[ale] Advice

Jeff Hubbs hbbs at comcast.net
Wed Jan 4 09:08:18 EST 2006


"Yes, I know that the M$ would probably crash more, but going off the 
theory of many small crashes vs one large crash is more digestable"

Cough, gasp, choke. Seeing as the "many small crashes" you think would 
be so much better would involve your mission-critical systems, I'm 
hoping that you guys are getting comfortable with the idea of a life 
without sleep.

The way I see it, your company's IT is already halfway in the dumper and 
your software vendor (PDS) is going to complete the deal for you, shut 
the door, and cotter-pin it shut behind you. The consolation here is 
that you're in the same boat as almost everybody.

What I hear you telling us is that your RS/6000-based app has grown an 
alternate Windows-based front end, probably for no other reason than to 
have a Windows-based front end. I don't automatically accept that such a 
thing is more efficient or effective than the character-based original 
(which I assume you'd ordinarily run on Win32 via terminal emulation), 
but apparently someone at your shop did. But, the real issue with the 
vendor is that it's probably killing them to be RS/6000-based; no one 
really wants to add an RS/6000 and AIX where neither existed and lots of 
shops are going to be looking for ways to get *off* of them if they have 
them because they're expensive and arcane.

So, let's assume that you're stuck with at least *some* Windows on 
client and server because of app choices that have already been made. 
The question becomes, what to do about your mail system. I can't find 
such a thing as an HP ProLiant DL180, but their low-end 1U server, the 
DL140, can be had with two 3.6GHz Xeons and two 146GB drives; I have a 
real difficult time understanding how that can't manage to serve e-mail 
for a 50-seat shop (granted, I'd like to see a better drive arrangement, 
i.e., hot-swap, but this speaks to the issue of buying hardware with 
built-in constraints and proprietary components). Is there a reason why 
you can't buy what you actually need, as opposed to what HewDell offer?

There are too many variables here for simple better/worse determinations 
based solely on Windows vs. Linux. I dare say that an Linux-based e-mail 
operation *can* work *far* better than a Windows-based one, but the 
people involved would have to know fundamentally how to *make* it far 
better; *that* aspect of it doesn't come from Linux or OSS but from the 
talent of the people involved. The issue here is WinHewDell nullifying 
the talent.

I also question this thin-client business. As much as the concept is and 
was cool, the simple fact of the matter is that there is nothing cheaper 
and more ubiquitous than your basic commodity-part PC. You can *use* 
such a PC as a thin client in either Lin- or Win-world, but in doing so 
you're running back to a big-heavy-metal-in-locked-room scenario. This 
is not inherently bad on its face, but ask yourself if the heavy metal 
and software that you're going to be using has the kind of 
it-will-never-die well-deserved reputation that was once associated with 
the IBM S/370 and DEC VAX hardware and corresponding OS software that 
would be associated with such a role.

I really can't recommend that you go one way or the other, but I do 
recommend that you decide whether you're going to let the industry and 
"industrial correctness" (that is to say, the same logic that leads 
people to buy Dell or HP servers without question) control your 
computing or whether *you* are going to control your computing.

One thing you might want to do is get together with your 
line-of-business managers and ask them, how many hours of e-mail outage 
will you accept per year and how long is the longest single outage you 
will accept? Take their answer seriously and see if you can "build" to 
that through any means necessary. If you break those limits in practice, 
then you have an automatic mandate to fix the problem. It may well be 
that not even Linux on commodity server hardware can get you there 
without clustering.

Please don't make the mistake of equating "downtime" with "lost money." 
"Lost money" is money you had in your hand and then ran through the 
shredder. Money "failed to be made," which is what downtime actually 
corresponds to, is not the same as "lost money." Regarded in this way, 
downtime is an obstacle to making as much money as you theoretically could.

It also seems to me that this is perhaps more a human resources problem 
than a technology problem. One or more persons' knowledge envelope 
sharply constrains this company's ability to research and implement 
effective solutions. Whatever happens, this place should not be left 
with something that no one knows how to run, fix, and improve.

Anyway, none of this passes for a comprehensive answer to your 
question(s) but maybe I've said something here that might be helpful.

Jeff


Rob Fauls wrote:

> Ok, so ive been trying to stay away from asking you guys on here 
> because that?s not really what you are all about, but I?ve been trying 
> to arrange a few different things for my company and im faced with a 
> difficult answer. The whole thing, without going into huge detail, as 
> I will below, boils down to windows or linux for the mail server 
> keeping in mind that the company is not currently staffed with anyone 
> that can support linux, they don?t want to outsource work, and there 
> isn?t time for us to send the tech to training.
>
> That was the short version, here?s the medium/long version. We have a 
> trucking company that is based in Atlanta, but operates throughout the 
> country. We have a software package that currently runs off a mixed 
> unix/windows system. The main database and character version of the 
> software runs off a RISC 6000, while the windows version operates off 
> a windows 2003 server. Fyi, when I talk about the two systems, the 
> unix system is the backbone of everything, the windows system simply 
> pulls the database off the rics. On the client side, there is a 
> character-based telnet login, and on the windows side, a gui-based 
> program that runs on the server through remote desktop/terminal 
> services. There have been threats/promises that the unix side will be 
> done away with and the system will only be updated on the windows side 
> after a period of time, thus far, PDS, the software manufacturer, has 
> not removed support for the unix side, and they have not stated a 
> specific time that they will discontinue support. They have only said 
> that it will be done because the new system is windows, not 
> unix?bottom line (nothing to do about this, its economics, so moving 
> on) As a company, we are considering a move to thin-clients that will 
> login to a 2003 server. We have to stick with windows here because a 
> good portion of our software simply wasn?t made for anything but 
> windows and the people that works for us don?t want to learn a new 
> system or new software and it isn?t worth pushing them over. 
> Considering the thin clients, I think it?s a good move to get us on 
> board with the windows 2003 server for the PDS server. Chapter 2: The 
> other half of this, and the one that im really concerned about, is the 
> server that we are about to purchase. We have been looking around for 
> a number of months at possibilities and options for our mail server, 
> and started to go with a linux machine to begin with, but because we 
> bought the wrong system (an HP DL-180 instead of a DL-380?the 180 
> didn?t have the power we wanted out of it or the raid function we 
> needed) the project was questioned and there were arguments made that 
> we should stick with windows because there wouldn?t be nearly as much 
> of a learning curve associated with it as there would be with a linux 
> system. I understood where they were coming from. Their current 
> full-time IT guy isn?t trained in linux and to be frank, is limited in 
> his general knowledge of networking altogether. He is great for what 
> they do now, he knows pds back and forth and can fix the day-to-day 
> problems that we have. He has trouble learning new tricks, though?old 
> dog. I feel that this issue can be fixed with simple training, but I 
> am unsure whether or not they will want to go through the ?hassle? of 
> training and retraining a linux guru as new things come out. I 
> understand the problem there is that the company stagnates and will be 
> left to drown as new tech comes out and they aren?t able to integrate 
> it because they didn?t keep up with the changing times. In any case, 
> now, the contract was handed off to the company that is doing the 
> setup/install/qos for our VoIP system. They have quoted out a system 
> that comes out to approx 10k for a dl-380, 3x36gig 15k rpm HD, 4x512 
> ram, and a 3.0 intel along with M$ smb server 2003 with exchange with 
> 35 licenses. For a M$ product, its great. The question is, do we need 
> to go with this system or would we be better off switching to a linux 
> server and using open source to get all this done. As far as I am 
> concerned, I think that setting aside the fact that we currently have 
> nobody on site to support it, there isn?t a con to linux, but the 
> simple fact that 95% of out business is done over email confirmations, 
> if an email server goes down, another has to be able to pick up the 
> slack without fail. As far as I see, the advantages of windows are 
> that we will be able to utilize the remote desktop feature to be able 
> to utilize thinclients, but with pds we already have this. We also 
> have the ability to integrate the blackberry enterprise server and to 
> synchronize schedules and setup meetings, and all the other ?cool 
> things? that exchange does that I haven?t seen from a single package 
> in linux. Maybe im blind. Maybe I don?t know what im talking about, 
> but I guess that?s why im coming here in the end, because I have 
> totally confused myself and im too new at linux to be able to give 
> them a straight answer. The biggest thing stopping me, as I think ive 
> said enough that you probably could quote me without saying it is that 
> im nervous about leaving them without an on-site person to handle the 
> email server and if they loose an hour, they have lost a huge chunk of 
> money that could easily offset that months profits entirely. Yes, I 
> know that the M$ would probably crash more, but going off the theory 
> of many small crashes vs one large crash is more digestable. The 
> contract was going to be signed on wed (today/tomorrow, depending on 
> when you read this) but it has been put on a hold until I talk with 
> some of the people tomorrow morning. I know a guy that is willing to 
> help setup the server and will get everything running on it, the only 
> problem is that he already has a full time job, so I cant count on him 
> and I don?t have anyone else around?maybe its time to look at the 
> employment mailing list that I seem to remember ale having? **pulls up 
> mozilla** I think ive prattled on enough to make everyone?s ears ache 
> and heads explode. Thanks for any input, guys.
>
> -Rob
>
> P.S.- Still offering that ftp site, just still don?t know how I should 
> setup everything or what to tell you guys.
>
> Equipment that is in the office or will be (to give an idea of the 
> hardware we have if it will change anyone?s decision?don?t think it 
> will, but as I said, I don?t know what to do so im kind of throwing 
> everything I can think to say at your guys)
>
> 2x cisco 3560 48-port
>
> 1x cisco 1840 router
>
> ~35 workstations (win XP)
>
> ~35 ip phones (cisco 17-something)
>
> RISC 600 server (PDS database/old pds system)
>
> Windows 2000 server (not sure what it does, to be honest)
>
> Windows 2003 server (Windows PDS)
>
> 10 monkeys in a room with only 8 typewriters
>
>------------------------------------------------------------------------
>
>_______________________________________________
>Ale mailing list
>Ale at ale.org
>http://www.ale.org/mailman/listinfo/ale
>




More information about the Ale mailing list