LHCb Online Meeting November 15th, 2000 Event Building Status Niko Neufeld LHCb Online Meeting November 15th, 2000

Done Evaluate AceNIC throughput performance Implement event building in the NIC and run it between 2 PCs Tested 1 on 1 event building over a switch (Foundry FastIron) in the CMS testbed Fixed protocol with Readout Unit Presented results at DAQ 2000, submitted paper to TNS, finished and submitted Mariannas thesis (let’s keep our fingers crossed for her!) Niko NEUFELD CERN, EP

Alteon Tigon 2 Features Development environment Dual R4000-class processor running at 88 MHz Up to 2 MB memory GigE MAC+link-level interface PCI interface Development environment GNU C cross compiler with few special features to support the hardware Source-level remote debugger Niko NEUFELD CERN, EP

Test Setup PC/Linux CPU Mem PCI CERN Network GbE NIC Niko NEUFELD CERN, EP

Performance of Alteon NIC Can fill the wire at any given frame size (from 64 to 9000 bytes) Can send out frames at a frequencies of up to 1.4 MHz For frames bigger than 512 bytes more than 95% of nominal bandwidth available for data (practically 100% for >8000 Jumbo frames) Niko NEUFELD CERN, EP

NIC 2 NIC Performance Niko NEUFELD CERN, EP

Event Building Algorithm Assembles events out of fragments from a known number of sources Handles an adjustable amount of events concurrently (limited only by buffer space) Implements “Implicit + Time-out Completion” Uses “scatter/gather” capabilities of NIC’s DMA engine to concatenate the fragments into the host’s memory Niko NEUFELD CERN, EP

Algorithm ? Polling Start Procedure New fragment New event fragment NO still in the table Fragment out of time Collect the fragment YES Decrement sources Add new event descriptor Check for missing fragments in previous events Niko NEUFELD CERN, EP

Performance NIC 2 NIC Average time per fragment 11.65 us Niko NEUFELD CERN, EP

Event Building Achievements Event building on a smart NIC at a frequency of incoming fragments of almost 100 KHz has been demonstrated Event building at Gigabit speed for fragments bigger than ~1100 bytes Code Optimization ongoing (9 us/frag have already been achieved) Niko NEUFELD CERN, EP

Being done Reorganisation of the software NFS exported volume for easy plug-in in test-bed environments Consistent use of several firmware versions Improve debugging/monitoring tools Add printf like capability Niko NEUFELD CERN, EP

To be done* Learn more about flow control in the Foundry switch Do a full scale test on the CMS test stand (16 on 16) *(hopefully before the LHCb week ) Niko NEUFELD CERN, EP

GigE simulation of the Readout network Very simple Ptolemy simulation exists Uses Atlas GigE Base classes Uses simple crossbar switch Is modeled after the Myrinet simulation To be done: add all functional features of the existing models (throttling etc.) Study Atlas behavioural switch models Niko NEUFELD CERN, EP