Re: Configuring ERChangeNotification/JMS Help needed
Re: Configuring ERChangeNotification/JMS Help needed
- Subject: Re: Configuring ERChangeNotification/JMS Help needed
- From: Dov Rosenberg <email@hidden>
- Date: Thu, 27 Oct 2005 14:21:24 -0400
- Thread-topic: Configuring ERChangeNotification/JMS Help needed
We ended up changing the JMS Changenotification framework to be a bit more
lightweight than the one in Project Wonder. Instead of sending around
Snapshots we simply send the globalids and invalidate the individual Eos
that are affected on the local instance. Our app is such that the initial
refetch is not too bad. We are using JMS to signal each instance that their
local copy has been updated or deleted and to update their . New Eos get
picked up as a matter of course.
We have noticed a performance hit, but it hasn't been too bad yet. We are
investigating the removal of JMS.
Dov
On 10/27/05 2:07 PM, "Chuck Hill" <email@hidden> wrote:
> Hi
>
> On Oct 27, 2005, at 10:37 AM, Wolfram Stebel wrote:
>>> On Oct 26, 2005, at 5:13 AM, Wolfram Stebel wrote:
>>>
>> ...
>> As a result of all mails, i removed JMS and ERChangeNotification
>> very fast
>> from my project.
>>
> :-)
>
>>
>>>> But thats a minor reason!
>>>
>>> See David's comment, "If you ever use a lot of instances it will slow
>>> down everything". This might not be the best way to get you where
>>> you want to go.
>>
>> JMS was announced to be fast...
>>
> As far as I know it is fast. It is EOF synchronizing the cache and
> all that extra fetching that is not fast.
>
>
>>>>
>>>> 2. I have 3 Applications working on the same database, i.e. they
>>>> might
>>>> change the same data entries concurrently.
>>>>
>>>
>>> So deal with it. This framework will _not_ eliminate the need to
>>> handle concurrent editing issues. It will just _reduce_ the
>>> frequency of occurrence. ChangeNotification framework or not, you
>>> still have to code for these.
>>>
>>
>> I have one critical point in the moment, which i reprogrammed now.
>> I thougt
>> of a general and final solution, but it seems to be a little, eeeeh,
>> optimistic (locking)...
>>
> Optimistic locking good.
>
>
>>>>
>>>> locations. When those change, the customer needs direct feedback of
>>>> the change.
>>>>
>>>
>>> What does direct feedback mean? They need a message to popup? An e-
>>> mail? Or just that they need to see fresh data.
>>>
>> Data should be updated in the second application, so that a refresh
>> in the
>> browser will reflect the change...
>>
> You probably don't want to do that. It would mean that every page
> request could potentially result in fetches to the database.
>
>> For now i have to re-login, thats semi-optimal.
>>
>>
>>>> 3. I asked on this list how to handle this issue, and i was told to
>>>> do what
>>>> i'm now implementing.
>>>>
>>>> So, what other tips do you have?
>>>>
>>>>
>>>
>>> 1. Implement code to handle optimistic locking failures
>>>
>> So, why turn optimistic locking into pessimistic?
>>
> I don't understand your meaning.
>
>
>>
>>> 2. Determine the data freshness requirements and implement refreshing
>>> fetches and new EC creation to meet those needs.
>>>
>> Normaly i (re)fetch depending on current application state, but
>> some basic
>> data like a logged in client and person...
>>
> ... I would not want to change while they are logged in.
>
> Chuck
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden