Tuesday, February 27, 2018

apache 2.2 - System Requirements of a write-heavy applications serving hundreds of requests per second

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.

No comments:

Post a Comment

linux - How to SSH to ec2 instance in VPC private subnet via NAT server

I have created a VPC in aws with a public subnet and a private subnet. The private subnet does not have direct access to external network. S...