My org has a 1 Gb/s Internet circuit but for some reason when we use the AWS CLI to try to copy data to S3, the most we ever see if 4-5 MB/s. This seems really slow, so I'm trying to figure out what the deal is.
I've connected a PC as close to the WAN as possible to eliminate firewall limitations, etc and I get the same result.
The only thing between the ISP and my test PC is a Cisco 3945E router. The ISP delivers our circuit as single mode fiber, which is connected directly to a gig port on the 3945E, which advertises our /24 via BGP. Our /24 is addressed on another gig port on the router, which is connected to a switch and that switch is where I've connected my test PC. The router is the only layer 3 device between the test PC and the Internet.
I've looked at every port (router WAN, router LAN, switch port facing router, switch port facing test PC) and none are showing any errors or meaningful utilization. I've also looked at router and switch CPU / memory utilization and all are fine. The router is just doing straight routing, it's not doing NAT or any other services.
If I go to speedtest.net on the test PC, I'm seeing ~600 Mb/s up/down which is a little less than I'd expect, but is still way more than what I'm seeing to S3.
A traceroute to the S3 endpoint I'm hitting seems "normal" - there are 6 hops in my ISP's network before traffic is getting to AWS.
By way of comparison, from my home PC which has 1 Gb/s service through AT&T GigaPower, I'm seeing about 25 MB/s, which again is slower than I'd expect but still faster than what I'm seeing at work.
Is there anything else I should be checking to try to determine why our throughput to S3 is so slow?
No comments:
Post a Comment