[dpdk-dev] 17.02 Roadmap

[De Lara Guarch, Pablo]

> -----Original Message-----
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Thomas Monjalon
> Sent: Monday, October 10, 2016 1:43 PM
> To: O'Driscoll, Tim
> Cc: dev at dpdk.org
> Subject: Re: [dpdk-dev] 17.02 Roadmap
> 
> Thanks Tim for the interesting inputs.
> Some of them may require a dedicated thread to continue the discussion
> based on some preliminary specifications or drafts.
> 
> 2016-10-10 16:13, O'Driscoll, Tim:
> > Elastic Flow Distributor: The Elastic Flow Distributor (EFD) is a flow-based
> load balancing library which scales linearly for both lookup and insert with
> the number of threads or cores.  EFD lookup uses a "perfect hashing" scheme
> where only the information needed to compute a key's value (and not the
> key itself) is stored in the lookup table, thus reducing CPU cache storage
> requirements.
> 
> What is the scope of this library? Just apply rte_hash to a flow table?
> Or is it also sending the packets in some queues?
> Does it depend of librte_distributor?

This is a going to be flow-level load balancer/filter based on perfect hashing,
but it doesn't store the key as a regular hash table.
It won't be an extension to the distributor library.

Will send an RFC soon (next week), where we can continue the discussion,
with code, documentation and sample app included.

> 
> > Extended Stats (Latency and Bit Rate Statistics): Enhance the Extended NIC
> Stats (Xstats) implementation to support the collection and reporting of
> latency and bit rate measurements. Latency statistics will include min, max
> and average latency, and jitter. Bit rate statistics will include peak and
> average bit rate aggregated over a user-defined time period. This will be
> implemented for IXGBE and I40E.
> 
> Are they retrieved from hardware or just computed in software?
> Could we have some drivers hook to compute them in a generic layer?
> 
> > Run-Time Configuration of Packet Type (PTYPE) for I40E: At the moment all
> packet types in DPDK are statically defined. This makes impossible to add
> new values without first defining them statically and then recompiling DPDK.
> The ability to configure packet types at run time will be added for I40E.
> 
> The packet types are part of the API.
> Although not yet convinced by the packet types, I do not understand how
> to benefit from some run-time defined packet types.
> 
> > Packet Distributor Enhancements: Enhancements will be made to the
> Packet Distributor library to improve performance:
> > 1. Introduce burst functionality to allow batches of packets to be sent to
> workers.
> > 2. Improve the performance of the flow/core affinity through the use of
> SSE/AVX instructions.
> 
> The packet distributor looks more and more like a DPDK application
> at the same level of BESS, VPP, OVS, etc.
> 
> > Add MACsec for IXGBE: MACsec support will be added for IXGBE. Ethdev API
> primitives will be added to create/delete/enable/disable SC/SA, Next_PN etc.
> similar to those used in Linux for the macsec_ops. Sample apps (l3fwd,
> testpmd, etc.) will be updated to support MACsec for the IXGBE.
> 
> How the ethdev interface and the cryptodev capabilities will be linked for
> MACsec?
> 
> [...]
> > Create Crypto Performance Test App: A new app, similar to testpmd, will be
> created to allow crypto performance to be tested using any crypto PMD and
> any supported crypto algorithm.
> 
> Good idea :)
> When I read "testpmd", I tend to think that it could test any PMD,
> including crypto. Are you saying that we should read dpdk-netdev-test?
> And you would introduce dpdk-cryptodev-test?
> 
> [...]
> > Optimize Vhost-User Performance for Large Packets: A new memory copy
> function optimized for core-to-core memory copy which will be added. This
> will be beneficial for virtualization cases involving large packets, but it can be
> used for other core-to-core cases as well.
> 
> Is it an enhancement of rte_memcpy or something else?
> 
> > Support New Device Types in Vhost-User: Support will be added to vhost-
> user for new device types including vhost-scsi and vhost-blk.
> 
> Great!
> Is it only related to networking or should we expect some storage-related
> code or drivers to come up?
> 
> > Interrupt Mode Support in Virtio PMD: Support for interrupt mode will be
> added to the virtio PMD.
> 
> I guess you mean Rx interrupt in virtio PMD to avoid 100% polling in the VM?
> 
> > Virtio-User as an Alternative Exception Path: Investigate the use of virtio-
> user and vhost-net as an alternative exception path to KNI that does not
> require out of tree drivers. This work is still at an experimental stage, so it
> may not be included in 17.02.
> 
> Interesting. Please share more details of the design you are thinking of.