Tags
In June of 2013 Cisco released the newest member of the ISR family, the ISR4451-X router (Cisco – blog post). At that time it was the only model in the ISR44xx line, but today it now has some new siblings. Today, October 2, 2014, Cisco has introduced some companious, namely the ISR 4321, 4331, 4351, and 4431.
The below graphic from Cisco that shows the new line and the performance characteristcs of each. You will note, for example, that the ISR4321 says 50–100 Mbps. The base version of the device will support 50 Mbps, but to move to 100 Mbps on that box you just need a software license. This will allow a network to grow with it needs without having to replace hardware, you will just have to add a license file. Back in the ISR G1/G2 days we need to sometime replace hardwre in order to gain additional performance, but that is not the case with the ISR 4000 series. Much of that comes thanks to Cisco using IOS-XE on these devices. A quick blurb IOS-XE from the previous link is :
IOS XE retains the exact same look and feel of IOS, while providing enhanced future-proofing and improved functionality. In IOS XE, IOS 15.0 runs as a single daemon within a modern Linux operating system. Additional system functions now run as additional, separate processes in the host OS environment. The operation, support and management of IOS XE does not require re-training from classic IOS.
Multi-Core CPUs and SMP
Running IOS and other applications as separate processes also enables load balancing the multi-core CPU, allowing each process to use a different core. IOSd within the IOS XE environment supports multiple threads and multi-core CPUs.Control Plane and Data Plane Separation
IOS XE introduces an opportunity to enable teams to now build drivers for new Data Plane ASICs outside the IOS instance and have them program to a set of standard APIs which in turn enforces Control Plane and Data Plane processing separation.IOS XE accomplishes Control Plane / Data Plane separation through the introduction of the Forwarding and Feature Manager (FFM) and its standard interface to the Forwarding Engine Driver (FED). FFM provides a set of APIs to Control Plane processes. In turn, the FFM programs the Data Plane via the FED and maintains forwarding state for the system. The FED is the instantiation of the hardware driver for the Data Plane and is provided by the platform.
What is nice about this new ISR 4000 line, besides the performance increases, is that Cisco has reduced product line and reduced RU space. When looking at Modular ISR Migratiion Paths graphic below, we can see that the Cisco 1921/2901 no roll into the ISR 4321; the 2911/2921 roll into the 4331; the 2951 is similiar to the 4351; the 3925/3945 roll to the 4431, and the 3925E/3945E roll to the 4451. What is also nice, in case you have not noticed, that most of the 4400 line is either 1U or 2U in space, whereas much of the old 3900 line used 3U of space. Reducing the network rack space is big in the data center where ever RU can sometimes count.
Looking at the read of the ISR4000 series one may notice a few things. The big thing is that the Network Interface Modules (NIM) are larger than what we find on the ISR G1/G2 EHWIC. That means that the old ISR G2 modules will not work in the new ISR 4000 line. Some other key features of the ISR 4000 line are build-in managmenet interface, PoE+ capable ports (4451/4431 can hand out PoE+ without a module), internal service cards for DSPs, and what really caught my attention – On-line insertion and removal feature to chage cards. There is now a software shutdown command for the card that will allow you to remove it. I have not seen this demonstrated, but this is exciting!
When it comes to the models, the third number is the key to the number of slots. The 4451 has 5 total slots, the 4431 has 3 total slots, the 4421 has 2 slots, and so on. Keep that in mind when you are ordering and you will be able to know what you need quickly.
Below is a graphic of some of the connectivity options. All the normal ones we use today are there, some of still on the roadmap for release, but all the old standards are there – T1/T3/Ethernet, etc. We also have support for all our voice functions throughout the whole ISR 4000 line (from the 4321 to the 4451), as well as support for UCS E-Series blaces and Switching modules. One new voice card for ISR 4000 router is a 2xFXS / 4xFXO card – single card with all those ports! No more taking up one EHWIC slot for FXS and one EHWIC for FXO! Another benifit of the larger NIM slot is increased density, for example there will be an 8–Port T1/E1 NIM module.
Since these are running IOS-XE, we also have the ability to take advantage of Service Virtualization. Instead of having to have hardware to do your WAAS deployment, the router can take care of that in software. For WAAS though you will need local storage, and for that Cisco has the NIM-SSD (1 or 2 200GB SSD), a NIM-HD that is a hot-swappable 500G or 1TB drive, and also a SSD-MSATA-200G drive that does not use a NIM slot – this is for the 43xx series of ISR. Today WAAS is available, but there are sure to be more VM Service Virtualization in the future (graphic shows Engery Wise).
For more information on Service Containers, below is a slideshare from Cisco Systems that will provide more information.
For some more information on the ISR4451-X router, you can watch a Techwise video:
[youtube=https://www.youtube.com/watch?v=gFnoMjmUy-Y]
Ben Williams said:
Quite underwhelming, but the biggest takeway is that Cisco now licenses throughput on the their routers. No bueno
jjfry18018 said:
Other vendors do the same as well, it is not just Cisco. What quickly comes to mind is Juniper has the MX5/10/40 and does the same based on a license.
It is actually a nice feature for companies as they can pay as they grow versus having to “roll a truck” to upgrade hardware at sites.
Ben Williams said:
Yes, valid point, but still the ISR g3 is quite underwhelming in terms of throughput. I think they’re going the right path by separating the control and data plane CPU infrastructure, but I’d expect ASR-type speeds and feeds. These specs barely compete with the current generation of ISR routers.
Duane Grant said:
Jeff,
This to me is a bit different than disabling ports and slots as is done in mx5/10/40 and the asr9001-s. In those platforms you still have the forwarding capacity of the ports you license I.e. Mx5 will run 40gbps through the 1x20g card. In the 4351 you have 200mbps hard capped in the virtual asic.
It’s also worth noting that you might experience higher throughput with an isr2 if you are running larger packets and not that many features. In my use case with pat, qos, bgp+bfd and larger packet sizes, I see ~700 Mbps in a 2951. With the same services on the 4351 I will only see 200mbps of throughput because of the shaper unless I pay for the 400 Mbps shaper which is still under what I’m seeing today in the Isr2.
Overall, I’m not sure I like this “hobbling” of boxes, but if it ultimately saves money it’s fine. I’m just not sure that’s where we will end up, because like I pointed out above, you’re getting less performance in a 4351 than a 2951 in certain (many) scenarios and I’m not sure that everyone is aware of that.
Blake said:
Nice boxes, but I can run a gigabit of traffic through a 1921, so regardless of what their marketing people think a 4321 with a hard-enforced limit of 100mbps is not a replacement, let alone an upgrade…