Why does EA access EAP file when checked out?

[ebeb]

Hi!

Everytime I have to do home office, I'm connected via VPN to our company network.

The EA project is a EAP (Jet DB) on a shared Samba drive. And I'm using SVN as VCS for packages.

Now, I've checked out a package. Everytime I click on an element it takes 5-10seconds to just load the element note.

I just did a quick Wireshark capture and realized that there is massive traffic going on. Its about 400 packets, 50% TCP traffic (Jet DB) and 50% netBIOS/SMB file traffic.

So I'm wondering, why is this necessary?

Once I checked out the package, I'd expect all information within the XMI or in RAM after loading it.

Is there a way to get EA to somehow cache information that has already been read?

Thanks, Jan

[Geert Bellekens]

Jan,

Have you looked at the WAN-optimizer?
I'm not sure it if only applies to DBMS, or also to shared eap files though.

Geert

[ebeb]

Wow...EA is full of surprises! I didn't know that.

But, according to the EA help, it only works for DBMS, unfortunately.

So still, my question remains, why does EA has to access the DB for checked out packages? Isn't it a local-only thing?

[Paolo F Cantoni]

Wow...EA is full of surprises! I didn't know that.

But, according to the EA help, it only works for DBMS, unfortunately.

So still, my question remains, why does EA has to access the DB for checked out packages? Isn't it a local-only thing?

Hi Jan,

The problem is that the EAP file is on the wrong side of the VPN.  EAP files (accessed with native Jet drivers) are designed by Sparx to be local to your system.  You can use them in shared configurations as you are, but you have to be aware of the limitations.

The correct way to think of EA repositories is to think of them operating in two modes:  Native Jet and Server (as in Client-Server).  Not Jet and DBMS (Geert and Sparx - they're all DBMSs - all that's changing is the drivers).

Only EAP files can operate in Native Jet mode.  Consequently, if you are using native Jet access, the traffic between the applicaiton and the disk will be the same regardless of where the EAP fiel sis located.

To access an EA repository in server mode, you need Corporate edition or better.  We've done extensive testing on various configurations of repositories and editions and now believe we have a good understanding of what's going on.

So, if you create a separate access path to the same, Samba based, EAP file, but using an ODBC connection, you should be able to access the EAP file in server mode.  Consequently, you should be able to use the WAN Optimizer as Geert suggests.  I haven't used it myself, but from the reports by forum members I see no reason why it shouldn't work via the ODBC connector.

Also, from the way in which you expressed the original problem, you may have a misunderstanding of the way package check-in/check-out works for EAP files.  Just as with server-based access, Check-in involves the export of the package from the EAP file as an XMI. followed by the check-in of the exported XMI to the VCS.  Check-out involves the reverse process:  Check-out the package XMI from the VCS, import the checked-out XMI into the EAP file.

Consequently, in your original scenario, check-out involves the transmission of the XMI to your local machine over the VPN, immediately followed by the storage of the imported package information back across the VPN to the EAP file on the Samba drive.

HTH,
Paolo

[ebeb]

Paolo,

thank you for your detailed reply!

To access an EA repository in server mode, you need Corporate edition or better.

Got the Corporate Ed.

We've done extensive testing on various configurations of repositories and editions and now believe we have a good understanding of what's going on.

So, if you create a separate access path to the same, Samba based, EAP file, but using an ODBC connection, you should be able to access the EAP file in server mode.  Consequently, you should be able to use the WAN Optimizer as Geert suggests.  I haven't used it myself, but from the reports by forum members I see no reason why it shouldn't work via the ODBC connector.

I'm not really sure what you mean by "separate access"? Do you mean I should install the WAN Optimizer to access the EAP file and then configure EA to connect to the WAN optimizer?

I'd appreciate if you could explain this a bit detailed how to do it or to point me to the appropriate doc.

Also, from the way in which you expressed the original problem, you may have a misunderstanding of the way package check-in/check-out works for EAP files.  Just as with server-based access, Check-in involves the export of the package from the EAP file as an XMI. followed by the check-in of the exported XMI to the VCS.  Check-out involves the reverse process:  Check-out the package XMI from the VCS, import the checked-out XMI into the EAP file.

Consequently, in your original scenario, check-out involves the transmission of the XMI to your local machine over the VPN, immediately followed by the storage of the imported package information back across the VPN to the EAP file on the Samba drive.

Ahh, yes, I think I had a misunderstanding. Thanks for clarifying that!

Jan

[Paolo F Cantoni]

Paolo,

thank you for your detailed reply!

[size=18]...[/size]

I'm not really sure what you mean by "separate access"? Do you mean I should install the WAN Optimizer to access the EAP file and then configure EA to connect to the WAN optimizer?

I'd appreciate if you could explain this a bit detailed how to do it or to point me to the appropriate doc.

[size=18]...[/size]

Hi Jan,

At present you probably have a UNC (or a mapped drive) path to the EAP file on your EA Start Page.  Go to the Start Page and use Connect to Server... to create another, ODBC, connection to your EAP file using the Jet 4.0 OLE DB provider.  That will establish a server based access mechanism for you.

Make sure you're able to read and write to the repository using this, new, access mechanism.

Now (that you've proved there's no problem) you can install the WAN optimizer and configure it appropriately.  As I mentioned, I've not used it but it would appear to work.  Just follow the instructions.

When you're at work, use the direct access mechanism, when you're remote, use the ODBC connection.  You could use the ODBC connection all the time, but it's probably faster to use Jet direct.
HTH,
Paolo

[ebeb]

Ok, at first I had some problems because the driver config page asked for an Access MDB file and username and PW.

But actually, I changed the file filter to *.* and then selected the EAP file. Username remained unchanged with no PW.

Then, after quite a while, the project opened, and everything works as expected (but still slow).

Then I downloaded the WAN Optimizer, but it says:

Code: [Select]

Note: Optimization is not supported for .EAP filesPlus, it's only available as a Windows Service. So it won't work on our Linux servers.

What now?  :-?

[Paolo F Cantoni]

[size=18]...[/size]
Then I downloaded the WAN Optimizer, but it says:

Code: [Select]

Note: Optimization is not supported for .EAP files

They may have meant EAP files using Jet directly.  It may not be the case with ODBC connections in server mode - since it should be using ODBC protocols.  Someone else will have to comment on that.

Plus, it's only available as a Windows Service. So it won't work on our Linux servers.

What now?  :-?

Unfortunately, as I said, I've not used the WAN optimizer so I can't help there.

If the WAN optimizer can access EAP files in server mode, then there may be a solution; but we'll need to get that information from either Sparx or a more WAN optimizer knowledgeable user.

Sorry,
Paolo

[ebeb]

Hi,

I just found the easiest and most obvious solution: Replication.

I just set the shared EAP as master, create a replicas, and sync them.

NOTE: Make sure not to have any other EA instances having the same EAP open  when trying to set it as master. This will lead to VERY strange behavior.  :-/

[Geert Bellekens]

Jan,

Just a warning note,

We recently had a topic from a user who had gotten himself into a very big mess using replication.
I think it works fine if you are the only one working on the model, but it is not suitable for multiple users.
If I were you I would look into either

• setup EA on a "real" database and use the WAN-optimiser
• control the model using a VC and use the VC to manage the syncronisation.

Geert

PS. You can do both too  

[ebeb]

Oh...

That was too good to be true, wasn't it?  

Yeah, using a DB and VC was my second option, I thought about it, but I chose replicas because of two reasons:
1) Replicas are still much faster that using a DB with WAN optimizer because the WAN optimizer says it good for bandwidth limited WAN and won't help with high latency which is IMO the problem in our case. I measured the ping from my home office which takee 200-400ms round trip.
2) It just is too time consuming to set up a DB, create a windows VM to satisfy the non-linux WAN optimizer and to move the data from the EAP file into the DB.

[EDIT] Could you please point me to the thread where that guy messed up the replicas/master? Link, name of the topic? The forum search doesn't work well

[Geert Bellekens]

that was this one: http://www.sparxsystems.com/cgi-bin/yabb/YaBB.cgi?num=1259789485/

Geert