- NANOG - lists.nanog.org

Re: PC Routers (was Re: /24s run amuck)
by james 15 Jan '04

15 Jan '04

: Be real carfull with Zebra, I got stung big time !!! What I run is actually Quagga, despite saying Zebra. Would you share your experiences ? James Edwards Routing and Security jamesh(a)cybermesa.com At the Santa Fe Office: Internet at Cyber Mesa Store hours: 9-6 Monday through Friday 505-988-9200 SIP:1(747)669-1965

8 11

Looking for power metering equipment...
by Alex Rubenstein 15 Jan '04

15 Jan '04

Preamble: We run a colocation center. We sell power to customers. Question: We are looking for something that sits in the PDUs or branch circuit-breaker distribution load centers, that, on a branch-circuit by branch-circuit basis, can monitor amperage, and be queried by SNMP. Considering there are several hundreds of circuits to be monitored, cheap and featureless (all we need is amperage via SNMP) is fine. Looked at things like Square-D PowerLogin stuff, but thats very pricey, and does about 30x what we need. Pointers? URLs? Experiences? Thanks.

6 6

RE: PC Routers (was Re: /24s run amuck)
by Michel Py 15 Jan '04

15 Jan '04

> Deepak Jain wrote: > With a network boot OS for each POP, you can do > version control much much more easily. This is seriously flawed, IMHO. I'd encourage my competitors to do it: after the master image gets corrupted all it takes is a bozo tripping the right circuit breaker and the entire POP is kaput. > QOS, priority/custom queueing are all KERNEL/underlying > OS functions. This also is flawed, IMHO. What if you want to do queing or QOS based on BGP? Michel.

2 1

RE: PC Routers (was Re: /24s run amuck)
by Michel Py 15 Jan '04

15 Jan '04

> Alexei Roudnev wrote: > Purchase SuperMicro U1 server, with 2 9 Gb SCSI > disks (hot swappable). Suddenly that cheap router ain't cheap anymore. > Now, say, announce A crash Cisco IOS. 99.9% Internet backbones > are Ciscos, so this announce breaks few Ciscos around and die > - so you never know about it (and will not have a chance to be > happy that _this announce crash Ciscos but do not crash ZEBRA). > Not bad, of course - you are alive, all Internet is alive. > Now, say, announce B crash ZEBRA (and do not crash Cisco). It > will spread until it reach first ZEBRA on it;'s road - _your_ > ZEBRA. So all Zerbras in Internet crash at once (and you are > unhappy). Another variant: announcement C crashes vendor "A", but not vendor "B" and not Zebras either (put whoever you want for "A" and "B" but there's only two of them on the backbone, mostly). Unfortunately, it takes a few minutes to crash, so it has enough time to propagate all over the Internet before the first "A"s begin to crash. As more "A"s crash "B"s will quickly be overwhelmed and the entire internet soon is down because no matter who coded it, when half of the backbones takes a hike the other half follows. Your Zebra is still up, but it does not do you any good because the entire Internet is down and you're only a small leaf on its edge, so nobody knows that you are still up, and they don't care anyway because no matter what they can't get anywhere. And yes this happens; not to beat or pick on any of the parties, I remember the entire AT&T frame relay network being down nationwide for more than 24 hours (for parts) because (as it is rumored) someone pushed a bad piece of software on a Stratacom switch. [me puts the asbestos suit on] <flame_bait> <FUD> If you have vendor C or vendor J, and all vendor C or J routers crap out at the same time, you're safe. Yes, you were down but so was half of the rest of the world, so it's obviously not your fault but vendor C or J's fault. If you have a zebra one a homebrew PC and all Zebras crap out: 1. You can't blame C or J. Worse, you can't blame anybody else. 2. As a coincidence, a sales droid from C or J will see your boss in the following days, take him/her to a nice restaurant, and will inevitably say "this would not have happened if you had a real router instead of this garage-built crud". 3. The sales droid is full of it, but _you_ are deep into it. </FUD> </flame_bait> Life is not fair, is it? Michel.

2 1

Re: PC Routers (was Re: /24s run amuck)
by Michael.Dillon＠radianz.com 15 Jan '04

15 Jan '04

>If someone were to take *half* the software innovations which have been >made over the past 15 years (a decent fib, interrupt coalescing, compiled >packet matching rulesets, etc) and applied them as if they knew something >about networking and coding, they could very easily produce a box using >off the shelf PC hardware which woops up on a 7206vxr for somewhere less >than $2000. Do you have any evidence that these improvements are not being done? The people building supercomputer arrays using Linux have a need for consistently high pps and bps that is greater than anything we see today on the Internet. They've been working on these types of improvements in device drivers and the OS (Linux, *BSD) for years now. You might not find this stuff in a standard enterprise distro like RedHat or SUSE but it is trvial to source this stuff and integrate it into your own build of the OS. A lot of this discussion has been people guessing about performance issues but few people have taken the time to put together a few boxes with Linux or *BSD and either Zebra or Quagga to trial them. We all go through detailed evaluations when buying C or J boxes, so it's not a waste of time to trial some Z or Q boxes as well to see what they can do. In the end, the resulting performance is affected by so many factors that it can't be predicted without testing. For instance, any weaknesses in the software might be completely nullified by the greater CPU power of a PC platform. And lets not forget that there are other platforms like ARM and PPC. Here's an ARM development system with PCI http://www.simtec.co.uk/products/EB110ATX/intro.html and here's a PPC one http://www.artesyncp.com/products/PM-PPC-440.html And if anyone thinks that ASICs give C and J an speed advantage that others can't touch, then guess again. Nowadays those ASICs are probably programmable ASICs which is a fancy way of saying that they are mostly made up of FPGA cells. It is not that difficult or expensive for people to design and build their own ASIC using cheap FPGA technology from companies like Xilinx. This is basic sophomore level electronics and is simple and cheap enough that people even hack their own MP3 players using FPGAs http://www.pjrc.com/tech/mp3/fpga/ PC-based routers may not be magic bullets but I think we should take them a lot more seriously especially if you want to innovate and offer something that differentiates you from other network operators. In a world where everybody runs C and J networks, there is only one flavor available, vanilla. --Michael Dillon

1 0

Re: PC Routers (was Re: /24s run amuck)
by neil＠DOMINO.ORG 15 Jan '04

15 Jan '04

> This also is flawed, IMHO. What if you want to do queing or QOS based on > BGP? That doesn't make any sense. You could only do the signalling for such a requirement in BGP and that isn't too hard to implement but the actual work to do QoS/queuing are in the kernel/OS/architecture irrespective of vendor or platform. Regards, Neil.

1 0

Re: PC Routers (was Re: /24s run amuck)
by neil＠DOMINO.ORG 15 Jan '04

15 Jan '04

> It is not a joke - we had such scenario few years ago (it was 'gated vs > Cisco and WellFreet vs Cisco'). And such scenario make Juniper back-bone a > little dangerous (but I believe that JUNIPER debugged such problems long > ago, so it is not a case today). Yes this has happened a few times, also things like very long AS paths and address family interfaces would take GateD down alot. [I must confess to not have used Zebra] The fix for this that I deployed in GateD was to add large chunks of code to ignore anything that it didn't understand, some of these were complete botches but it stopped the gated.core's from appearing. Regards, Neil.

1 0

RE: /24s run amuck
by McBurnett, Jim 15 Jan '04

15 Jan '04

Ok, I am often outgunned and off target here. But I have to ask this: 1. If filtering is used, as suggested by someone, what happens to the small/mid-sized company that is multi-homed out of an ISP's /20 or larger block? In this case, I can see an ISP with a /20 bust that up to /21s smaller to accommodate this user. 2. Wasn't /24 filtering something that a few large ISP's did a few years ago and everyone complained? I don't have a reference here but I seem to remember some flack about that. 3. What happens in the case of a carrier that has given /24s to a downstream out of different blocks? I guess the real question is this: If X company can not be reached, how/who would you complain to? And would this be like the RR and AOL email filtering lists where we all complain, and this filtering is an effort by some to force others to clean up their act? Am I out in Left field? Jim

4 4

IP Subnet Management?
by Michael Wiacek 15 Jan '04

15 Jan '04

Hey everyone, I've been trying to come up with an algorithm to describe the assignment of IP subnets. I have something in a proof of concept form that will break a block of addresses into subnets at a user's request. The thing is that the assignments it makes are provably optimal. Within the limits you may place on an assignment, there will never be more than one empty subnet of any particular size. I am curious as to what other things I should try and put into this. Right now, if you request a /16 from a /8, the /16 is considered assigned and further changes within it are not possible. Would 'children' so to speak be useful for you? What would your ideas be? How have you tackled this problem? Feel free to respond off list if you'd like :-) Appreciate your time! Mike (sick of spreadsheets) Wiacek

3 2

RE: PC Routers (was Re: /24s run amuck)
by Michel Py 15 Jan '04

15 Jan '04

> alex(a)pilosoft.com wrote: > o) lack of unified tools to configure and manage: > Each of those tools has varied degrees of documentation, > different configuration interface, vastly different > 'status' interface, different support mailing lists, etc. > It is much easier for a given organization to find a > cisco/juniper/etc expert than to find someone who's > experienced with Linux/FreeBSD network toolchain, and I > don't foresee that changing anytime soon. You summarized it very well. > There are linux and freebsd distributions that aim to > minimize the "OS" layer to suit router better. Linux > also has a filesystem that spreads writes across the > flash area, so you are not likely to write single block > 100000 times in your life. This is exactly where Cisco got their act together: they understand that what's above is exactly the kind of thing that many people are willing to pay more for in trade for not having to research the issue. In the end, time is money: each organization has a finite number of good techs; sometimes there is value in paying more for gear and have your senior techs available to deal with issues that are directly related to the business, instead of optimizing the OS filesystem so it does not wear out the flash. Michel.

1 0