Standards and the Challenges of High Speed Cable Link proprietary encryption handshakes
High Speed Cable Link proprietary encryption handshake issues are often adversely affecting the interoperability of a wide variety of Ethernet, InfiniBand and FibreChannel OEM products being installed in various DataCenters. Many cables using PCB-based plugs are embedded with EEPROMs that have codes for cable assembly performance level, assembly length, manufacturing identification and OEM product interoperability limiter. The SFP+ EEPROM specification, SFF-8472, allows for standard coding and private coding fields.
A painful example of this phenomenon is when an end-user DataCenter manager needs to quickly add more server or storage and switching equipment and buys it from a different and new OEM source, maybe based on fastest delivery. As the DataCenter network engineers are readying to connect their new equipment they find that they need to very quickly buy more cables from their primary Infrastructure wiring supplier who had been qualified and involved in the original cabling installation and other upgrades.
The DataCenter manager buys high speed SFP+ cables from their primary supplier and they arrive within a few days per typical turn around time. These cable products have already passed Industry Plugfest interoperability testing with flying colors and yet, woefully, they will not work between the original equipment and the new sourced equipment due to encryption no-go code. At this point, the DataCenter manager finds out that certain Tier 1 OEM devices connected to another source Tier 1 OEM devices will not handshake and operate. Then he or she contacts the OEM and finds out the same cables will cost up to ten times more compared to their primary supplier and will not be available for several weeks. They sometimes find out that certain OEM supplied cables will not handshake with other sourced OEM equipment. These kinds of incidents do not generate End-User satisfaction.
Leading edge data rate products use very complex technologies and components which are even challenging to have working well within a single source OEM’s product family. We have heard of this problem occurring at recent Industry plugfests. Being empathic that tuning components, circuits, systems and maintaining signal integrity is increasingly byzantine, it is understandable why some major OEMs try to control the whole environment of their products including cable assemblies. It is understandable that OEMs offer lifetime warranties to ensure operability and optimum performance of their equipment only. However if every OEM product did only proprietary handshaking, the market would dramatically slow down causing much bigger issues for every business in the DataCenter and Networking market as a whole.
Many key members of the Ethernet Alliance and IEEE 802.3 community leadership have been expressing their growing concerns of the adverse effects of cable link encryption and even the viability of their organizations and standards. They seem most importantly concerned about open system interoperability and the maximum expansion of Ethernet within new market segments, especially the DataCenter market. Trying to coming out of this great recession quicker as a community, it seems that we need more clever ways of working together and solving issues that could cause a more serious dragging anchor effect and market crystallization if unaddressed.
I suggest that some open dialog about this issue within the key communities occurs soon. It seems that members of these industry organizations should be persistent with requiring open handshaking of multiple sourced cables as a part of plugfest certification that would solve quick, cost sensitive cable link installation problems within heterogeneous DataCenter installations.
The alternative may lead us back to 1970’s when OEMs mostly had their own captive cable assembly divisions for supplying links and very few OEM devices worked with other OEM devices.
Leave a Reply