Slow in compiling the split info for this weeksmeeting.

[cherrypits@xxxxxxxxx (John Cherry)]

Tim Anderson wrote:
> Here is the first pass, it has Troy, Joe and my inputs. We will have to get a
> volunteer for Clustering since John did not do one before he left. And Peter
> is still working on the standards stuff.
>   
Sorry about the CGL clustering evaluation, but circumstances beyond my
control have kept me from tackling this.  A quick summary is this...

    * All the requirements that are addressed by TIPC are now mainlined
      in the kernel (this includes about half of the requirements).
    * Several others (like Synchronized/Atomic Time Across Cluster) can
      also be considered mainline with implementations such as NTP.
    * Every NEP provides a clustering solution, so whether the the
      requirements are actually met or not, clustered solutions are
      being provided.

Problem 1: TIPC is not a clustering "solution."  It contains all the
tools and protocols with which to build a clustering solution, but it is
not a clustering solution on it's own.

Problem 2: The open clustering solutions, such as LinuxHA and OpenAIS
may provide standard interfaces, but they may not have all the
capabilities provided by an underlying protocol like TIPC.  In other
words, you could say that you provided OpenAIS (or some other AIS
solution) and TIPC, but if these are not integrated, you would not be
meeting the CGL clustering requirements.

Problem 3: Not much analysis has been done in determining whether
distros actually meet the CGL clustering requirements or not.  For
instance, I have seen registrations where LinuxHA and TIPC are listed as
the implementation that addresses just about all the clustering
requirements.  Besides the lack of AIS interfaces, LinuxHA and TIPC have
completely different notions of "cluster membership".

Recommendation: At the summit in Japan, I would recommend that the CGL
Clusters document be re-visited in light of the SCOPE recommendations.
The CGL Clusters document should reflect more of a solution-orientation
rather than a capability-orientation.  Without the integration of these
clustering capabilities, an open clustering solution is just not
possible and registration against the individual capabilities is
meaningless.

John
>  
>
>   
>> -----Original Message-----
>> From: lf_carrier-bounces@xxxxxxxxxxxxxxxxxxxxxxxxxx 
>> [mailto:lf_carrier-bounces@xxxxxxxxxxxxxxxxxxxxxxxxxx] On 
>> Behalf Of Tim Anderson
>> Sent: Monday, November 26, 2007 8:12 AM
>> To: lf_carrier@xxxxxxxxxxxxxxxxxxxx
>> Subject: Slow in compiling the split info for 
>> this weeksmeeting.
>>
>> I am working on this during my travels and should have it 
>> emailed out when I
>> arrive in Tokyo.
>>
>>
>> --
>> Tim Anderson
>> MontaVista Software, Inc.
>> (480) 517-9060 
>>
>> _______________________________________________
>> Lf_carrier mailing list
>> Lf_carrier@xxxxxxxxxxxxxxxxxxxxxxxxxx
>> https://lists.linux-foundation.org/mailman/listinfo/lf_carrier
>>
>>     
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Lf_carrier mailing list
>> Lf_carrier@xxxxxxxxxxxxxxxxxxxxxxxxxx
>> https://lists.linux-foundation.org/mailman/listinfo/lf_carrier