Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

NOTE: I am a self-taught PHP developer who has little to none experience managing web and database servers.

I am about to write a web-based attendance system for a very large userbase. I expect around 1000 to 1500 users logged-in at the same time making at least 1 request every 10 seconds or so for a span of 30 minutes a day, 3 times a week. So it's more or less 100 requests per second, or at the very worst 1000 requests in a second (average of 16 concurrent requests? But it could be higher given the short timeframe that users will make these requests. crosses fingers to avoid 100 concurrent requests).

I expect two types of transactions, a local (not referring to a local network) and a foreign transaction. local transactions basically download userdata in their locality and cache it for 1 - 2 weeks. Attendance equests will probably be two numeric strings only: userid and eventid. foreign transactions are for attendance of those do not belong in the current locality. This will pass in the following data instead: (numeric) locality_id, (string) full_name. Both requests are done in Ajax so no HTML data included, only JSON. Both type of requests expect at the very least a single numeric response from the server. I think there will be a 50-50 split on the frequency of local and foreign transactions, but there's only a few bytes of difference anyways in the sizes of these transactions.

As of this moment the userid may only reach 6 digits and eventid are 4 to 5-digit integers too. I expect my users table to have at least 400k rows, and the event table to have as many as 10k rows, a locality table with at least 1500 rows, and my main attendance table to increase by 400k rows (based on the number of users in the users table) a day for 3 days a week (1.2M rows a week).

For me, this sounds big. But is this really that big? Or can this be handled by a single server (not sure about the server specs yet since I'll probably avail of a VPS from ServInt or others)? I tried to read on multiple server setups Heatbeat, DRBD, master-slave setups. But I wonder if they're really necessary. the users table will add around 500 1k rows a week.

If this can't be handled by a single server, then if I am to choose a MySQL replication topology, what would be the best setup for this case? Sorry, if I sound vague or the question is too wide. I just don't know what to ask or what do you want to know at this point.

share|improve this question
As a programmer, you can test all of these conditions yourself to determine whether your present hardware and configuration will satisfactorily support the anticipated load. If you would prefer to get expert advice that is specific to your application's needs, you should hire an expert systems administrator. – danlefree Jul 27 '12 at 10:52
If you have that many users and are collecting time critical data it sounds like you're going to need multiple servers for fault tolerance anyway. – Matt Aug 16 '12 at 8:35
Also for the performance, the only way you can tell how your application will perform is by load testing it. If you can break your numbers down to bytes written/read per transaction/second you might get a ballpark idea of the minimum but even then your implementation can vastly impact those numbers – Matt Aug 16 '12 at 8:43

i would start with some oracle/mysql whitepapers. they usually get quite deep in details when it comes to benchmarking starting from this search you can get a quick overview about capabilities of either the mysql-server and the hardware related to such performances [the first three results are good starters].

share|improve this answer

We have some canonical questions about Load-Testing and Capacity Planning

How do you do load testing and capacity planning for web sites?

How do you do load testing and capacity planning for databases?

Can you help me with my capacity planning?

Basically, you have to get some monitoring and apply the scientific method.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.