Omar Sultan over at the Cisco Data Centre blog in this post claiming that the Cisco 6500 has a fantastic investment protection story.
Let’s face it, the Catalyst 6500 is the poster child for investment protection. It has been shipping since 1999 and is still happily chugging along.
Urrrmppffff. Excuse me while I barf up my my breakfast. Sorry about this. Hang on, won't be a sec. Right, where were we.
The real story isn’t quite so simple. While I agree that that the original Catalyst 6000 (not 6500) was released in 1999 as a Layer 2 device, it has had more than half a dozen supervisors. The first Supervisor 1 wasn’t very good, Supervisor 1A (which fixed all the bugs) then the Supervisor II and IIA, Supervisor III (in several variants) and then on the Supervisor 720 (in its many variants) and, in back to the future blitz, the Supervisor 32 with a cutdown feature set. Lets not forget the Switch Fabric Modules (WS-C6500-SFM, WS-X6500-SFM2)
While the chassis has been constant (except for backplane upgrade for Cat6000 and certain models amd early 6500′s around 2000/2001) you could make the claim that every Supervisor upgrade is a “rip and replace”…..if you were a marketing person. After all, the chassis doesn’t get replaced….. or does it ?
For those of us who actually work in the field, replacing supervisors IS A RIP AND REPLACE procedure. As each supervisor added features and drastically changed the technical landscape, I rarely just “replaced the supervisor” as the risk was took great. We would always buy another chassis and “cutover” to the new unit.
Lets not forget all the line cards – the 61xx cards using the 32GB shared bus, the 630xx POE, 65xx cards that connected using single fabric lines and the 67xx cards that finally, ten years later meet the promises that were made in 1999. Promises that seem conveniently forgotten.
Oh yes, I remember the Cisco Account Mgr telling us in tones of sepulchural joy about the amazing features of the C6000 backplane, and its future capabilities. How we purred over the specification and got excited. As the years passed, we forgot about the promises and lived with the reality. (And they are also true of the Nexus 7000 as I wrote about early last year)
The beautiful future we were going to have, holdings hands as we ran through fields of beautiful flowers. 
The architectural limitations
As someone who has spent over ten hours this week explaining the limitations of the C6500, and how little bandwidth the backplane has and how constrained this makes your designs. The constant review of all the design tricks you need to make it run the way you intended, especially under high load.
Only two fabric inputs per card, no load balancing of those inputs. The Fabric has head of line problems under load, the shortcomings of oversubscribed line cards. The Supervisor 32 with truly limited performance. Lets not forget these things. What about all those line cards, and the rules and tricks you need to know about their performance ? How many queues per port ? Per port group ? Which backplane is it using – the shared bus or the fabric, how many fabric connections does it have ?
Here’s to the Catalyst 6500
But anyway, here’s a toast to the Cisco Catalyst 6500. Its a teenager now with hair in its armpits and a gruff voice. For all its cranky ways, and tantrums, at least it has worked most of the time. It hasn’t been a brilliant product, or the best at anything. But for most of us, that’s what we want – something that just works.
On the other hand, I can still get support. Maybe that’s what “investment protection” really means.
Note to Cisco / Omar: This means less marketing and more concentration on what really keeps Cisco in business – technical support. The IT manager is usually happy when the engineers are not grumbling about the product. Not because of some rubbish about “investment protection”.
Rant over.
PS: Cisco makes a lot of money out of maintenance
A related assertion I see folks trying is that Cisco doesn’t innovate because their switches are X years old. I will tell you, that as a company, we are quite pleased to see a data center filled with dusty Cisco chassis. It means we have given our customers the ability to add capability and capacity based on their needs. It means we respect the investment customers have made with us and it means our engineers are smart enough to protect that investment.
This really means “we made a shed load of money out of selling maintenance on the same old kit ever year. We’ve been raking it in!”



Perhaps just a typo, but the Supervisor 32 is not Layer 2 only.
Thanks Brad. Not sure what I was thinking. Updated.
Hey Greg:
Always appreciate a good rant.
In this case, it also helps make my point. I did not say the Catalyst is a monolithic box that has been unchanged in 10 years. It has evolved over the years in, as you point out, an number of incremental steps, and sometimes the upgrades were not an inconsequential effort. And we have learned from that and come up with things like ISSU on the 6500 and a whole host of features on the N7K.
My point, however, is that the change is driven at the customer’s pace because they want or need new features need additional capacity, not because we got bored and EOL a chassis and decide to not support it and pour our investment into our cool new platform. The Catalyst gives customers granular control to only apply upgrades where they need–they can keep a consistent operations and management environment without having to forgo the ability to incorporate new technologies as needed, where they are needed. For example, f you need to deploy a chassis full of 10GbE is it properly simpler to buy a new chassis already loaded? Probably, but I don’t find most customer deploy a new technology that way. They will take a new technology and deploy it incrementally and sparingly where it is cost-effective or functionally required, then then shift buying patterns as the new technology becomes mainstream. For the rest of the network gear, if its doing what the business needs it to do, leave it alone and let it get dusty.
As for the postscript, a good number of our customers buy their support through channel partners, self-spare and the like, so I don’t think we are “raking in” all that you think we are. But again, it helps make the point. Customers can continue to feel comfortable with their investment because of continued support (heck, we just stopped supporting Catalyst 5xxx hardware last year–12 years after we introduced them and 5 years after we stopped selling them). The lack of hardware churn means that customers can invest in their own spares to save money. The lack of platform churn also means they are not constantly wasting time dong test and qual on new systems. The operations and management consistency means that customers don’t have to keep sending folks off to get retrained and the network folks themselves have highly marketable skills.
Finally, I do agree with you Greg, that, ultimately, we want the engineers happy with the product and we continue to focus on that through both engineering and TAC. But, I am also guessing, somewhere up the line, whomever is paying the bills is also fairly pleased that we worry about investment protection too.
Have a good weekend,
Omar
Cisco Systems
1- you may not have said it’s monolithic, but it was certainly implied in the article.
2 – Cisco maintenance remains “reassuringly expensive”, so much so that it gives partners a chance to offer it cheaper. This has caused me problems for years, since partners consistently do maintenance badly. Bean counters always choose the cheapest and are quick to complain when we can’t meet SLA’s. Of course, “I told you so” doesn’t work very well. (That’s a no win position there).
3 – What you call “investment protection” isn’t a proper definition. But that is why I have a blog so I can complain and imagine that I have changed the world. Which, of course, I didn’t.
Sigh.
Anyone have an idea on future sales life on the 6500 series and is
vPC going to be available or only VSS on the 6500s
Is there going to be a small 7K box available ?
My work (college) uses 6509s and 6513s for port density aggregation – we will continue to need to provide 500 ports @ 100/1000 not 500 ports @ 10 Gb.
A 10 story dorm/residence hall very likely has 2000 ports and has 4 6513s. Currently we use HSRP with multiple vlans and run 1 Gb & 10 Gb fibre from the 6513s to data centre.
DS2