RE: citrix and windows 2000 terminal services on clients --- do w

MSiebenschuh

Well Known Member
RE: citrix and windows 2000 terminal services on clients --- do w

Not to fan the flames but let me also put this out there (I am not a
consultant I am the CNC here).

Put simply there is no way a "farm" can be a single point of failure. In
your instance the "farm" hasn't failed it has simply slowed down. EVERYONE
is still working.

As far as the code is concerned - if you are maintaining the code on all of
your fat clients like you should they should all have the same code at the
same time - thus if one box has bad code they all should. We do a weekly
PROD update. With my 26 TSE's I can rename the current PROD directory,
deploy a new full package and I'm done dealing with a couple hundred users.
If there is "bad code" I can roll back in less than 10 minutes. Unless you
have very savvy full client users that won't be happening.

Maintenance is indeed much better with Citrix because it is simply less
boxes to take care of (even for a small business (2 boxes (total $50,000) or
100 ($100,000 for cheap PC's) you figure it out).

Beyond this you need to look at all the other benefits a Citrix farm
provides - in our case we have 3 divisions running 3 different software's
(One World, World, Custom)all of them use the farm. They are now all
supported by a central back office rather than 3 back offices. You can also
maintain all the other apps your company uses much more efficiently (i.e.
Word, Outlook, etc. we know how many service packs this stuff gets!).

After running our Financial users (about 100) on fat clients for 1 1/2 years
I have finally moved them all to Citrix - guess what - less phone calls.

That is just my opinion but I find Citrix to be easier to troubleshoot,
maintain, and update (for ALL users not just local users).

Mark Siebenschuh
 
Re: citrix and windows 2000 terminal services on clients --- do w

Not so fast.

YOU CANNOT PULL 50 USERS OFF OF 1 TSE AND PUSH THEM ONTO 1 SINGLE OTHER TSE
BOX WITHOUT PERFORMANCE IMPLICATIONS. Now, in your case, you may bring down
1 of your 26 TSEs down and move the 20 or so users over to one of the other
25 TSEs. Performance may not suffer. (Which makes me wonder how you could
bring 26 servers down one at a time during the day anyway????)

Your price calculation on hardware alone is WRONG. A decent Dell box (PIII,
256MB, 7.2 GB HD) runs me about $650 * 100 users= $65000. (This assumes all
of my users will need a new PC anyway). And you neglected the 3rd failover
server is an additional $25000. This brings the TSE solution (HARDWARE
alone) to $75000 which is exactly $10000 more than the fat client solution.
(And your assumption that each of your users won't need a PC anyway. NEVER
MIND the licensing premium you pay for the Citrix / TSE client licensing.)


With regard to the single point of failure on the code. If you have 150
users, chances are very good that each user or a group of users use a
discrete subset of all of the OneWorld applications. One bad ESU is not
necessarily going to affect all users. Its going to affect maybe 3 to 10
users. On the fat client solution, I can back out that change on a single
PC without any worries just by checking out a good copy of that app. I
can't do that on a TSE because you CANNOT CHECK OUT OBJECTS while other TSE
users are using any other app on the TSE. That is a single point of
failure.

I would argue that the majority of the IT people that are proponents of the
all TSE solution just do not want to or are incapable of properly managing
the OneWorld codebase. IT IS NOT THAT ONEWORLD NEEDS TO BE ON A TSE FARM TO
FUNCTION WELL.



----- Original Message -----
From: "MSiebenschuh" <[email protected]>
To: <[email protected]>
Sent: Wednesday, December 13, 2000 7:22 AM
Subject: RE: citrix and windows 2000 terminal services on clients --- do w
~~0:2583


> Not to fan the flames but let me also put this out there (I am not a
> consultant I am the CNC here).
>
> Put simply there is no way a "farm" can be a single point of failure. In
> your instance the "farm" hasn't failed it has simply slowed down. EVERYONE
> is still working.
>
> As far as the code is concerned - if you are maintaining the code on all
of
> your fat clients like you should they should all have the same code at the
> same time - thus if one box has bad code they all should. We do a weekly
> PROD update. With my 26 TSE's I can rename the current PROD directory,
> deploy a new full package and I'm done dealing with a couple hundred
users.
> If there is "bad code" I can roll back in less than 10 minutes. Unless you
> have very savvy full client users that won't be happening.
>
> Maintenance is indeed much better with Citrix because it is simply less
> boxes to take care of (even for a small business (2 boxes (total $50,000)
or
> 100 ($100,000 for cheap PC's) you figure it out).
>
> Beyond this you need to look at all the other benefits a Citrix farm
> provides - in our case we have 3 divisions running 3 different software's
> (One World, World, Custom)all of them use the farm. They are now all
> supported by a central back office rather than 3 back offices. You can
also
> maintain all the other apps your company uses much more efficiently (i.e.
> Word, Outlook, etc. we know how many service packs this stuff gets!).
>
> After running our Financial users (about 100) on fat clients for 1 1/2
years
> I have finally moved them all to Citrix - guess what - less phone calls.
>
> That is just my opinion but I find Citrix to be easier to troubleshoot,
> maintain, and update (for ALL users not just local users).
>
> Mark Siebenschuh
>
>
>
>
>
> --------------------------
> To view this thread, visit the JDEList forum at:
>
http://198.144.193.139/cgi-bin/wwwthreads/showflat.pl?Cat=0&Board=OW&Number=
2583
> *************************************************************
> This is the JDEList One World / XE Mailing List.
> Archives and information on how to SUBSCRIBE, and
> UNSUBSCRIBE can be found at http://www.JDELIST.com
> *************************************************************
>
>
 
Re: citrix and windows 2000 terminal services on clients --- do w

Who's lashing out? Not me. I am simply giving Melo the same informed
opinion from the CON side of the argument. I am trying to give him a broad
and experienced opinion. I think a large number of the posts in this forum
substantiate that OneWorld on TSE is not as well-heeled as some folks would
imply. All, Melo has to do is to look on one of the current threads in this
forum regarding TSE OCM mappings. For more fun, do a search on the
archives.

Re the second point: The expression goes: "THE DEVIL IS IN THE DETAILS". I
prefer the expression: "ALL GENERALIZATIONS ARE BAD."

Re the third point: more power to you.

Re the fourth point: (low self-esteem?). I did not call anyone a loser.
That's not nice and not appropriate for this forum. We all have are
strengths and weaknesses. And, I am not a consultant.

I have several of my 300 local users on TSE clients. All of my 150 remote
users are on TSE. Every OW installation is different and every user/client
is different. MY POINT IS THAT "YOU DO NOT HAVE TO HAVE ALL USERS ON TSE"
to run OneWorld effectively, which is what his vendor implied to him. Would
you not agree?


----- Original Message -----
From: "Mark Siebenschuh" <[email protected]>
To: "'JD Nowell'" <[email protected]>; <[email protected]>;
<[email protected]>
Cc: "Mark Siebenschuh" <[email protected]>
Sent: Wednesday, December 13, 2000 10:38 AM
Subject: RE: citrix and windows 2000 terminal services on clients --- do w
~~0:2583


> Ok this will be my last response to this board but I feel we need to show
> all points after this I will concede ;-).
>
> First point - I DID SAY THERE WOULD BE A PERFORMANCE HIT - I just made it
> clear that they would all be working. I don't take down 26 servers in a
day!
> I was just saying if I needed to I could ask my users to log off and in 10
> minutes I would have an entire company back to working just like the day
> before. (If I want to bring them all down in a day without any real impact
I
> just take a server out of the published app. Tell the users to log out and
> back in. Now the box is free for me to mod. When I'm done load balancing
> takes care of the rest.) Your fat client's wouldn't have a backup of the
> directory (unless you are pushing out your packages in a special way) so
> they would have to hack around checking stuff out (we don't allow our
users
> to do that and I really don't feel like going to a hundred machines and
> doing it for them).
>
> Second point - I wasn't really trying to get into detailed money calcs. If
> we wanted to get nit picky we could find really cheap TSE's and even
cheaper
> PC's I was being generic. We actually use winterms so there is no hardware
> issue. If a user has a problem with it they just get a whole new winterm.
> You might spend a few hours trying to figure out if it is an OS issue, a
OW
> issue, a bad HDD issue, a bad HDD controller issue, etc. In the end the
> initial outlay for TSE is probably more but the maintainability and lower
> maintenance give at least a break even with less effort.
>
> Third point - actually I check out stuff to TSE when absolutely needed (no
> this is not supported by JDE we even had to mod the OL app to do this -
> please don't go off on this fact since we all do what we need to in order
to
> make OW work for us)with all my users logged on and it works fine. But you
> do make a good point about not affecting all users with a code change. In
> the case of just 2 TSE's you would definitely be correct - in our case I
> have my financial users on certain boxes and our distribution users on
other
> boxes to prevent this very problem you say confronts the TSE's configs. In
> your example it sounds like at the end of a few months none of your fat
> clients have the same code base with all the checkouts to fix things.
>
> Fourth point - I guess the rest of us losers need to hire you to show us
> how to maintain a stable working system with a minimum of impact since we
> can't manage a "code base" - but wait that would make you a consultant
whom
> you seem to vehemently hate for some reason.
>
> My reply was made in an effort to give Melo an understanding of the
benefits
> of TSE in a large environment not to blast others on this list as lazy
know
> nothings (I have no idea why you are striking out like this). Some of
these
> smaller companies might have a CNC/DEV/BackOffice that consist of 2
people.
> Just because they take an easy to maintain route at a little more expense
> doesn't mean they are lazy they might be working harder than you or I.
Again
> I must reiterate the benefit of TSE as far as other business apps that are
> used in addition to OW, the low maintenance, and the ability to restrict
> users to what you need them to have (fat clients mean that users are
> installing 20 copies of AIM, the latest and greatest screen saver and who
> knows what other widget that goes out and blows a reg key. (Because they
> have to be local admins to pick up packages so that means they can install
> software). No my solution isn't optimal for all companies - which one is?
I
> think the key to this list is showing what you have and how well it works
> for you and letting others draw their own conclusions. With that I sign
off
> to further replies about this. Until the next hot topic...
>
> Mark Siebenschuh
> Enterprise Software System Administrator
> HP9000 (V-class/K-class/D-class)
> Oracle 8.0.5
> JDEB733 Base (in the middle of the XE upgrade)
> 26 Dell 6450's TSE/Citrix server farm
>
>
>
 
RE: citrix and windows 2000 terminal services on clients --- do w

Mark,

Well said. I think the point here is that Citrix with load balancing is a
high availability solution. Even if you loose a server you can still run
your business. Is performance worse? Maybe. There are also many arguments to
having a single point of failure. If I have a problem with a package do I
want to visit 100 machines or 1? The centralized administration can save you
hundred of man hours per year.

I am also baffled by the inference that if you use terminal server you do
not know how to manage the code. My personal feeling is that if you have 100
machines with 100 different copies of the code you are not a genius. You are
an idiot. PROD should be PROD, DEV should be DEV, regardless of the box you
are on. Developers are the only exception. If you are running multiple
copies of the code you are a fool. Good luck at your next job.

Checking out objects to terminal server is not a problem. Neither is package
deployment. You just need to be creative and use the system instead of fight
it.

Paul Ross
OneWorld Hired Gun
[email protected]
Phone: (513) 984-3093 x3097
 
Re: citrix and windows 2000 terminal services on clients --- do w

why push them all onto another single TSE ?

If I have 3 Terminal Servers to support my 100 users - (40 users per TSE is a sizing, by the way, I created) and one fails over - loadbalaning will place 1/2 of the users on one terminal server and the other 1/2 on the second terminal server. This means that I will have 50 users on each of the 2 remaining terminal servers. That is, of course, a 25% increase in the number of users - and I would expect a 25% degradation of performance. Likely I will still experience <1 second performance delay from the Citrix Server however. If 1 second is not acceptable to you - then 1/2 the users to get 1/2 second !

Remember - when I sized TSE we measured 1 second performance degradation - and for the benchmarks we found that 1second for each user was being exceeded for 80 users concurrently. We therefore halved that figure for our sizing methodology. Of course, it all depends exactly on what one is running - but 50 users should easily be supported on a terminal server, though a slight performance degradation will occur.

Its all very basic math. If I have 1000 users across 25 terminal servers (as in Marks case) - then if I lose 1 server, the remaining users will be balanced across the remaing 24 servers meaning that most servers will now have 42 users on them - rather than 40. Thats a very acceptable 5% increase and I'd bet will lead to less than a 5% performance degradation. More servers, more reliability and more fault tolerance.

Hope this thread will educate people !

Jon Steel                                       
Xe Upgrade Specialist - AppzBiz
 
Re: citrix and windows 2000 terminal services on clients --- do w

wrong wrong wrong

heres a configuration I just calculated for a VERY good OneWorld TSE on http://www.dell.com

Dell Poweredge 1550
Dual 1Ghz PIII pentium with 256k Cache
2Gb SDRAM (4 DIMMS)
No Keyboard/Monitor
2 x 9Gb Ultra 3 1" 10,000RPM SCSI on RAID 1
PERC3 Raid DCL RAID Card - 64Mb Cache
No OS, Intel Pro 10/100 NIC with Adaptive Load balancing

Total Price - $9,110 or $330 per month business lease

I'll bet that the above server will support 40 concurrent OneWorld users HACKING 5 line Sales Orders at an average of 1 order per minute easily. I'll also put the money on the fact that you'll only see the CPU's at 50% over time.

A lot different to your $25,000 eh ? What you configuring - Xeon Processors for a Terminal Server ? Hoping to increase your sales commission there ?

Heres a little something to think about - just "checking out an object on the fly" without TESTING it first is NOT a proper way to manage deployment. I certainly hope that you're not working on any of my customers ! If you deploy any code - whether its an ESU or custom code - without first FULLY testing them in a CRP Environment - then you deserve the downtime.

Jon Steel                                       
Xe Upgrade Specialist - AppzBiz
 
RE: citrix and windows 2000 terminal services on clients --- do w

to all who responded to my citrix/tse request for information, thank you
very much. indeed our team is learning a lot reading through the discussion
threads.

Melo
UNILAB
 
Re: citrix and windows 2000 terminal services on clients --- do w

Your changing Melo's requirements and previous hypothetical factors already
stipulated in previous posts. The requirement was for 150 users (not 100
and not 1000). Mr. Siebenbuch's assessment was that this would require 2
TSEs (quad processors (not dual), 4GB memory (not 2GB)) @ $25,000. I would
agree that 2 servers of this size could support 150 users. (The size of the
server you are referring to would support 40 users per TSE not 75). The
question is whether if 1 of the 2 servers was down, could you support the
other 75 users on the single TSE. My opinion is that NO you could not. You
would have to have a third server to absorb the load. I AM NOT ARGUING THAT
THERE IS NO SUCH THING AS LOAD BALANCING (WE ARE RUNNING IT AND IT FUNCTIONS
AS DESIGNED). Load Balancing does not absorb a load you don't have hardware
for. So throw another TSE (quad processors (not dual), 4GB memory (not
2GB)) into the calculation and another $25000, not $10000. My original
point was that the cost of TSE/Citrix/TSE servers does not provide enough
benefit in terms of helping to support the application LOCALLY. If you are
capable of managing the codebase, you should have no problem managing 135
fat clients for LOCAL users.

Now for 1000 users, the economics of support change. I think it is more
economical NOW to have 500 to 750 TSE local users (assuming Win 2K and
servers I have described above) and you could justify it in a cost/benefit
sense. Incidentally, up until Win2K and B7332, you could not get more that
40 users onto OneWorld TSE. Less than 1.5 years ago, your solution for 1000
users would have required 50 TSEs not 20. Even 20 TSEs is not real
practical. There have been too many problems with OneWorld on TSE, and I
know because I was out in the field, trying to support the real world
situations that JDE white paper writers don't seem willing to make reference
to.

My final word is that anyone may question our opinions in this forum. If
Melo would like to verify MY opinion, he need look no further than the
HONEST comments of contributers to this forum found in the ARCHIVES section
attesting to the challenges of OneWorld on TSE (most of whom are not
consultants or vendors).



----- Original Message -----
From: "altquark" <[email protected]>
To: <[email protected]>
Sent: Wednesday, December 13, 2000 4:52 PM
Subject: Re: citrix and windows 2000 terminal services on clients --- do w
~~2583:2662


> why push them all onto another single TSE ?
>
> If I have 3 Terminal Servers to support my 100 users - (40 users per TSE
is a sizing, by the way, I created) and one fails over - loadbalaning will
place 1/2 of the users on one terminal server and the other 1/2 on the
second terminal server. This means that I will have 50 users on each of the
2 remaining terminal servers. That is, of course, a 25% increase in the
number of users - and I would expect a 25% degradation of performance.
Likely I will still experience <1 second performance delay from the
Citrix Server however. If 1 second is not acceptable to you - then 1/2 the
users to get 1/2 second !
>
> Remember - when I sized TSE we measured 1 second performance degradation -
and for the benchmarks we found that 1second for each user was being
exceeded for 80 users concurrently. We therefore halved that figure for our
sizing methodology. Of course, it all depends exactly on what one is
running - but 50 users should easily be supported on a terminal server,
though a slight performance degradation will occur.
>
> Its all very basic math. If I have 1000 users across 25 terminal servers
(as in Marks case) - then if I lose 1 server, the remaining users will be
balanced across the remaing 24 servers meaning that most servers will now
have 42 users on them - rather than 40. Thats a very acceptable 5% increase
and I'd bet will lead to less than a 5% performance degradation. More
servers, more reliability and more fault tolerance.
>
> Hope this thread will educate people !
>
> Jon Steel
> Xe Upgrade Specialist - AppzBiz
>
> --------------------------
> Visit the forum to view this thread at:
>
http://198.144.193.139/cgi-bin/wwwthreads/showflat.pl?Cat=&Board=OW&Number=2
662
> *************************************************************
> This is the JDEList One World / XE Mailing List.
> Archives and information on how to SUBSCRIBE, and
> UNSUBSCRIBE can be found at http://www.JDELIST.com
> *************************************************************
>
>
 
Re: citrix and windows 2000 terminal services on clients --- do w

children,
Please quit YELLING at each other.

NT 4.0 SP5, SQL 7.0, One World B7321 SP12.4
 
RE: citrix and windows 2000 terminal services

Thomas,

1. Have you implemented the Xe Java thin-client solution? Or for that
matter, anybody else on the list?
2. Is it ready for prime time, i.e. reliable?
3. How would be the configuration for Melo's requirements: "out of around
150 (clients), 135 will be on the lan and 15 will be remote"?

Thanks,

Nico Wohldorf

WOHLDORF SYSTEMS, INC.
25 San Miguel Avenue, Suite 2204
MPO Building, Ortigas Center
1600 Pasig City, Philippines

Direct/Voice: +63.2.633.1855, ext. 215
Facsimile: +63.2.638.5646
Cellular: +63.918.903.6411

E-Mail: mailto:[email protected]

URL: http://www.wohldorf.com

"Excellence in Integrated Management Systems"
 
Re: citrix and windows 2000 terminal services

Have you actually had experience with this? If so, I think there are many listening who would like some particulars on your experience.

>>> <[email protected]> 12/14/00 8:14:15 PM >>>
From: "Stirling, Thomas"
<[email protected]>
To: "'[email protected]'" <[email protected]>,
"'[email protected]'"
<[email protected]>
Subject: RE: citrix and windows 2000 terminal services
on clients --- do w
~~2583:2694
Date: Thu, 14 Dec 2000 15:53:22 -0500

Have you looked into Xe and the Java Thin-client answer?
It's quite
exciting. With a third party topology (Web-Boost) it
out-performs Citrix
over WAN. The great thing is that you don't have the
outlay of TSE Servers.
 
RE: citrix and windows 2000 terminal services

Installed Xe and a couple of clients I'm with are looking seriously at
the numbers. So far it looks good. There's a great article on "WWAT"
on the knowledge garden under "findings."
 
RE: citrix and windows 2000 terminal services

Nico-
Can't say I have; however, from what I'm reading
it looks very promising. I have a couple of clients
that are about ready to go with it.
 
Back
Top