Mac Mini for FMS 15

[TJ53]

Is still the newest Mac Mini a good choice as a server for FMS 15?

This is for for about just 5 simultaneous users to upgrade from FMS 11 that is running fine on an older Mac Mini.

Thanks in advance!

[Josh Ormond]

What are the specs on the mini you are comparing the new ones to? The newer minis are kind of underpowered as a database server. 5 users may be ok. But what are the chances you will grow to 10 users before you buy another server?

The new minis are only dual-core and limited on RAM.

 

[Steven H. Blackwell]

I would not recommend use of the mini Mac as a FileMaker Server.

 

Steven

[Rick Whitelaw]

I use a 2011 Mac mini but have installed extra ram and a faster hard drive. And this mini can expand further. The problem with the new Mac minis is that they're not expandable. Another thing to point out is that I'm the only user. For my purposes it's reliable and ample. However if I were to decide to expand to multiple users and web direct I would consider another machine. The Mac mini is light years away from being anything near "enterprise". Word has it that Apple doesn't currently offer server-grade hardware although the MacPro is a very powerful machine. 

[Wim Decorte]

The Mac Pro *is* powerful but also very limited in many ways, mainly in disk space.  And you'd waste a lot of money on video card hardware that  you don't need.

The Mac Mini does not make for a good server if you are looking expandability or performance.

[Lee Smith]

2 hours ago, Steven H. Blackwell said:

I would not recommend use of the mini Mac as a FileMaker Server.

 

I was going to ask this earlier, but hadn’t gotten to it.

6 minutes ago, Wim Decorte said:

The Mac Mini does not make for a good server if you are looking expandability or performance.

Is this the same concern as OAM?

[Josh Ormond]

3 hours ago, Wim Decorte said:

The Mac Pro *is* powerful but also very limited in many ways, mainly in disk space.  And you'd waste a lot of money on video card hardware that  you don't need.

The Mac Mini does not make for a good server if you are looking expandability or performance.

Just as an informational point. Our office moved from an older Mac Pro to the new model, and reports were that there was a noticeable drop in overall performance. It simply did not handle the constant bashing of 60 users very well. 

We are in the process of getting things ready to be deployed on a server grade server running Windows. 

[Rick Whitelaw]

Had I 60 users I would definitely move to a proper server grade machine. Probably Windows although I have little experience with the platform.

[Mike Duncan]

13 hours ago, Wim Decorte said:

The Mac Pro *is* powerful but also very limited in many ways, mainly in disk space.  And you'd waste a lot of money on video card hardware that  you don't need.

The Mac Mini does not make for a good server if you are looking expandability or performance.

I really wish it weren't so, but Apple just does not offer hardware for servers. Even if you rack mount a mini or pro, it is just not quite the same. If the mini had a better processor available, I think it would be a closer fit.

[BrentHedden]

10 hours ago, Josh Ormond said:

Just as an informational point. Our office moved from an older Mac Pro to the new model, and reports were that there was a noticeable drop in overall performance. It simply did not handle the constant bashing of 60 users very well. 

We are in the process of getting things ready to be deployed on a server grade server running Windows. 

There have been a lot of people that have come this same way.  I recently moved our FMServers from a couple of MAC minis to a virtual Windows environment.  The resources are more plentiful and definitely cheaper for a Windows option and any Apple hardware can offer, especially when you need to move to something beyond a small workgroup.

I certainly love using my MAC for development more than a PC, and would be hard presses to change that.  But for servers, Apple just falls short.  

[Steven H. Blackwell]

Perhaps one day we will have additional options for server OS and platform.  But that day, if it is to come, remains in the future.

 

Steven

[TJ53]

16 hours ago, Josh Ormond said:

What are the specs on the mini you are comparing the new ones to? The newer minis are kind of underpowered as a database server. 5 users may be ok. But what are the chances you will grow to 10 users before you buy another server?

The new minis are only dual-core and limited on RAM.

 

Thank you all for your feedback. Specs for current Mac Mini running FMS 11:

(I think it was bought around 2009)
OS X 10.6.8
2.53 GHz Intel Core 2 Duo
4 GB RAM

The server is hosting three databases that are pretty simple, plus one database that only stores files and images of the other three databases, accessed by five simultaneous users using FMP 11 (no WebDirect users needed in the future). It is unlikely that these databases will grow in size or users.

This is for a small workgroup so it is very unlikely that they would want to spend $3,000 on a Mac Pro plus the FM15 licenses plus my rates for doing the upgrade. So if a Mac Mini is not recommendable I would have to do some research about a cheaper option using the Windows platform ... which I'm not very familiar with!

 

[Steven H. Blackwell]

OS 10.6.8 will not support installation of FIleMaker® Server 15.  That requires OS 10.10 or 10.11.  Your hardware may not support the installation of that OS.

Steven

[TJ53]

Thank you Steven, these are the specs of the current Mac Mini that has FMS11 installed (Josh Ormon asked for the specs so I posted them here). I don't pretend to install FMS15 here since this is obsolete, but to buy a new server for the upgrade.

My question is, given those specs and the databases being performing reasonably well currently, would buying the newest Mac Mini for upgrading to FMS15 perform at least as well as currently? or should I just forget about buying Mac Mini and should spend more money and buy a Windows server? I know that Mac Minis are no longer recommended, but this is for just 5 users and 4 small databases that are not expected to grow.

[Josh Ormond]

It's really hard to say if it will be ok or not. I would suggest, if you go that route...test it heavily before you deploy. Test both the conversion of the file to .fmp12 and the performance on it with FMS 15.

Are they using IWP ( or have a need for WebDirect ) at all? If the answer is yes, I would NOT use the Mac Mini.

[TJ53]

Thank you Josh,  so it looks like it would be too risky going for the Mac Mini. 

So ... any general advice about what would be the least expensive hardware and OS server software option recommendable? this is for a small workgroup and they are very concerned about the cost of upgrading from 11 to 15 ... actually they may not even upgrade and stay on 11 if they can't afford the new machine plus the licenses for FM15.

[Buckie]

Nothing risky in there at all. Just throw in an SSD if you need more concurrent users or simply leave it as is. I run three FileMaker Server instances: version 10 on a G4 mac mini with 764-something megabytes of RAM handling about 10 concurrent users, another 2008 mac mini server running version 14 (serving just a couple of users but also being a full-featured server with several websites, DNS, calendar and email) and a late 2012 mac mini server running version 15 (6 users+webdirect) and the same stuff as the 2008 one. All run perfectly fine without any hiccups (the G4 one is not lightning-fast but it's also not slow for everyday FMS operations) and I think that Mac Mini is a really powerful machine, especially the later models. So I say just go for it. It'll be fine unless you need to serve hundreds of users.

Edited June 9, 2016 by Buckie

[Rick Whitelaw]

23 hours ago, Josh Ormond said:

Just as an informational point. Our office moved from an older Mac Pro to the new model, and reports were that there was a noticeable drop in overall performance. It simply did not handle the constant bashing of 60 users very well. 

We are in the process of getting things ready to be deployed on a server grade server running Windows. 

I have my 2007 MacPro running El Capitan and it's now 8 cores @3ghz. It's a monster and I have tons of storage on it. 3.5tb and room to go bigger. No reason to buy a new one. 

2 hours ago, Buckie said:

Nothing risky in there at all. Just throw in an SSD if you need more concurrent users or simply leave it as is. I run three FileMaker Server instances: version 10 on a G4 mac mini with 764-something megabytes of RAM handling about 10 concurrent users, another 2008 mac mini server running version 14 (serving just a couple of users but also being a full-featured server with several websites, DNS, calendar and email) and a late 2012 mac mini server running version 15 (6 users+webdirect) and the same stuff as the 2008 one. All run perfectly fine without any hiccups (the G4 one is not lightning-fast but it's also not slow for everyday FMS operations) and I think that Mac Mini is a really powerful machine, especially the later models. So I say just go for it. It'll be fine unless you need to serve hundreds of users.

A couple of problems with this. A G4 Mac mini will not run an OS that will accept a recent version of FMS. An SSD will make reading faster but not writing so I attach less importance to its relevance to number of users. I attach much more importance to RAM which you don't mention except for the older machine. I can't imagine a machine with "764 something mb" being capable of handling 10 simultaneous users.

my 2 cents . . .

[Buckie]

Not sure why you think SSDs won't make writing faster. They do make everything faster, a lot depends on what SSD model you choose. Some of them do indeed have a reading/writing speed difference, sometimes reading is up to two times faster, but you can find models that excel at writing. There also may be an issue with older SATA hardware in the computer that limits the maximum speed/bandwidth.

As for the RAM, it's of course very important and the G4 could use some more but it's physically impossible. Some of my users will never upgrade to version 12+ and I don't see any good reason to buy a new piece of hardware for an aging software. It's generally only noticeable when everyone starts performing scripts all at once but it doesn't happen often and for everyday business work it's just fine. It used to give me troubles when an FMS would stop responding and that used to happen at random, sometimes months would go by, sometimes only days: I finally fixed it by defragmenting the disk from which the database was served. Not only did it solve the freezing problem, it also improved performance for the clients a lot. And that's another point why SSD would be better - they don't suffer from fragmentation (not that you could stick one into a G4 anyway). The G4 runs OSX 10.5.8, the last one officially supported by FMS v10, it has all unnecessary services and daemons disabled including the java applet bundled with the FMS to save memory. And it actually has 1 gigabyte of RAM, I confused it with a PowerMac G4 that used to run the server before that Mini.

As for the other two machines, the 2008 one has 8 and the 2012 one has 16 gigabytes of RAM. So there's a lot of headroom. Both run OSX 10.11.

So to recap, the Mini is a very powerful machine in general. Unless your users are going to perform very intensive scripted operations very often you'll be fine with one. It will probably mostly idle.

[Wim Decorte]

12 hours ago, Buckie said:

 that Mac Mini is a really powerful machine, especially the later models. So I say just go for it. It'll be fine unless you need to serve hundreds of users.

That is just not true.  The later Mac Minis are NOT powerful machines; they seriously lack in processing power.  And it won't need hundreds of users to bring it down.  Depending on the design of the solution

And even with a very efficient design I doubt that will get  more than about your 10 concurrent users out of it.  Disk i/0 typically is only the #2 constraint; if an FMS deployment is going to run in performance problems it is typically for lack of processing power.  The Mac Mini simply does not have many cores (less than its predecessors).

The only metric that counts is to use the FMS stats log and get a baseline for the performance.  Especially if you already have a server; that should make things a lot easier.

2 hours ago, Buckie said:

It's generally only noticeable when everyone starts performing scripts all at once but it doesn't happen often and for everyday business work it's just fine.

I think your solution is not very typical in that respect then.  At the core of most solutions are lots of scripts that are executed by the users on their machines, on the server through PSoS and server-side schedules...

[Buckie]

I can host a sample database file made by anyone on one of the Minis if anyone's wanting or willing to test the performance.

I have the stats.log from all of the machines and I can post them here. Not sure if that is going to be helpful though.

[Joost Miltenburg]

Just my two cents :

I have multiple clients running FMServer (14) on a MacMini (late 2012) server configuration. SSD's, memory maxed out and Core-i7's in them. Usually between 20-30 users on them with several databases several GB's in size ( one even 30GB+ )

They hum along nicely with no user ever complaining about slowness. 

[TJ53]

Thank you all, so my conclusion is that for FMS15 scalable deployments Mac Mini is no longer an option. The only Mac environment option would be the Mac Pro, that seems to be a waste of money and resources and still has some limitations (as Wim Decorte mentioned above).

Therefore, it no longer makes sense using a Mac as FileMaker Server. From now on the recommended and cost-effective option is Windows Server.

[Buckie]

Just want to point out that "Windows Server" is by no means "recommended". We have our opinions, but here's the real official recommendation:

https://www.filemaker.com/products/filemaker-server/server-15-specifications.html

Make of it what you wish.

Edited June 24, 2016 by Buckie

[comment]

39 minutes ago, Buckie said:

here's the real official recommendation:

Actually, those are requirements, not a recommendation of any kind.

[Mike Duncan]

38 minutes ago, comment said:

Actually, those are requirements, not a recommendation of any kind.

To be fair, there are minimum requirements listed on that page as well as recommended hardware. While the minimum requirements are met by a mini, the recommended CPU is at least 4 core, which is not available on a mac mini.

Why, Apple, do you tempt me with your OS and server.app and then not give me the hardware to run it on

I'll keep wishing...

[Rick Whitelaw]

On June 10, 2016 at 3:52 AM, Buckie said:

Not sure why you think SSDs won't make writing faster. They do make everything faster, a lot depends on what SSD model you choose. Some of them do indeed have a reading/writing speed difference, sometimes reading is up to two times faster, but you can find models that excel at writing. There also may be an issue with older SATA hardware in the computer that limits the maximum speed/bandwidth.

As for the RAM, it's of course very important and the G4 could use some more but it's physically impossible. Some of my users will never upgrade to version 12+ and I don't see any good reason to buy a new piece of hardware for an aging software. It's generally only noticeable when everyone starts performing scripts all at once but it doesn't happen often and for everyday business work it's just fine. It used to give me troubles when an FMS would stop responding and that used to happen at random, sometimes months would go by, sometimes only days: I finally fixed it by defragmenting the disk from which the database was served. Not only did it solve the freezing problem, it also improved performance for the clients a lot. And that's another point why SSD would be better - they don't suffer from fragmentation (not that you could stick one into a G4 anyway). The G4 runs OSX 10.5.8, the last one officially supported by FMS v10, it has all unnecessary services and daemons disabled including the java applet bundled with the FMS to save memory. And it actually has 1 gigabyte of RAM, I confused it with a PowerMac G4 that used to run the server before that Mini.

As for the other two machines, the 2008 one has 8 and the 2012 one has 16 gigabytes of RAM. So there's a lot of headroom. Both run OSX 10.11.

So to recap, the Mini is a very powerful machine in general. Unless your users are going to perform very intensive scripted operations very often you'll be fine with one. It will probably mostly idle.

 

On June 10, 2016 at 3:52 AM, Buckie said:

I finally fixed it by defragmenting the disk from which the database was served.

It's not necessary to defragmenter a disk in OS X. The is takes care of it daily.

[Buckie]

All right. This, this, this and also this one and another one all have 4 cores. Take note that 4 cores is a recommended configuration that's supposedly going to be used for heavy load, lots of users and lots of hosted databases. If your goal is to serve 5-something users with a single solution then probably what they call "minimal" requirements would suffice. Minimal doesn't mean everything is going to be sluggish and crash every other 10 minutes, it's going to be perfectly fine with loads like those. Usually those recommendations work like that: nobody's going to officially recommend a minimal configuration that is unstable or unusable.

Secondly, the "not necessary to defragment" is a myth. It is necessary in certain circumstances and can be very useful. In my particular case the "OS" didn't take care of FileMaker files in particular that were left severely fragmented for whatever reason, even though there was still ample free space on the volume (about 40 gigabytes out of 80). As I said earlier, that helped with FileMaker Server v10 becoming unresponsive from time to time until reboot.

My offer still stands if anyone cares about actually testing the performance instead of arguing whether there should be 8-core Xeons installed or 16-core ones.

[Wim Decorte]

3 hours ago, Buckie said:

Take note that 4 cores is a recommended configuration that's supposedly going to be used for heavy load, lots of users and lots of hosted databases.

That is a very faulty conclusion that you are taking there.  FMI's recommendations are pretty meaningless given the wide variety of solutions (and their efficiency) that people can come up with, and the wide variety of deployments.

The biggest server I have deployed has 24 cores plus hyper-threading.  That is after seeing serious stress on the processors when the server was a mere 8 core.  That's for 130 concurrent users and 75 files.  Very heavy loads.

 

3 hours ago, Buckie said:

My offer still stands if anyone cares about actually testing the performance instead of arguing whether there should be 8-core Xeons installed or 16-core ones.

We're not arguing; people here are sharing from experience.  A 4-core or 8-core machine simply may not be enough for the solution at hand.  There's no platform favoritism here, just practicality.  There's also no mystery around choosing the right server: look at the FMS stats log.  Solve the bottleneck that it shows, if any.  Very often the bottleneck will be on the processing side (as expressed in elapsed time per call and wait time per call).  If these are elevated, go to a machine with more cores.

If you make heavy use of PSoS and/or server-side schedules, go for more cores.

This is a good read, and it comes with a performance testing application: http://www.deskspace.com/understanding_filemaker_server_performance_introduction.html

[Buckie]

I agree with that. There's no way you can have 130 users with 75 files running on a Mini. However, TJ originally mentioned "just 5 users" and I presume a single solution for which a decently-specced Mini is capable enough.

And of course the logs are very useful however instead of going too deep in the numbers I like to perform daily tasks on the solutions that I host from time to time to see if the actual day-to-day user experience is sub-optimal. May not be the scientific approach but it's a good meter nonetheless.

dsBenchmark is interesting, however I'm not sure how to interpret the graphs. I ran both quick and std benchmarks with 2 and 4 tokens and 50 "users" and the blue line seems to represent the number of users per token... maybe, and I'm not sure what "period" means. Anyhow, if I don't log in into the machine and start opening browsers/activity monitors and such, the "period" line would stay below 1 most of the time and "dequeue" would stay at 0 all the time. Don't see much in the way of processor load or RAM usage (but the cache is set at 512 MB so I probably shouldn't). FMS stats show normal performance with wait time per call staying at about 250ms most of the time. I have a feeling I'm doing something wrong there.

[BrentHedden]

On June 26, 2016 at 2:49 AM, Buckie said:

dsBenchmark is interesting, however I'm not sure how to interpret the graphs.

This is beyond the original question/conversation, but I can give you my 2cents on using dsBenchmark.  It is a great and invaluable tool for measuring performance at a server level.  It also great as a learning tool for how tokens work.  But it WILL NOT tell you how many users can connect to a server before it breaks.  All of the users it creates are virtual server-sided ones.  Which is great of comparing different hardware or in my case various VM settings to optimize the best performance.  It will steadily increase the number of server-side 'users' until it reaches the maximum (in your case 26, although you stated 50??) or the red Dequeue (wait time per call) is over a certain timeframe (3 seconds?). I would set the tokens to zero unless you are planning on programming a token type of system in your solution.  Otherwise it'll skew the results.

[typewriter]

My 2 cents on the original post:  you say you have a database of 5 users. I guess a well-deployed MacMini should do the job fine. If it runs fine now under FMS11, it will do so with FMS 15 on a new MacMini.

As others have pointed out here, with the newest Minis you don't have any upgrade options other than putting in a bigger SSD of replacing the machine if it runs out of steam. If you get the mini with the fastest processor and maxed out RAM (16 GBytes) you will have a compact, energy efficient and quiet work horse (if you can get an older Core i7 model, you're better off). Sure the Mini is not enterprise-grade. But a database with 5 users isn't either, just keep seeing things in the right perspective.

Speaking of processor load, think of a FileMaker Server as a highway: FMS handles incoming calls in a FIFO way, with multiple processor cores handling more threads simultaneously like having more driving lanes available. If the server becomes overloaded it doesn't stop, but the excess traffic is stacking up just as more vehicles that the highway can handle causes the traffic to jam. Unfortunately, FMS doesn't have many tools to 'throttle down' the incoming calls, so users are not aware of what's happening. That is something FMI should be working on imho, MacMinis or not.

 

 

 

 

[Wim Decorte]

2 hours ago, typewriter said:

Unfortunately, FMS doesn't have many tools to 'throttle down' the incoming calls, so users are not aware of what's happening. That is something FMI should be working on imho, MacMinis or not.

How do you see that happening?  Something like a "please wait" displayed to the client?

I would rather see things move the other way with the ability to load balance across multiple servers.

[typewriter]

9 hours ago, Wim Decorte said:

How do you see that happening?  Something like a "please wait" displayed to the client?

 

Something like that, yes. Perhaps in the form of a couple of Get () functions that pass on system data to the client process. Then it's up to the developer to handle the queue and pass that information back to the user in a polite way.

 

[TJ53]

Thank you all for your insights. 

So my conclusion is: for very particular cases of small deployments (5 users), small and simple solutions (not processor intensive), and not scalable (Mac Mini not possible to upgrade), then a Mac Mini can do the job.

But ... for FMS15, Mac platform is not the best choice for server hardware in terms of cost and scalability, so from now on PC hardware and Windows Server make more sense for FM Server 15 scalable deployments.

I suppose the key words here are scalability (not possible with Mac Minis) and cost (Mac Pros are too expensive and for same investment a more powerful PC can do a better job).