How to calculate required switch speed based on network usage?

I have a 48 100Mbit/s port HP Procurve Switch 2610 (J9088A) that can handle 13.0 million PPS (packets per second) and features wire speed switching capacity at 17.6Gbps.

First off, what does that REALLY mean?

Where do I start when trying to figure out if my office (with 70 employees) will be well setup with this switch? How to calculate through-put based on a user average load of X MB per day?

90% of the folks will only be sending email, access random websites, etc... the other 10% will be conducting heavier tasks like moving image files (10 MB) across network shares, constant external FTP streams through the switch to a server etc...

Is this switch good enough?

Let's assume absolute worst-case scenario for estimation here:

You said 10%, or 7 employees, would be moving image files across network shares. That is an absolute max of 1GBps per connection, but will most likely max out around 300Mb/s depending on storage hardware. If they're doing it across the Internet, it's going to cap out at the lower of either your uplink bandwidth or the remote server's bandwidth.

Even estimating that they use all 1Gbps for those 7 employees, that leaves 63 employees to share 10Gbps. In order to use that up with email and websites, your Internet connection would have to be 10Gbps, and each employee would have to be using 158 Mbps. Continuously. That's not even remotely feasible through web browsing and email, by about 3 orders of magnitude.

The above is all calculated with the assumption that every employee is doing all they can to use up all your switch's capacity at the same time for an extended period.

Yes, the switch would be more than good enough.

EDIT: For reference, We have over 100 employees and over 100 servers. Our core switch has never registered more than 450,000 packets per second. Even if every single packet was 1,500 bytes, that's still only around 5.3 Gbps.

How to calculate through-put based on a user average load of X MB per day?

It's pretty much irrelevant to base effective bandwidth of real world usage on the averaged bandwidth of users across a whole day with this number of users.

What's not clear is how you intend connecting 70 users to a 48 port switch.

90% of the folks will only be sending email, access random websites, etc

Even an un-switched lan is going to have latency times several hundred times smaller than the latency of the internet connection - and the bandwidth discrepancy will be almost as large. So using using ports from an expensive switch has very little value.

Note that unless you are planning to add multiple switches - and need to think about spanning trees for multiple paths, high end switches have little performance benefits over cheapo basic ones (yes there maybe differences in build-quality/longevity - but IME, they don't justify the price differential). The other place where expensive switches make good sense is where you need really low latency switching. You're not going to see the beneifts for remote web/email access and large local transfers.