[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-ha-dev
Subject:    [Linux-ha-dev] SA/Phoenix(RSCT)/HACMP (was: Re: Availability resource management
From:       TEREKHOV () de ! ibm ! com
Date:       2000-10-12 12:18:59
[Download RAW message or body]





> David, glad you asked that :-)

Peter, glad you answered that :-)

David,

As Peter pointed out SA is NOT a cluster enabling technology.

SA provides Availability Resource Management function in
both cluster and non-cluster (single system) environments.

On OS/390, SA uses (in Peters terms):

1. XCF for "node/network up and down status info
=A0 (i.e., a heartbeat component)" and "grouping,
=A0 multi-node synchronization (i.e., group services)"

2. MQSeries for "reliable communications"

3. Netview for "resource monitoring/control,
=A0 generalised event delivery (i.e., event management)"

Something like Phoenix(RSCT) or "heartbeat" =A0certainly needs
to provide XCF functionality for SA on Linux.

As for Netview part ("resource monitoring/control,
generalised event delivery (i.e., event management)"),
well, for Linux, we would probably need to provide a) an API
and b) runtime in order to:

a) enable creation of resource specific monitors and
=A0 controllers from scratch;

b) enable reuse of existing monitoring and control
=A0 facilities (scripts, RSCT, Tivoli, CIM, etc..)
=A0 via prepackaged plug-in modules.

This API would serve the same purpose as Microsoft
MSCS (Wolfpack) Resource API. It would be nice to
discuss the API here informally before coming to a decision..

Note that not only SA but other products such as FailSafe,
LifeKeeper, HACMP/Linux, etc.. could benefit from having
a common API.

As for overlap with HACMP, SA and HACMP are two different
approaches to manage resource availability.

IMHO, the strength of SA is in support of SHARE ALL
cluster model (we come from "shared all" Sysplex) versus
SHARE NOTHING model via cross-system dependencies,
grouping, goals and conflict resolution, sophisticated
relationship and failover/failback model and server capacity
(server farm) management.

It would be interesting to know whether you see any value in
those features in a Linux space..

Sure, I understand your confusion related to HACMP and SA
overlap, certainly this and perhaps other questions such as
OpenSource vs. proprietary code need to be clarified.. Please
provide your input and help address those issues in the best
interest of customers, linux-ha community and IBM. Please
understand that, being a developer I am not in position to
comment on these issues..

regards,
alexander.

To: =A0 =A0 =A0 =A0linux-ha-dev@lists.tummy.com
cc: =A0 =A0 =A0 =A0Alexander Terekhov/Germany/IBM@IBMDE, dbrower@us.ora=
cle.com
From: =A0 =A0 =A0 =A0Peter Badovinatz/Beaverton/IBM@IBMUS
Subject: =A0 =A0 =A0 =A0Re: [Linux-ha-dev] Availability resource manage=
ment for
Linux (1/2) =A0Link

David, glad you asked that :-) =A0The easiest point to make is on the P=
hoenix
front, since Phoenix (or, RSCT as the IBM pubs will call it) does not
itself have any resource control "intelligence." =A0It is essentially j=
ust
the plumbing infrastructure used to monitor and control resources. =A0T=
hus
the comparison to OS/390 System Automation (I'll use SA) would be as
underlying support services used by the SA 'driver.' =A0I am not famili=
ar
enough with SA to comment on what actually exists in that space on OS/3=
90.
One could conceive of Phoenix being used in a support role if an SA
component were to be brought to Linux, as something certainly needs to
provide:
- node/network up and down status info (i.e., a heartbeat component)
- reliable communications, grouping, multi-node synchronization (i.e.,
group services)
- resource monitoring/control, generalised event delivery (i.e., event
management)
These of course being the Phoenix components, which themselves make NO
decisions about placement or control of resources. =A0Likewise, Heartbe=
at
could also slot in here, and other items. =A0Key was that the writeup a=
s
given did not specify a specific underlying infrastructure (unless I mi=
ssed
it in there.) =A0I know that OS/390 has it, but I do know it isn't Phoe=
nix!

Now, as to HACMP, it DOES make these resource decisions, also, newer
versions of HACMP use Phoenix as the underlying support infrastructure.=

However, HACMP's decision making is much more stylised than is done by =
SA,
and is fixed by "simpler" protocols and rules. =A0HACMP is in many ways=

similar to FailSafe in its resource management model (no, not exactly, =
but,
it is close enough, and that helps place it for the audience on this li=
st.)
SA provides a more flexible and dynamic model, albeit at some level of
configuration and conceptual cost (i.e., I need to read the description=

again myself :-) =A0Simply put, IMHO, HACMP and SA are two different
approaches to placing resources, and to making sure that they're availa=
ble,
based upon the user's desires. =A0I also see them as mutually-exclusive=
, you
would need to pick one or the other.

From the writeup, I don't think configuring SA would be that easy, and =
I
know that HACMP is not trivial either.

As to the specific availability of Phoenix, HACMP and/or SA on Linux, a=
nd
IBM stratagies, absent any official IBM announcements, I cannot say
anything :-( =A0The archives of this list have about everything I can, =
to
this point, say.

These have been the opinions of:
Peter R. Badovinatz -- (503)578-5530 (TL 775)
Clusters and High Availability, Beaverton, OR
wombat@us.ibm.com
and in no way should be construed as official opinion of IBM, Corp., my=

email id notwithstanding.


Please respond to linux-ha-dev@lists.tummy.com

Sent by: =A0 =A0 =A0 =A0linux-ha-dev-admin@lists.tummy.com

To: =A0 =A0 =A0 =A0linux-ha-dev@lists.tummy.com, Peter
Badovinatz/Beaverton/IBM@IBMUS, Alexander Terekhov/Germany/IBM@IBMDE
cc:
Subject: =A0 =A0 =A0 =A0Re: [Linux-ha-dev] Availability resource manage=
ment for
Linux (1/2)



Hmmm, this seems reasonably thought out, but I'm a little
confused by what might seem to be overlap with existing
AIX functions in the Phoenix/HACMP space. =A0On the
surface, I might rather have those facilities on Linux
sooner rather than waiting for something like this later; and
I'd think that moving the AIX stuff to Linux would better
align with IBMs strategies than S/390 stuff.

Can you and Wombat comment on the overlaps?

thankx,
-dB

TEREKHOV@de.ibm.com wrote:
>
> Hi,
>
> I am working for OS/390 System Automation department
> (IBM German Software Lab). The suite of products we develop
> here provides high availability and automation solutions for
> systems and applications (OS/390 Apps) running on S/390
> mainframes in both single system and cluster (IBM Sysplex)
> environment. In recent times several development groups in
> the Lab have been undertaking quite a lot of activity in
> the direction of Linux for S/390.. That will enable our
> customers to have Linux system images running together with
> their legacy OS/390 system images on the same hardware which
> could provide many benefits in heterogeneous environments
> (e.g. DB2/CICS/IMS running on OS/390 with business logic
> apps running on Linux/390). We expect that in such
> environments customers would benefit from having some kind
> of cross platform high availability (automation)solution
> which would not only cover OS/390 alone or Linux alone but
> would also close the gap between two platforms...
> (e.g. would manage cross system/platform dependencies,
> provide unified scheduling, administration, etc.)
>
> Recently we have evaluated some existing Linux
> technologies/products which may fit in such heterogeneous
> environment and become somehow connected with S/390 HA
> offerings... But because of differences in resource
> management concepts currently implemented on the Linux
> side such an integration does not seem to be an easy job
> (from S/390 HA perspective).
>
> Currently we are evaluating the possibility of
> providing S/390-like availability resource management
> (automation) on Linux (with focus on Linux/390).
> The document below explains some concepts related
> to availability resource management. It would be
> interesting to know what people with Linux/Unix HA
> backgrounds do think about feasibility of these concepts
> (resource status model, relationships, grouping/failover,
> etc.) from a Linux/Unix perspective.. Thank you in advance
> for your feedback.
>
> regards,
> alexander.

--
Butterflies tell me to say:
"The statements and opinions expressed here are my own and do not
necessarily
=A0represent those of Oracle Corporation."

_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.tummy.com
http://lists.tummy.com/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/



=



_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.tummy.com
http://lists.tummy.com/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/

[prev in list] [next in list] [prev in thread] [next in thread]