I have a pretty fast datafeed (usually 2ms) see screenshot.

But every now and again, CME will get a burst of 1000+ trades in a couple ms. and it blows out my latency to 50+ms.

Is this the CME or the data provider? i have a 10Gbe internet connection and dedicated server.

Just wondering if people have been able to avoid these spikes during the day? And how you did it?

LukePoga 

Is this the CME or the data provider? i have a 10Gbe internet connection and dedicated server.

All three.

CME does exhibit traffic spikes and send packets with heterogenous size, and bursts are naturally more expensive to process.

2 ms is actually pretty slow. CME's data feed uses a zero copy protocol.

Your dedicated server might have a 10 GbE uplink but there's nothing telling what bottlenecks sit between that uplink and CME.

LukePoga 

Just wondering if people have been able to avoid these spikes during the day? And how you did it?

The bulk of the issue is resolved at the application layer.

CME 10Gbe Glink is $144K/year

I often wonder how much faster it will be tick to trade. i can program my own FPGAs and all that. but its no use because of high entry price to CME DMA!

let alone the transaction fees!

i was wondering does CME offer rebates for being a market maker? and what is the entry point for being an individual market maker? i assume its not possible because its not a charity!