Cisco Open Network Environment
This is something new Cisco is investing in. When you talk about what open networking is you will get different views of what it should be able to do. In fact Cisco is developing a framework of API’s or virtual overlays over the physical and virtual networking infrastructure so the infrastructure can be steered by software. It’s software defined networking or even application defined flows if you will.
Imagine a setup where we introduce a new metric on networking links like the effective COST of a link in Euros. Every morning a network admin receives prices and selects the cheapest link to use for that day. If you would want to automate that, you would have to contact Cisco and add a new request to IETF concerning a new networking metric. Afterwards routing protocols would have to change to take into account this new metric. However with ONE, it’s possible to create an algorithm yourself and instruct your devices via API’s to change it’s routing table… the sky is the limit. You could even construct your own encryption algorithm if you wanted to.
There are several flavors regarding Cisco ONE :
- Openflow which is agentbased
- onePK which is built on open API’s
- Overlay technologies such as the Nexus 1000v
Starting with the overlay space, this is an interesting concept. The introduction of VXLAN created virtual VLAN’s in fact which can be layered on top of your existing Layer2 or Layer3 network. This way you can create up to 16 Million VLAN’s! (16 000 000!). Your complete datacenter is the mobility domain. ESX hosts can be interconnected via L3 and your VXLAN’s will bridge them together (over L3 or L2).
In the future Cisco will develop a VXLAN gateway appliance able to bridge your non-virtual vlan’s over L3/L2 networks
Currently the onePK API is being developed with northbound API for app developers and southbound API towards the devices. The goal is to have a unified API for IOS/XE, NXOS, IOS XR,
… It can exist in the box or as an appliance. This will be available first quarter 2013. More available on www.cisco.com/go/one
Nexus 2000/3000/5000 Roadmap
The Nexus 5010/5020 series will be end of sale at the end of this month. GoldCoast will be the last supported NXOS version for these hardware platforms.
The Nexus 5500 is the current platform to be focussing on with it’s unified ports, L3 daughtercard.
Keep in mind though that for MPLS, OTV or LISP you still need a Nexus 7000 L3 engine. The latest N55-D160-L3 daughtercard has larger table sizes for IPv4 hosts and multicast mac’s : from 8000 to 16000 and 4000 to 8000 entries.
I want to mention the NGA appliance. As the Nexus 5500 series do not support netflow in L3 mode, the NGA appliance was introduces. This appliance generates netflow packets. Input to the device
needs to be generated by a SPAN port on your Nexus 5500 switch.
The latest generation of Nexus 5500 series switches support 24 FEX. Keep in mind when using L3 mode, this amount drops to 8 with the old L3 card and 16 with the newer L3 daughtercard.
Cisco introduced the brand new 5596T Nexus switch, with 32 10Gbase-T interfaces and 16 SFP+ interfaces. These are all FCoE capable, 30m at release (safe harbor) and 100meters in later releases. If you plan to go 10Gbit all the way, this might be your cup of tea.
At the roadshow a new GEM was introduces, the 40Gbit QSFP+ GEM for Nexus 5500. In fact it’s not really a 40Gbit native GEM but allows you to save on optics and power consumption. It’s really 4 wires 10Gbit which are routed into QSFP+ cables individually. You still need to configure a portchannel of 4 interfaces …
Nexus 2000 Products updates
Next to an update to the Nexus 5500 series, the Nexus 2000 series are developed further. The new B22 Blade FEX are now supported in HP,Dell and Fujitsu bladecenters. Keep in mind you need a Nexus 5500 with NX5.2(1) in order to extend these. Nexus 7000 does not support the B22′s. Expect to see high density 10Gbit FEX with QSFP+ uplinks in order to have 40Gbit uplinks.
One of the announcements I was happy about is the pricing strategy around the Nexus 2000 series. Cisco decided to price them all the same at 9000 dollars list price.
The Nexus 2248TP-E-1GE was introduced as a next in line to the 2248TP-1GE. The key differences are optimized buffers and enhanced counters. An excellent improvement for bursty workloads such as video on demand or storage. As all Nexus 2000 series cost the same, just buy this one when you need new Nexus 2000 switches!
The Nexus 2232TE is a 1Gbit/10Gbit switch which is FCoE capable! It’s exactly the same price as the 2248TP but only has 32 10Gbit ports. So you price per port might increase. Otherwise this Nexus will introduce 10Gbit to the server. Another cool announcement if you ask me!
Nexus 6000 was mentioned, which will be a 4RU Nexus with higher port density and a 40gbit base. It should do L2/L3 out of the box. Expect to see this one arrive first halve of 2013. No extra information was provided however.
Last but not least, Cisco mentioned further development around the Nexus 3000. This ultralow latency switch now reached less than 300ns line rate switching and routing… even with ACL’s! That’s impressive … but wait … when going in warp mode (yeah, marketing!) with less than 4K IPv4 host entries you can achieve around 100ns… and when you go into warp mirroring mode you can achieve 50ns latency…. let me repeat that : less than 50ns latency. That’s really amazing.
However warp mirroring mode is about copying data from one port to one or multiple others, there is no real switching involved.
When you are challenged by choosing a stable NXOS, you should still go with 5.2(1)n1(1) as it’s still the long lived release.