Re: BizTalk server Topology & scalibilty
- From: "Jon Fancey" <JonFancey@xxxxxxxxxxxxxxxxxxxxxxxxx>
- Date: Fri, 8 Apr 2005 05:03:01 -0700
Re-reading the original question, I think I was answering a differen one :-)
apologies Halo.
"Jon Fancey" wrote:
> Yes that's true, thanks. I hadn't considered splitting the send and receive
> sides of the same adapter. From my reading I was only thinking of the
> orchestraton/adapter split. I'm still not sure why you'd *want* to do it
> though along these lines? It seems more logical to me that you'd want to
> split orchestrations up across different hosts as needs dictate rather than
> an 'orchestration server'? This would allow for more predictable perf and
> control over relative priorities.
>
> "Matt Milner" wrote:
>
> > Actually, if I understood the question correctly, this is definitely
> > possible. Receive adapters are bound to one or more hosts, send adapters
> > ard bound to a host, and orchestrations can be bound to a host. In
> > addition, tracking can be configured on a host basis. So you can definitely
> > have host that is enabled for tracking and that has all of the receive
> > locations bound to it. You can have another host that you setup your send
> > ports on and that you bind all of your orchestrations to. There is some
> > trickiness in terms of the send ports getting bound to default hosts and not
> > being able to change the host for the MSMQT adapter, but for hte most part
> > it is straight forward.
> >
> > Matt
> >
> >
> > "Jon Fancey" <JonFancey@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
> > news:A1DE7E78-1DA3-4C27-9E82-EA0CDD0F4BC2@xxxxxxxxxxxxxxxx
> > > HaloX,
> > >
> > > I'm afraid the simple answer to your question is No.
> > >
> > > What you are essentially asking is to split different but related
> > > activities
> > > within the BizTalk engine across multiple engines on different servers. As
> > > Hugo points out BizTalk is stateless, that is to say all state is
> > > maintained
> > > only in the SQL server databases (unless one inadvertently starts creating
> > > stateful adapters, orchestrations, etc but that's another story). This
> > > includes orchestrations, pipelines etc. So, in a non-deterministic way the
> > > load from receive to send *may* be processed by different servers, for
> > > example if the orchestration instance is dehydated to the SQL store and
> > > subsequently rehydrated on another box.
> > >
> > > But this is non-deterministic and as such you aren't able to *specify*
> > > particular servers for particular tasks at this granularity. The only way
> > > to
> > > achieve this is to perform routing between servers (by name) within say an
> > > internal service such that a received request is passed by an
> > > orchestration
> > > on box A to box B which will send a reply. This would require some
> > > hoop-jumping though to avoid a request/reply pattern just passing straight
> > > back through both boxes for the response. It probably wouldn't buy you
> > > much
> > > either if perf or scalability is what you're after.
> > >
> > > However, if you are not using a request/reply pattern, say you are using a
> > > one-way request and one-way response, then you could just deploy each of
> > > these to different machines (groups - that don't share the same SQL
> > > config)
> > > and configure it to achieve what you want. The connection (if any) between
> > > the two though then has to be handled by you explicitly i.e. you'll have
> > > to
> > > do your own correlation.
> > >
> > > Hope this helps clarify it a bit for you. I am interested why this is your
> > > desired config?
> > >
> > > "HaloX" wrote:
> > >
> > >> Looking through the BizTalk documentation and books I see recommended
> > >> configuration of hosts. There are the following configurations:
> > >>
> > >> 1) Single server: Single server runs ALL BizTalk and SQL operations
> > >> 2) Two Server: A single server runs SQL while another runs all BizTalk
> > >> operations.
> > >> 3) Multiple Server: Multiple servers running BizTalk and configured
> > >> with
> > >> each machine host running specific isolated functions to scale out the
> > >> receive, processing, transmitting and tracking. One or more (recommended
> > >> for
> > >> this configuration) servers running SQL in cluster.
> > >>
> > >> Now, the issue is I have a SQL cluster running SPS. I want to leverage
> > >> this
> > >> cluster because it is hardly tasked and I will have BizTalk run on a
> > >> seprate
> > >> SQL instance. I would like to have two machines running BizTalk:
> > >>
> > >> * One configured a Message Box Master Recieve, and Tracking
> > >> * One configured for Transmit and Orchestrations
> > >>
> > >> Will BizTalk support this configuration. Nothing in the documentation
> > >> identifies if certain configs AREN'T supported, but from my experience
> > >> this
> > >> doesn't mean much. In SPS there are certain configs not supported and
> > >> nothing in the initial documentation identified this.
> > >>
> > >> Thanks
> >
> >
> >
.
- References:
- BizTalk server Topology & scalibilty
- From: HaloX
- RE: BizTalk server Topology & scalibilty
- From: Jon Fancey
- Re: BizTalk server Topology & scalibilty
- From: Matt Milner
- Re: BizTalk server Topology & scalibilty
- From: Jon Fancey
- BizTalk server Topology & scalibilty
- Prev by Date: HAT question
- Next by Date: RE: ANNOUNCING BIZTALK SERVER "NEWUSER" NEWSGROUP
- Previous by thread: Re: BizTalk server Topology & scalibilty
- Next by thread: Re: Rule No fire
- Index(es):
Relevant Pages
|
Loading