If you connect with a 2x1Gbps link aggregation/Gig EtherChannel the max. theoretical speed you should get in any single point-to-point file transfer test across it is 1Gbps. Think of it as a two-lane highway, with a 1Gbps speed limit on each lane. Traffic to or from a specific MAC address will be directed down one or the other lane, not both. GEC improves overall throughput when you have multiple traffic flows going across the link, but any single flow will be limited to 1Gbps max.
If you can create two separate L3 links between the 3550-12G and the 5224, and can do OSPF equal-cost path load balancing, then in theory the traffic from one point to the other point should be able to be spread across both links. But I don't have any personal experience with the 5224 -- and the user guide makes it look like it's a Layer2-only switch. So equal-cost path load balancing with OSPF is out.
You could do a really large file copy or transfer across the link, something several hundred MB in size, to stress it. And either time it and do the math, or use PRTG to watch the throughput on one end of the connection in real-time. Here's a link to their site:
http://www.paessler.com/products/prtg
If you use PRTG, set the sampling interval really short to get a better sense of the real-time throughput.
Hope this helps.
