Multi-Vendor Networking – The Two Edged Sword

Published On 2011/02/22 | By Kurt Bales | Rant

A couple of weeks back, when we recorded Episode 33 of Packet Pushers Podcast, one of the items we had on the list of topics to discuss was that of multi-vendor networks and the recent Gartner report on the topic.

Due to various reasons this topic was taken off the list, but I still had a few thoughts on the topic so I decided to write this blog post to discuss some of them.

The Gartner Report

In late 2010 I attended a HP Australia executive briefing and breakfast in Sydney. The keynote speaker at this event was Mark Fabbi, and I had the honour (luck?) of sitting at the same table as him during the event. Mr Fabbi was the author of a paper called “Debunking the Myth of the Single Vendor Network”. The presentation was aimed at C-level executives along with the various Magic Quadrants that we Network Engineers love to laminate and stick to the walls of our cubicles!

Whilst telling us (well me at least) something we already knew, he laid out in a fashion suitable for presentation to upper management for discussion and approval to add addition vendors to your network.

Suggestions have been made that vendors with deeper pockets have been pushing the agenda for these reports, but I think that at the end of the day there are compelling reasons not to have vendor lock-in.

The right tool for the right job

At eintellego we specialise in building multi-vendor networks. I like to call this the “Chasing Amy School of Network Design”. Anyone who has seen that movie should know what I am referring to, but basically why should I limit my choice of devices to one particular vendor. If there is a better tool to complete the task I am trying to do then I will not put blinders on because it does not come from my vendor of choice.

By way of an example, for quite a few years we had been selling and supporting the Cisco PIX and ASA platforms for our customers because we felt they offered the best overall hardware firewall solution for our customers. Other vendors had products that were competitive but none of them were compelling enough for us to move away from the ASA. Then in mid-2009 in a space of maybe 3 weeks I had two separate customers ask me on my opinion of the Juniper SRX platform. I had previously looked at Juniper equipment and was quite impressed with the Junos Operating System, but when it came to firewalls I was not impressed with the ScreenOS options. I could have easily just dismissed the new product because it wasn’t from Vendor X who I was used to buying from, but after several recommendations from colleagues as well as requests from customers I decided to take a look. And I am very glad I did!

What I found was a product that met several of the short comings I had with the ASA platform (dynamic routing with redundancy etc), was based on the Junos Operating System, and had quite an impressive throughput and capabilities for the price. Today at eintellego we will recommend and sell SRX solutions to our customers before an ASA solution because we feel that in 2011 it is truly a better option.

Multi-Vendor for the sake of Multi-Vendor

One point I should make clear is that I do not advocate multi-vendor when it is not required. For a long time the Security Industry Best Practice promoted the idea of separate firewalls from separate vendors. While there are sensible reasons behind this approach (vulnerability in one platform will be safe on another), often during the implementation one of the two devices was treated worse than the other. Maybe installing the second box was to achieve a “tick in the box” for some industry security certification, or maybe a choice by a previous admin who no longer is there to champion the cause for the device, but sometimes one of these boxes was modified and updated far less than the other.

Why should we make our lives complicated for ourselves? Why would I force myself to use, say, an ASR and an MX series router as my border routers purely to meet the requirement of multivendor. Valid reasons may exist for this solution, buy “multi-vendor” is not that reason.

I need to learn another System?

I often hear from engineers that they do not want to learn another operating system. I usually laugh at this point at tell them their future doesn’t look that bright. Even within a single vendor you have several new operating systems now, an depending on the direction your network takes you, you may find yourself working with IOS, NX-OS, IOS-XR and/or possibly IOS-XE.

When I train my staff in networking concepts, I try to re-enforce the theory behind the solution or protocol more than the commands required to complete the solution. If you understand why Spanning Tree or OSPF behave a certain way or respond to an event in the network, it does not matter what operating system you are on, all you need to do is work out where to find those options. Knowing the right commands is only 10% of the job!

As an example, we recently rolled out 500+ HP E-series switches for a customer. Prior to this project my experience with the (former) ProCurve range was limited mostly to simple Layer 2 designs with a little bit of AAA and SNMP thrown in. After about an hour of playing on the device, I was already comfortable with the CLI and ready to start “translating” the config I was after into the correct dialect.

ProCurve CLI is very similar to IOS you say? Well how about Junos then? I think we can all agree that Junos is significantly different to many of the other competitors, but it only took me a day or two in the lab before I was comfortable enough to deploy some of these in a semi-production network to start getting some real world experience.

If you’re working in the network field DO NOT be afraid of new tech. That doesn’t matter if it from the same vendor or a new vendor.

My Vendor is the only one who does X?

Personally I think this one falls under the same heading of “The right tool for the right job”, but sometimes people have engineered themselves into a corner and decide to continue implementing the same solutions because to change or re-design is “too hard”.

Maybe you deployed a proprietary protocol and now you are locked into that vendor. In a previous post I discussed how I was not opposed to proprietary solutions, particularly when they are the only option for a solution. Does this mean that you should stick with this solution when new alternatives exist? The same driving force behind the implementation of the proprietary solution should drive the review of new alternatives – The right tool for the right job!

I understand that certain environments have strict design/innovation schedules, and that once a network is built it is extremely hard to get changes made, but we should always be looking for the best way to do our jobs and to design our networks. Don’t be designing networks based on the text-book from 10 years ago!

Wrap up

Multi-Vendor is not scary or hard, as long as you do it for the right reasons.

Feel free to share your comments, flames or multi-vendor nightmares :)

Like this Article? Share it!

About The Author

Kurt is a Network Engineer based out of Sydney Australia. He a the Senior Network Engineer at ICT Networks where he spends his time building Service Provider networks, and enterprise WAN environments. He holds his CCNP and JNCIE-ENT, and currently studying for his JNCIE-SP and CCIE R&S. Find out more about Kurt at http://network-janitor.net/ and you can follow him on Twitter.
  • http://twitter.com/SyntaxTerror @SyntaxTerror

    When you look beyond Cisco's range of switches and routers, their idea of a stable/functional CLI or GUI leaves a lot to be desired (I've just spent 2 weeks battling the ACE interface, I've been defeated). You're not doing yourself any favours by going with Cisco in these circumstances and there's a lot of other great players in the field, may as well use that as a starting point to branch out!

  • Pingback: Junipers Redefines the Networking Core and Edge with Single-Tier QFabric « Wikibon Blog

  • Mark Johnson

    If you havn't look at Arista Networks by far the best products iData Center focuses Ethernet switches on the market – the CLI looks identical to IOS so no retraining is required…