February 18, 2014
Dear OPRA Market Data Recipients:
The OPRA Participants have updated their traffic projections for July, 2014. These projections are
based on 1-second and l00-millisecond intervals. The use of the 100-millisecond interval will better
reflect system utilization during bursts of traffic, but results in a larger projection. These figures are
from the 12 active Exchanges.
OPRA will increase the potential traffic rate from SIAC at the opening on the indicated date.
The bandwidth required to receive data via your SFTI connection is reflected in Gigabits.
Two redundant streams of data are available from SIAC. These projections are for one stream.
1-Second:
7/8/2014
1/6/2015
7/7/2015
1/5/2016
Required
Capacity
Messages Per
1-Second
22,938,500
25,343,500
26,994,500
29,360,500
Bandwidth
Gigabits Per
Second
Bandwidth Plus Total
10% for
Messages Per
Retransmissions Day
5.51
6.08
6.48
7.05
6.06
6.69
7.13
7.75
26,448,050,000
27,557,855,000
29,903,390,500
30,788,430,000
100-Milliseconds:
7/8/2014
1/6/2015
7/7/2015
1/5/2016
Required
Bandwidth
Capacity
Gigabits Per 100Messages Per
Millisecond
100-Milliseconds
3,705,000
0.89
4,077,450
0.98
4,581,850
1.10
5,024,500
1.21
July 2014 Output Rate
The maximum output rate on an individual Multicast Line for OPRA will be 1.5 Million MPS.
July 2014 Peak Packets
The peak packets for OPRA will be 2 Million packets per second.
Latency
The average latency for OPRA is 0.8 milliseconds. Message latency is measured beginning with the
time-stamp taken as an inbound Participant message arrives at the network entrance to the OPRA
environment, through processing by the system into a consolidated message for Data Recipients, to
the time-stamp taken as the outbound message arrives at the network exit from the environment.
These time-stamps are taken and correlated by a process external to the data processing applications.
If the external process cannot correlate an inbound message to its corresponding outbound message
or measures negative latency for a message, the message is excluded from broader latency
calculations such as average message latency.