network slowness troubleshoot

Question

I want some suggestion regarding our network issue:

From our office location to our DC ==> it's ~270msc(office and DC lies in different countries). As, you can already see here latency is the major concern for us. At working hours i've complains network is too slow while download/upload of files(even small size files as well).

Network wise: cisco routers in office, we've 100Mbps internet(ipsec-BGP) ===> connection towards our DC.

Earlier, i've did some theoretical calculation of throughput = ~2.5Mbps what i got out of 100Mpbs pipe.

Proposals to solve issue:

1) Increase the bandwidth upto 1Gpbs, will that help for such long distance latency?

2) Replacing cisco routers in location with juniper srx(in remote office), not sure how it's gonna help but it's point.

3) Using WAN-Optimizer(im totally new to this) anyway will this help?

I would like to simulate high latency(as of ~270msc) in my lab, this way i can be sure if replacing cisco with juniper may be any help. In, order to do so is there any such commands to simulate latency between devices in same lab(although their latency is ~10msc, i want to those devices to see next hop as ~270msc)

All suggestions are welcome :)

Thanks & Regards,

Madhu

Answer 1

Option 3 is your only real choice.

You are limited by physics. As you probably know, you are running into delay-bandwidth product limitations, so Option 1 --increasing your connection speed won't help.

Option 2 is great news for Juniper, but it will make absolutely no difference in performance.

You might consider adjusting the maximum TCP window size on the computers in your office, but option 3 may be easier.

Answer 2

Have you already accounted for MTU issues?

Often, an IPSEC or other tunnel over the public Internet will result in a path MTU of e.g. ~1420 bytes to destinations across the tunnel. If your hosts are all configured for 1500 bytes MTU they can usually work out the lower path MTU in the absence of ICMP filtering, but it's also frequent that folks block the necessary ICMP fragmentation-needed messages required for Path MTU Detection to function correctly. The upshot is worse performance all around.

The work-around typically involves configuring a feature such as TCP MSS adjusting or clamping (terminology differs) which changes the advertised MSS during the TCP handshake process so it won't exceed the tunnel MTU.

If you don't know that you've already done this, chances are you haven't! It may be an easy fix to significantly boost TCP performance.

Answer 3

First and foremost is to verify bandwidth utilisation on ISP. Internet link . And futher look for any packet drop in your entire end to end infrastructure or any congestion in network and even verify CPU ,RAM and disk resource utilisation in network devices like firewall, routers and switches connected in your topology .