To Enerprise Architects...

I would like to hear from the experienced IT folks, if you had unlimited budget and all the skill sets necessary which hardware, OS , DB and middleware would you implement a new EnterpriseOne 8.12 and why ?

Secondly, if you knew that the next step would be to upgrade to Fusion Applications would your answer to the first question change if so to what ?


Some examples:
IBM pSeries, AIX, Oracle, Red Stack (Oracle OAS..)
Dell, Windows, MS SQL, Blue Stack (IBM Websphere..)
IBM iSeries, OS400, DB2/400, BlueStack.
HP(Intel Server), Linux, Oracle, Red Stack.


It is my understanding that "currently" the most common implementations are (AS400, DB2/400, BlueStack ) due to World legacy and existing AS400 skillset, and Intel, Windows, MS SQL with Blue Stack due to cost (cheaper).

Thanks in advance for your replies.

Unlimited budget? All the bells and whistles? ok, here's my wish list:

SAP ERP 6
32 CPU Unix server
Oracle 11g

Let the holy wars begin.....

SAP ERP 6? I thought the question was unlimited budget, not unlimited time....

Question is limited to EnterpriseOne 8.12 implementation.
Please no SAP stories in here !... THANK YOU.

Ooooo, this will be fun

The answer of course is ... it depends.
Most of all it depends on number of users, transaction volumes, and required uptime.

Given my druthers I'd want ...
1. Primary Database / Server
64 bit Oracle running on 64bit Linux or Windows on a 64bit Intel/AMD box. Four - 8 processors, 64 GB memory, Multiple Raid 1+0 drives (database spread across several RAID 1+0 subsystems, journaling/redo logs on additional separate drives.
2. Fallback/Reporting Database / Server
Rather than a RAC setup (too complicated - anything too complicated will fail when you need it most) - I'd want a duplicate hardware/database setup where we would use Oracle's Data Guard software to automatically update database transactions from primary DB to the secondary DB and this Read-Only DB is then used for all enterprise reporting.

Reason for Oracle DB - it is still the most robust and less likely to experience locking problems than SQL Server.
Reason for 64 bit - Memory allocation for large databases under 32bit windows involves hokey compromises.
Reason for Intel/AMD hardware: easier to duplicate and/or find a replacement for in a hurry. Inadequate service from one hardware supplier? Just switch.

3. Enterprise Server:
I love a UNIX solution (AIX or HPUX) here.
Any Unix box should be rock solid and relatively low maintenance.
Windows based boxes have issues with nasty Windows Update, etc.
But really either Unix or Windows would be OK. Sizing of box (memory, CPUs) depends mainly on number of Web-Clients. Would not be very large really for us (4 processor, 16-32 GB ram, RAID 5 disk).

4. Deployment Server:
Anything with lots of RAID 5 disk. Doesn't have to be super powered or lots of memory - just lots of disk.

5. Web Server:
OAS on WinTel. Not experienced enough to offer much advice but my inclination would be to treat the OAS and Web Servers as commodity items - have several in place that are easily replaceable with better/faster units. Go with Red Stack because supposedly slightly easier to install/maintain than Blue (I'm not impressed with OAS so far).

6. Backup Server:
Have a backup server(s) with LOTS of disk. Backup strategy s/b to support fast, hot backups disk to disk, disk to tape then takes place off the Backup boxes.

7. LAN:
Separate 1 GB+ (10 GBs?) subnet between servers. This is in addition to primary network backbone.

Final Advice:
If you want performance and reliability - the place to invest is in your storage subsystems (raid 1+0, multiple channels, etc). Forget the box color . . . Your storage subsystems is where the payback is.

O yeah! Forgot about question on Fusion upgrade . . .
Don't worry about it. There won't be an "upgrade" per se.
EnterpriseOne customers will be offered a "migration" path to the latest version of E-Business Suite - its all marketing smoke and mirrors

Larry,

Great suggestions!

I don't have a whole lot of experience with Linux/AIX/Unix and I always claimed to be "allergic" to the "green-screen" before I started working with OneWorld Xe, but since there are no budget constraints, what about the IBM AS400 (iSeries/Systemi...whatever you want to call it)?

I am by no means an IBM zealot, but I've been really impressed over the last 6 years with it's automatic management of the database, disks, memory and batch processing. Not sure I understand fully it's processing of E1 interactive-type processing, but I have a hard time believing that it would be any better on a different platform.

I am also curious about the Deployment server specs: "Doesn't have to be super powered or lots of memory - just lots of disk." I currently have a total of about 130GB (OS: 10GB, Apps: 120GB). I am always trying to reduce the amount of time that package builds take. I thought that in order to accomplish that, I would need to invest in more CPU and RAM. Currently, the client portion of the package build takes between 3.5 and 4 hrs. Our server has 2 dual core 2.8GHz processors with 2 GB RAM.

Also, the JAS server choice, you suggest using OAS, yet you make the comment that you are "not impressed with OAS so far". Any real reason for that or are you just hoping it gets better? I am currently using WAS/Windows and will be researching the benefits of moving to OAS/Linux next year.

I don't mean to pick a part your comment. You have some really great insight, so I thought I would try and pick your brain a little. Getting reliable, accurate information about E1 can be really tough sometimes.

Thanks for your time,

Stewart Schatz

Stewart,

I am more of a developer than anything else so my response was a fun exercise mainly. In other words don't give my opinions too much weight

I didn't include iSeries because:
a) Very little experience with it
b) Its expensive bang-for-the-buck compared to other solutions (in my opinion)

Regarding your Deployment Server package build times. Its about what we see also. Given adequate resources of CPU and memory such as you already have I don't think you'll see much improvement from adding more CPU and memory. I think you'll find that the biggest time component in package build is in the retrieval of specs data from the database. Maybe I'm wrong ... I know a few people's full client package builds are around 2 hours - maybe they can chime in with what they did to optimize package build time.

I'm not overly impressed with OAS from an admin point of view because of:
1. Documentation - I knock its organization. I give it a B- grade.
2. Primitive interface - The admin interface looks like an afterthought (maybe I've been spoiled by Microsoft).
3. Documentation - Accuracy - I give it a 'B' grade.
4. Documentation - Completeness - I give it a C+ (this may be due to the organization though and the fact that companies now-a-days have documentation floating around in "pieces" in different places on their web sites.)
From a functional viewpoint OAS seems to work fine so far - we're not in production with it yet.
I haven't seen WAS admin interface or documentation remember - I might be even harder on it

[ QUOTE ]
I would like to hear from the experienced IT folks, if you had unlimited budget and all the skill sets necessary which hardware, OS , DB and middleware would you implement a new EnterpriseOne 8.12 and why ?


[/ QUOTE ]


Personally I like QuickBooks. The new versions are the best.

I'm going to pretty much agree with Larry.

Picking Oracle Database and installing it on an AMD Opteron platform with 64bit Linux - a few hundred thousand dollars

Picking a really nice state-of-the-art hard disk array, utilizing hundreds of arms and mirroring with another array to provide 100% redundancy throughout - close to a million bucks

Using Intel 64bit Quad Core processors on front end servers with Linux (or even Windows 2008 server maybe) and Oracle Application Server - some change from say a hundred thousand dollars

Putting time and effort into the backbone network and ensuring multiple, redundant connectivity and high throughput with a phenomenal and experienced network team - a few more hundred thousand dollars

Using experienced Independent consultants headed up by your internal project team instead of some fat bloated big-5 company thats ready to be sued - priceless.

There are some things money can't buy. For everything else, theres Oracle....

[ QUOTE ]
Unlimited budget? All the bells and whistles?
SAP ERP 6
32 CPU Unix server
Oracle 11g


[/ QUOTE ]

Yup

So how many 32 way boxes do you think you'd need to run your company under SAP as well as JDE does ? Probably quite a few !

[ QUOTE ]
[ QUOTE ]
Unlimited budget? All the bells and whistles?
SAP ERP 6
32 CPU Unix server
Oracle 11g


[/ QUOTE ]

Yup

So how many 32 way boxes do you think you'd need to run your company under SAP as well as JDE does ? Probably quite a few !

[/ QUOTE ]

Well you know what they say, the guy with the most toys at the end, wins! He DID say unlimited budget.....

Thanks for your response. I appreciate it.

If *I* had an unlimited budget, I'd be on a beach somewhere. If my company handed me an unlimited budget with which to implement EnterpriseOne 8.12, I'd certainly avoid ruffling their feathers by suggesting an SAP deployment.

If I knew the next step would be to upgrade to Fusion Applications, I'd go pSeries, AIX and Red Stack, simply because we already have the experience in house to manage that type of a setup and are doing so today.

But Charlie, it's FUN ruffling feathers!!! Throwing in a curveball answer without flaming anyone helps to lighten things up......

I think its more fun to add value. How's that for a curveball?

[ QUOTE ]
I think its more fun to add value. How's that for a curveball?

[/ QUOTE ]

LOL - unlimited ROI !!!!

Thats the problem with these "hypothetical questions" - someone stating "if you had unlimited budget what would you do" is about as informative as "if you had no budget what would you do".

I can spend a LOT of money on equipment to ensure resilience and performance - but where would the return on investment be ? What would be the value to the company ? How does anyone here know whether suggesting a multi-million dollar system will help your company reduce expenditure and over what term ?

I know companies that spent hundreds of thousands of dollars on OneWorld and saved millions against their bottom line. I also know one specific company that spent $100m on their implementation and it almost bankrupted them - they didn't get a single module running.

I get asked by lawyers all the time about the stability of ERP software as their clients try and point fingers back to the software company - but realistically every failed implementation is almost always due to bad management and a failure to understand the scope of implementing the product.

So, hypothetical questions really do nothing except maybe show a few peoples "favourite" platforms. Ask enough people in these forums, and you're going to end up with every platform with every high-end hardware configuration as the "right" platform in the end.

What is important is to have the right platform that matches the companies skills and strategic direction, and to ensure that your implementation partners keep the companies best interests at the forefront of the project goal. I don't think anyone can argue with that "hypothetical" reality.

With unlimited budget I'd spend a big chunk on getting the right people. It's very easy to make the "best" H/W & S/W combination run like a dog through misconfiguraation, scoping etc. Good people, doing good work on modest hardware may not always be able to beat bad people doing bad work with a huge heap of tin resource but where would you rather work?

That said, for an 8.12 install I'd take Wintel (because we know it), ORACLE db (because I trust it) OAS (break the rules here - we're familiar with WebSphere and gut feel is its a better product BECAUSE its harder to set up but I'm thinking about what's likely to be supported in the future). Lots and lots of tin, regional web servers, clustered DB with hot failover and separate reporting instances and tuned application servers. I want separation between the hardware and the functionality - treat tin as a consumable and buy the best that week for whatever it is I need.

On and some very serious network installation too - both in the server farm (server to server, server to backup) and to the regions.

My tuppence.

[ QUOTE ]
(break the rules here - we're familiar with WebSphere and gut feel is its a better product BECAUSE its harder to set up but I'm thinking about what's likely to be supported in the future).

[/ QUOTE ]

I am not about to get into an argument about OAS versus WebSphere, because they are both fine products. Still, does the difficulty of installation truly lend itself to being intrinsically "better"?

For instance, did you ever setup a Linux distribution (Slackware, maybe) with a kernel in the 1.x range, circa the mid-90's? How much fun did you have with the XFree86 setup? In that case, the "fun" may have been in getting it all to work...but once it was working, was it "better" than an OS/2 installation, or - no flames please, Windows? Subjectively you might say it was, depending on how much "fun" you had downloading and compiling source code into functional applications. ;-)

my bad for being a little unclear there! In spite of working in IT I don't actually look for pain and suffering ;-) What I meant was I have the vague feeling that having more parameters to tweak in Websphere than it seems to allow you in OAS (and I freely admit that I haven't had enough time with OAS to be sure that comparison is true) - anyway, having more parameters ought to, if you could actually understand what they did, allow you to tune better.


BTW my favourite was AIX 1 on IBM's PC platform and the incredible disappearing /tmp structure! (It's a long story but the bottom line is a compiler really ought to be able to tell the difference between a successful compile that produces a compiled output and one that reports no errors because it produces no output at all.)