The reality is that knowing your bandwidth needs depends on many things including your database size and schema. FM does many smart things for you like if you have 500 huge images, but the layout only shows thumbnail sizes, then FM only downloads a small size version of the picture and not the whole picture. So when you talk about a picture, you have to know the size of the picture in the layout to roughly determine the amount of bandwidth to download that picture. It won't be the full picture size. I've seen some relatively small databases with incredibly complex schema and lots of unstored calcluations that can bring a fast server on a local area network to a crawl with only 5 users. There are other well optimized large databases being used by 1000+ people that work well. And everything in between. Other questions are whether you are trying a solution all in a local area network that is very fast or if this is a wide are network solution (e.g., over the internet). If you have an internet solution, maybe your server has an incredibly fast connection at a server farm, but your end users are on a rural slow DSL with very slow upload speeds. It can render the solution unacceptable even though the server's bandwidth has nothing to do with the limitation. Other considerations are how fast the computers are that the users are on. If they are on a slow iPad version one, things can be very slow. A very fast gamer computer with the latest xeon processors will be much faster. If the only thing on a layout is a small version of the B&W pic and no other fields or schema involved, and a local area network, then no problem at all. The reality is that you'll have to create a development solution and test it as you go. Regarding your network, if it is one that you control (e.g., local area network), then there is no reason not to be using cheap gigabit ethernet switching technology to get the most out of FileMaker.
There was a guy at Devcon this summer that did some networking bandwith calculations. Someone like this guy can do the calculations assuming you really know what information FM has to pass to support a given layout down to details of whether security is turned on and indexing status of fields, etc. I think it will be much more than you will be able to provide about your final solution. So the reality is that many of us test our solutions as we scale them up and do a lot of things to opmize fields to avoid unstored calculations and complex schema. And as necessary, we replace network bottle necks, add memory to the server, use dual ethernet connections (one for WAN and one for LAN), etc.
I wish there was simple formula, but what you ask is actually a very complex question. And a detailed analysis by a networking specialists is going to be more than you'll be able to get an answer here, but there may be companies on here that could do such an analysis (obviously for the appropriate consluting fee). So my question is how important is this to you, are you willing to purchase test equipment, do you have a final solution for testing, and what kind of budget do you have? Or are you just asking a general question of technet developers on how well their solutions with container fields work with 500 or so users?
In my circumstance, I've worked with a 200+ user solution and fairly complex schema including container fields (inspection pictures). It worked reasonbly well over a 100 megabit ethernet switching system. While there were 200+ users, not all were doing queries at the same time and most were idle (but connected) at any given instance. The system had not been well optimized for large users support, but was in need of it, as well as upgrading to gigabit. However, it worked satisfactorily in that situation.
Thanks for taking time to give us your feedback. It is truly not a straight-forward situation but at this moment, I must remember the following :
1) image size, thumnail
2) unstored calculation
4) users' environment (if they are using 3G, iPad 1 etc)
These are some extra information about my current setup which I should have included it with my question:
The current solution, which available only on iPhone, is strictly accessed via the Internet and I will not know who the target audience is. I need to prepare to answer the hosting company the Internet speed that I require. Is it 1M, 2M, etc. The current .fmp12 size is 5.6MB and is expected to grow to at least 500MB. It has 6 tables with standard 1-M relationship.
I need to prepare to answer the hosting company the Internet speed that I require. Is it 1M, 2M, etc.
To add to the complexity, remember that internet speeds have to componenets: upload speed and download speeds. Go for a setup with synchronous speeds. Users will both be downloading and uploading.
Note however that the user's experience will be the result of their own internet speed and any latency between them and your server. You have no control over these but they can make or break the experience.
1 or 2 Mbps is not going to cut it, that is for sure. Not for 500 concurrent clients. Gut feeling: I would probably start at 50 to 100 Mbps. The only way to know whether bandwidth is a bottleneck is by monitoring, which is something that for your setup is going to be a must anyway.
I agree that 1 to 2 megabits per second is too slow. I have a 2 Mbps upload for my development server with a very limited number of clients hitting it and I would not want any slower. Generally you can get a server on a rack in some server farm and the basic minimum ethernet is 100 Mbps and I bet that would do. You could always test 10 Mbps if you're on a budget, but 100 is pretty much the standard connection now with gigabit being the high speed option.
Personally I have used www.MacMiniColo.net in Las Vegas using Mac Mini's with Mac OS X Mountain Lion (10.8.5) and FileMaker Server Advanced and the users have never been able to saturate the bandwidth. They are pretty inexpensive and start at $35 a month. And since you have the Mac MIni as a Mac OS X Server, you can also do mail, web hosting, DNS, FTP, and many other server services if you want. Of course the more services you have going, the slower the FileMaker Server performance.
Another company I hear very good stories on hosting FileMaker Servers is Foxtail Technology (www.foxtailtech.com).
There are quite a few other companies too. I know that Productive Computing, the well known FileMaker plugin developer, recently jumped in the hosting market. If they are as good at hosting as they are at plugins, then I bet they are a high quality good performance place for hosting.
There are also hosting companies out there that have poor customer support and/or limited bandwidth during peak loads. I won't mention names, but I've been burned by them in the past and just be careful. Not all hosting companies are the same and the cheapest often is not the best.
Oh, one last thing, you can always do a virtual server by a company that does not specialize in FileMaker hosting. You would be responsible for installing and supporting FileMaker Server. But there are pretty good deals at Rackspace and Amazon on virtual servers, as well as other places. I know IBM recently purchased Softlayer, which was previously the Planet. So IBM may have some good virtual server options too at www.softlayer.com.
Thank you thank you for all the valuable and professional inputs. This is indeed an intricate issue that must handle delicately as it can either make it or break it
From your response, it seems that 10 Mbps (which is also the starting bandwidth from a local hosting company) is what I should get started and upgrade accordingly when the need arises. It is my first instance, hence, many "unknowns" ahead of me.
The big companies have the same issue with roll outs and bandwidth because there are almost always more variables than can be answered for an accurate response. Also, the main reality you are looking at is not what the normal response times are, but what they are during peak usage. Sometimes hosting companies will offer a burst rate for those few peak times.
Regarding your 10 Mbps, I assume that is your upload speed. Remember for a server, upload is more important than download speeds. If it is at a server farm, then you may have the same (synchronous) speeds, but for most local installations, the download is much faster than upload (async). In the case of a server, the much faster download speed is irrelevant. Focus on upload speeds for your server.