Hannover and Citrix
by Volker Weber
Yesterday I talked to a friend about the difficulties of running Lotus Notes in multiuser environments. Traditionally Notes has installed in a directory /Notes directly off the root directory with a directory /Notes/Data to hold its data. The directory moved around, first to /Lotus/Notes, then to the /Programs tree. If you tried to run multiple user you moved the ./Data directory to the user' home directory, still sharing the program directory. You just needed to make sure that you were calling the correct notes.ini.
If you install the Notes Plugin for the Workplace Managed Client (WMC) on Linux, you will find that it wants to install to the user's home directory, thus allowing you to run more than one instance on the same machine. At first look this seems to violate the policy of installing shared components outside of the home directory. You will find however that only a minor part of the Rich Client Platform is shared. All the plugins are provisioned for each individual users according to policies on the management server. They are data.
This leads us to an often overlooked issue with the WMC. As you may be aware of, the WMC runs inside a Java Virtual Machine. This VM runs in the user space and from the outside it looks like 'all data'. This is not a problem as long as you are running a single instance of a WMC. As soon as you run more than one instance, you need to load each of them into a separate memory space. There are only data segments and no code segments that could be shared. At this point you can kiss your plans to run WMC or "Hannover" on Citrix goodbye.
There you have it: A good reason why Lotus will continue to deliver "Notes Classic".
Tags: citrix lotus notes hannover workplace wmc
Comments
Good point, Volker. Although I believe the Microsoft JVM supports JIT compilation, it's not clear to to me that this would make any difference.
Alleviated to some degree by this
The widespread usage of CITRIX will be definitly a big issue for WMC implementations.
CITRIX has gone mainstream in corporate environments despite the fact that you will need 1 Server CPU per 10 concurrent Users.
CITRIX is popular because it allows to transform the FatWindows Client World to a "controlled" server environement.
Although is it not smart to implement huge serverfarms but it is cheaper than migrating current Windodws applications.
WMC is about changing the computer paradigm from Fat Windows Clients to webbased SmartClient runing on LINUX or Windows wiht efficient use of todays client hardware ressources.
I haven't really thought this through yet, but what happens if there are noExecute flags in the memory for data segments? Does the JVM by-pass this? Would this be a possible security issue?
This is a day one issue with Java that's only now being partially addressed. When bytecode is loaded from .class files, the in-memory version isn't the same as the disk version which makes it difficult to put in read-only shared memory pages. This was also an issue with ROMable Java code. Some vendors have adopted their own solutions for that.
Stefan's link to the class data sharing support in 1.5 is a partial solution but it's specific to Sun's JVM. WMC uses IBM's JVM. I'm not sure if IBM's JVM supports class data sharing. And even if it does, the fact that each user on the same machine gets his own copy of the plugins makes sharing moot.
I wonder where the Notes DLLs are put in the Hannover environment? If they're buried in per-user plugins as well, you won't even get sharing for native code.
I figured this problem out a while ago, hence my question on here...
http://www.edbrill.com/ebrill/edbrill.nsf/dx/notes-fud-making-or-not?opendocument&comments#anc1
Post a comment
Recent comments
Volker Weber
on No surprises at the PhilNote at 21:53
Ben Poole
on No surprises at the PhilNote at 21:07
Pieterjan Lansbergen
on No surprises at the PhilNote at 21:03
Jan Tietze
on No surprises at the PhilNote at 20:48
Ben Rose
on No surprises at the PhilNote at 15:14
Ben Rose
on No surprises at the PhilNote at 14:46
Philipp Sury
on No surprises at the PhilNote at 13:40
Sascha Reissner
on No surprises at the PhilNote at 13:28
Ben Rose
on No surprises at the PhilNote at 12:10
Colin Williams
on No surprises at the PhilNote at 09:43
Adalbert Duda
on No surprises at the PhilNote at 09:33
Volker Weber
on No surprises at the PhilNote at 09:28
Jan Tietze
on No surprises at the PhilNote at 08:53
Ben Rose
on No surprises at the PhilNote at 08:14
Volker Weber
on No surprises at the PhilNote at 04:50
Amy Blumenfield
on No surprises at the PhilNote at 01:36
Richard Schwartz
on Peter O'Kelly: I am now a Microsoft employee at 02:49
Hanno Zulla
on Ente Chop Sui at 22:32
Kristof Doffing
on Ich helfe gerne aus der Autokrise at 15:03
Bruce Lill
on IBM is not going to send out a press release at 00:31
Thomas Cloer
on Events of the last few days at 12:38
David M Taylor
on Two screens or one? at 03:27
Henrik Heigl
on Events of the last few days at 18:51
Sven Scholz
on Two screens or one? at 16:44
Volker Weber
on New poll: how do you expect to be a year from now? at 21:53
Click here to subscribe
