4.0 can
> do
> all the proxying stuff that backhand can do. well, they actually hadn't
> heard of backhand, so they were just saying that it could do what I was
> describing, but it's been a while since the presentation.

As far as I know, BigIP cannot do all of the proxying backhand can do.
It cannot multiplex incoming connections over a single back end
connection and it cannot dissect an incoming connection and assign
subsequent HTTP requests to different machines.

> are these guys on track?

Dunno. BigIP is a good product. I use it and I like it. But, the "we
can do everything you can do only better" argument gets a little old.
BigIP and mod_backhand are fundamentally different. mod_backhand could
do database lookup _before_ assigning the request to a specific machine
based on the headers -- try that with BigIP. BigIP can advertise a VIP
and balancing things like FTP and SMTP -- try that wth mod_backhand.
You get the picture.

> also, I swore that there was a couple of slides in the ApacheCon
> presentation that showed the peaks when just using BigIP and how
> backhand
> could truly level the load across a cluster. I didn't see it in any of
> the
> slides at backhand.org. Does anyone have a nice pretty picture of
> something
> like that that I can show management? right now I'm just trying to
> convince
> them to compile it into the servers in our test lab, but everyone
> thinks F5
> is all we need and it's not worth even investigating...

There are advantages of have mod_backhand behind (in conjunction with) a
BigIP.

o concise resource advertisements and display.
o more complex content and load-balancing rules.
o more intelligent handling of heterogeneous machines
o Sun, Intel, Alpha several of each all different disks, CPUs and
memory.
o graceful handling of misbehaving machines
o Dual processor machine, boots with only 1 processor online
o byzantine cron job runs awry and trashes disk, memory or CPU.
o slow disk failure impacts machines performance.


--
Theo Schlossnagle
1024D/82844984/95FD 30F1 489E 4613 F22E 491A 7E88 364C 8284 4984
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7

> As far as I know, BigIP cannot do all of the proxying backhand can do.
> It cannot multiplex incoming connections over a single back end
> connection and it cannot dissect an incoming connection and assign
> subsequent HTTP requests to different machines.

Actually, the new BIG-IPs can load balance based on the URI (have to
in order to be competitive with CSS) but it's a bit more CPU intensive
than normal TCP forwarding.

> > are these guys on track?

sub(/track/, 'crack') ??

> Dunno. BigIP is a good product. I use it and I like it. But, the "we
> can do everything you can do only better" argument gets a little old.
> BigIP and mod_backhand are fundamentally different. mod_backhand could
> do database lookup _before_ assigning the request to a specific machine
> based on the headers -- try that with BigIP. BigIP can advertise a VIP
> and balancing things like FTP and SMTP -- try that wth mod_backhand.
> You get the picture.

::drool:: mod_backhand + port forwarder

Real quick: big-ip can't even come close to replacing the benefit of
having something proxy out the content. mod_backhand can act as that
proxy, but big-ip can't: it's just a fancy port forwarding machine with
almost no buffer (ie: your 20MB mod_perl process is still spending
time sending data to the client instead of serving another request for
a proxy server). -sc