If the problem of high peering intranet latency is solved

Question

I set up peering in Singapore region to connect two vpc of the same AZ, the ping latency between EC2 of the same vpc is 100ms, but after pinging through peering the ping latency becomes 700ms.i need to reduce the latency of the network between peering

Accepted Answer

Hi, those latency figures are very very high, so I think something else is going on - it's not the VPC and peering infrastructure.  To give you an idea, latency across the whole of Australia, from Perth to the AWS Sydney region, is well under 100ms.

Also note that a VPC is not associated with an AZ.  The subnets inside are individually associated with AZs.

Also if you're peering VPCs across accounts, the AZs don't necessarily line up by name.  What I mean is that AZ "a" in one account won't necessarily be the same physical AZ as "a" in another account unless you've explicitly aligned them.

Answer

I would expect the latency to be higher for cross-AZ traffic as compared to intra-AZ traffic because availability zones are separate from each to avoid common points of failure.

However, 100ms within the same AZ is extremely high; and 700ms (close to one second) between AZs is not normal. You should see single-digit millisecond latency between availability zones; I just tested this in the Singapore region between all three availability zones and that's what I'm seeing.

If the problem of high peering intranet latency is solved

Relevant content