Forum Replies Created
-
Posts
-
The incoming email address would be FABRICOfficeHours@admin.live.unc.edu
Please check your spam folder. The email definitely goes out, but some spam filters over-react. There should be an email confirmation to the address you specified.
You should have received a calendar invitation at the email address you provided in the office hours sign up and that should have contained information about a Teams meeting to join. Please revisit the invite and check if the Teams information was there. If you can’t find it please take a screenshot of the invite email and post here.
The thread I pointed you at was discussing using DPDK with SmartNICs, I belive. There is nothing we do to enable or disable DPDK – you are free to install it and configure it as any other software package. We currently are not able to provide things like NUMA topology information, but we are working on that.
Specific to FPGAs – none are available right now – we are in the process of sending them out and installing at sites. Programmatically (via API) we expect them to be available in early summer. Again – there is nothing that is stopping you from using DPDK with FPGAs once they are available – it is entirely up to you as an experimenter. The use of DPDK also strongly depends on the specific FPGA toolchain you choose to use. We do not dictate a particular toolchain.
Hello,
All CPUs in FABRIC are AMD, so Intel-specific DPDK optimizations aren’t available. For more information you may want to look at this thread https://learn.fabric-testbed.net/forums/topic/dpdk-tm-capabilities/
Fraida,
The curent portal version has a ‘Batch Upload’ option for adding project members. If you navigate to your project, then use the ‘Project Members’ sidebar link, you’ll see ‘Search’ and next to it ‘Batch Upload’. Let us know how this works for you. Note that adding a large number of project members locks up the membership manipulation (it is a lengthy process). You can continue doing other things in the portal, but you won’t be able to manipulate the membership until the batch process completes.
Very nice! We are discussing internally the change we need to make to allow full 9000 MTU – we plan to implement the change after the workshop next week.
@yoursunny I tested SALT to UTAH and this is what I have as the max MTU:
try: node1 = slice.get_node(name=node1_name) stdout, stderr = node1.execute(f'ping -M do -c 5 -s 8928 {node2_addr}') except Exception as e: print(f"Exception: {e}") PING 192.168.1.2 (192.168.1.2) 8928(8956) bytes of data. 8936 bytes from 192.168.1.2: icmp_seq=1 ttl=64 time=0.268 ms 8936 bytes from 192.168.1.2: icmp_seq=2 ttl=64 time=0.214 ms 8936 bytes from 192.168.1.2: icmp_seq=3 ttl=64 time=0.225 ms 8936 bytes from 192.168.1.2: icmp_seq=4 ttl=64 time=0.239 ms 8936 bytes from 192.168.1.2: icmp_seq=5 ttl=64 time=0.246 ms --- 192.168.1.2 ping statistics --- 5 packets transmitted, 5 received, 0% packet loss, time 4118ms rtt min/avg/max/mdev = 0.214/0.238/0.268/0.023 msAdvitha,
The answer depends on what you mean by ‘sharing’. The upcoming version of the portal will allow you to load the saved topology (there is already a ‘Download JSON’ option there now) so you can recreate someone else’s slice.
If you want to provide access to your slice to another user, you do need to put their public key into the VM slivers, which fablib can help you do. We are working on a feature that lets users use the public key of another user, if it is registered with the portal, to install that key into your slice, thus granting them access. This is a work in progress.
@yoursunny – we looked into SALT-UTAH – you are correct, the MTU was incorrectly set on the switch on the 25Gbps ports (for 100G it seems to be correct). We will remedy this next week and will check settings on other sites so the experimenters can get as consistent an experience as possible. Thank you for reporting this.
UTAH-SALT we will look into – that is a L1 connection and this is surprising, thank you for letting us know @yoursunny
Connection to MASS unfortunately goes through providers that do not support jumbo frames – it is concatenated from two L2 services from regional providers and no other options exist. We will start updating the topology advertisements you see to indicate the MTU a given link can support, currently feel free to ask us.
Due to Intel’s announcement about stopping further development of the Tofino chipsets we’ve had to alter the strategy for deploying them. Previously we planned to have them at several core locations, but now we feel this is no longer warranted since these products have a limited useful lifespan, more likely now used for education, so instead we identified several campus sites and we are discussing the deployment with them.
