Mass Attribute Updatage

Hey all, 
(yes I know updatage isn't a word)! 
 
I'm writing a linux c script which updates eDirectory attributes in real 
time. What's your input on the best way to handle a large queue of 
attribute updates? 
 
Should I use POSIX threads and establish several connections to the 
server? or should I create a queue and send the updates every thirty 
seconds? Obviously I don't want to have too great network or processor 
utilization... 
 
Possibly, some form of partition/replica solution? Having a filtered 
replica on the server running the script? 
 
Thanks in advance all, 
Regards, 
 

jgaldes@bigpond.com wrote:

> Hey all,
> (yes I know updatage isn't a word)!
>
> I'm writing a linux c script which updates eDirectory attributes in real
> time. What's your input on the best way to handle a large queue of
> attribute updates?
>
> Should I use POSIX threads and establish several connections to the
> server? or should I create a queue and send the updates every thirty
> seconds? Obviously I don't want to have too great network or processor
> utilization...
>
> Possibly, some form of partition/replica solution? Having a filtered
> replica on the server running the script?

How many attribute updates are you planning to do?

Whatever the number, NetWare will almost certainly cope ok, but users might
notice a performance degradation for a while if you do this during work
hours.

Schools and universities often create thousands of users in a single run of
a program with no delays introduced to reduce the impact on the server. In
addition to creating the objects and generating public and private keys
(time consuming stuff), they may be setting up to 20 or so attribute values
per user.

John
DevServ SysOp 24

Depending on what you want to do with this data, you may want to look
at flagging each attribute with the "Per Replica" Attribute Type flag.
This
stops the value being replicated around the replica ring - so reducing the
network traffic. There is also a "Schedule Sync Never" flag that
stops
the updating of an attribute, causing a sync of the replica.

As for the design of your code, your best bet maybe to have a single
thread handling the updates from a queue. This way your work will
be interleaved with any other work being sent to the server.

Roger Thomas, Dev SysOp 22

<jgaldes@bigpond.com> wrote in message
news:BJVNd.1909$jd1.299@prv-forum2.provo.novell.com...
> Hey all,
> (yes I know updatage isn't a word)!
>
> I'm writing a linux c script which updates eDirectory attributes in real
> time. What's your input on the best way to handle a large queue of
> attribute updates?
>
> Should I use POSIX threads and establish several connections to the
> server? or should I create a queue and send the updates every thirty
> seconds? Obviously I don't want to have too great network or processor
> utilization...
>
> Possibly, some form of partition/replica solution? Having a filtered
> replica on the server running the script?
>
> Thanks in advance all,
> Regards,
>
> James Galdes