[OOTB-distro] Clustering/load balancing Alfresco Commmunity

[Andreas Steffan]

On 05/06/2015 03:52 PM, Torben Lauritzen wrote:
> On 06/05/2015, at 12.18, Andreas Steffan <a.steffan at contentreich.de> wrote:
>
>> On 05/06/2015 10:57 AM, Torben Lauritzen wrote:
>>> The conclusion is that the various clustering things that are used in these systems are in some way incompatible; perhaps it is necessary for them to be in absolute sync and this cannot be guaranteed by the different strategies used in the available clusters.
>>>
>>> So my question is: what do we need to cluster? I guess an idea of the use cases involved is most useful.
>>>
>>> We have a client running a clustered setup on Enterprise (which they had to bauy elsewhere). They have one share instance, a load balancer in front of two Alfresco repo instances with two solr instances, and a shared database and filesystem. This seems to work reasonably well.
>> Essentiallly, Alfresco "native" clustering cababilities boil down to
>> transaction safe jvm memory caches sychronized between nodes. Hence, it
>> reduces the amount of network roundtrips. IIRC, this applies to both,
>> the repo and share. Getting this right is quite a bit of effort.
>>
> What do you mean, when you say Alfresco? And what do you mean by native? And why would you want “transaction safe jvm memory caches sychronized between nodes”?

Initially, you said "clustering would be a valuable addition to the CE"
and that Alfresco (the company) might not  be pleased if "we" succeed.
Hence, I concluded you consider (re-)building the functionality which
alfresco removed just before 4.2.a.

However, when they talk about clustering it is that "transaction safe
jvm memory caches sychronized between nodes" they removed from the CE
they have in mind.  Have a look at
http://docs.alfresco.com/4.1/concepts/ha-intro.html

  

>> Are you sure these capabilities address your painpoints?
>>
> This is kind of a strange question - it is like being offered a solution, and then afterwards being asked what the problem was :-) The issue is, that we need to be able to assure our customers that Alfresco Community will scale with regard to the number of concurrent user, whether it is 1.000, 5.000 or 20.000 users.
>

The technical solution really depends on a lot of  things. Due to the
effort involved, before coming up with the native clustering solution, I
would first make really sure it addresses your needs. E.g. user requests
are smacking down your RDBMs and there is no other, easier way to scale
up. If  your customer wants  assurance, I would suggest you set up a
load simulation and find the bottlenecks  first.

cheers
Andreas 

-- 
Andreas Steffan

Achter Billing 14
22399 Hamburg
Germany

skype: contentreich
M: +49 1793903615
T: +49 40 23943542
F: +49 40 23943542

http://www.contentreich.de

Contentreich : Alfresco ECM, Clojure, Groovy und WordPress - aus Spaß und für Geld