I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?). Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years. The manufacturers still seem to think low traffic routers don't need much memory or CPU. What projections are you using regarding the default free zone over the next 5 years when picking new hardware? -- Mark Radabaugh Amplex 419.837.5015 x21 mark@amplex.net
On Fri, 10 Jul 2009, Mark Radabaugh wrote:
I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?). Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years. The manufacturers still seem to think low traffic routers don't need much memory or CPU. What projections are you using regarding the default free zone over the next 5 years when picking new hardware?
1-10Gb is what you consider a small ISP platform? At the low end (maybe much too low for you), the top models in cisco's 2800 series can hold 1GB of RAM. That ought to be plenty for the next few years. ---------------------------------------------------------------------- Jon Lewis | I route Senior Network Engineer | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
Mark Radabaugh wrote:
I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?). Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years.
More like, ipv4 is going explode the routing table in the next 5 years? On a percentage basis v6 is in fact growing faster. but one of these things is growing at 1k-2kprefixes a week and the other is ~ 2k prefixes total. It's plausible that you need 500k v4 dfz routes at or before 2012 that would be right on schedule from the 250k mark... Fitting a curve to the v6 table growth is an interesting experiment in modeling your expectations. I think it's an excellent opportunity for a synthetic futures market.
The manufacturers still seem to think low traffic routers don't need much memory or CPU. What projections are you using regarding the default free zone over the next 5 years when picking new hardware?
IPv6 is going to explode the routing table in the next 5 years. More like, ipv4 is going explode the routing table in the next 5 years?
more like the routing table will continue to grow, mostly proportional to growth in multi-homed sites and richer inter-provider topology. randy
On 10 jul 2009, at 19:03, Joel Jaeggli wrote:
IPv6 is going to explode the routing table in the next 5 years.
More like, ipv4 is going explode the routing table in the next 5 years?
IPv6 is now at something like 1.2 - 1.4 prefixes per AS. So it will take a LONG time before we reach 100k v6 prefixes unless something changes. IPv4 is already growing very fast, and it's all the small stuff: < / 16. This is 90% of the allocations and 10% of the address space (from memory). This isn't going to be impacted much by the IPv4 depletion. There's always people going bankrupt so addresses flow back to the RIRs. The only way the v4 table is going to explode is if Comcast etc decide that rather than 1 /12 they're going to take 4000 /24s or some such. Now of course the Comcasts of this world really don't want to spend even $1/address, and I think a /24 will be more than $1 after the v4 space has depleted. But even worse: having 4000 contracts with 4000 different people with 4000 different lawyers... Suddenly switching to IPv6 doesn't seem like such a bad deal anymore. Bottom line: a router that can do 500k prefixes is probably ok for the next 3 years but not likely for the next 5. 750k or more should be enough for 5 years, though. And you don't want to buy the router you're going to use in 2015 today.
On 2009-07-10-12:42:24, Mark Radabaugh <mark@amplex.net> wrote: [...]
What projections are you using regarding the default free zone over the next 5 years when picking new hardware?
Geoff Huston, et al provide some useful trending: http://bgp.potaroo.net/index-bgp.html With that said, I've been treating hardware forwarding of 1MM v4 prefixes (or equivalent CAM carving of v6, MPLS, ...) as a minimum requirement for Internet-facing routers with a five-year shelf life. Platforms claiming in the 500-600k range seem prohibitive just tracking current v4 prefix growth, and moreso as v6 adaptation increases and end-users begin to realize that v4 and v6 routing is fundamentally the same, and begin to de-aggregate/advertise v6 space just like they do v4... -a
Let me be the devil's advocate: why would you need full Internet routing? Taking reasonably sized neighborhoods of your upstreams (AS paths up to X AS numbers) plus a default to your best upstream might do the trick. Ivan http://www.ioshints.info/about http://blog.ioshints.info/
-----Original Message----- From: Mark Radabaugh [mailto:mark@amplex.net] Sent: Friday, July 10, 2009 6:42 PM To: nanog list Subject: BGP Growth projections
I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?).
Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years. The manufacturers still seem to think low traffic routers don't need much memory or CPU.
What projections are you using regarding the default free zone over the next 5 years when picking new hardware?
--
Mark Radabaugh Amplex 419.837.5015 x21 mark@amplex.net
I would second Ivan's comment. Unless you are a major transit operator (which beats the "small ISP" requirement), you don't really need a full view, and can do we a limited view with a default route. Arie On Sat, Jul 11, 2009 at 10:27 AM, Ivan Pepelnjak <ip@ioshints.info> wrote:
Let me be the devil's advocate: why would you need full Internet routing? Taking reasonably sized neighborhoods of your upstreams (AS paths up to X AS numbers) plus a default to your best upstream might do the trick.
Ivan
http://www.ioshints.info/about http://blog.ioshints.info/
-----Original Message----- From: Mark Radabaugh [mailto:mark@amplex.net] Sent: Friday, July 10, 2009 6:42 PM To: nanog list Subject: BGP Growth projections
I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?).
Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years. The manufacturers still seem to think low traffic routers don't need much memory or CPU.
What projections are you using regarding the default free zone over the next 5 years when picking new hardware?
--
Mark Radabaugh Amplex 419.837.5015 x21 mark@amplex.net
On 2009-07-12-06:09:12, Arie Vayner <arievayner@gmail.com> wrote:
Unless you are a major transit operator (which beats the "small ISP" requirement), you don't really need a full view, and can do we a limited view with a default route.
Disagree. Protection against big-provider depeerings, interdomain capacity problems, etc is increasingly relevant to smaller sites interested in getting business done. While some will outsource this protection their (non-transit-free) provider, others enjoy maintaining this granularity of control themselves... -a
On 2009-07-12-06:09:12, Arie Vayner <arievayner@gmail.com> wrote:
Unless you are a major transit operator (which beats the "small ISP" requirement), you don't really need a full view, and can do we a limited view with a default route.
Disagree. Protection against big-provider depeerings, interdomain capacity problems, etc is increasingly relevant to smaller sites interested in getting business done. While some will outsource this protection their (non-transit-free) provider, others enjoy maintaining this granularity of control themselves...
Specifically, with full routes, us "small ISP" people can match ASNs with traffic in Netflow to see where our traffic goes/comes from, and thus do capacity/link/peer/transit/traffic planning and problem mitigation. -- Scanned for viruses and dangerous content at http://www.oneunified.net and is believed to be clean.
On Sun, 12 Jul 2009, Arie Vayner wrote:
I would second Ivan's comment. Unless you are a major transit operator (which beats the "small ISP" requirement), you don't really need a full view, and can do we a limited view with a default route.
Until something breaks or the next big depeering chicken fight which causes you to lose reachability to some portion of the net. As Randy might say, I encourage my competitors to design their network that way. ---------------------------------------------------------------------- Jon Lewis | I route Senior Network Engineer | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
Ivan Pepelnjak wrote:
Let me be the devil's advocate: why would you need full Internet routing? Taking reasonably sized neighborhoods of your upstreams (AS paths up to X AS numbers) plus a default to your best upstream might do the trick.
Ivan
We currently do exactly that - dropping anything longer than a /23 with default routes to cover anything missing. Since were replacing the routers anyway I would prefer to have the capacity to do things correctly rather than what, at least to me, seems a kludge. It is a very workable solution - just not one I am completely comfortable with. -- Mark Radabaugh Amplex 419.837.5015 x21 mark@amplex.net
Mark Radabaugh wrote:
I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?). Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years. The manufacturers still seem to think low traffic routers don't need much memory or CPU. What projections are you using regarding the default free zone over the next 5 years when picking new hardware?
I'll give you the Cisco product answer since that's what I know. I'd go with the ASR 1000 product line. At 1-10Gbps you've exceeded what an 7200 (even the G2) can handle. The largest of the ISR (3845) tops out at 1/2 Gbps at max CPU in theory (far less in reality). You don't want a software router though, especially for a SP and especially not for an Internet edge router. The ASR forwards in hardware. The 1002 with no internal hardware redundancy can handle 5 or 10 Gbps and costs a little more than a 7206 w/ NPE-G2 or a 7201 (with the 5Gbps ESP). This is one consideration for replacing my edge 7200s with. The 1004 version currently scales to 20Gbps and can handle redundant RPs. The 1006 module also currently scales to 20Gbps but can handle redundant RPs and ESPs. All the ASRs have internal software redundancy so crashes should be relatively painless in theory, even with a single RP. http://www.cisco.com/en/US/products/ps9343/index.html I'm looking at using the 1002 for my Internet edge and the 1006 for the core are smaller remote POPs. The platform has been out for a year or so and appears to be fairly solid. Justin
Give Vyatta on a decent x86 server a try. http://www.vyatta.com/downloads/appbrief/Vyatta_app_BGP.pdf -----Original Message----- From: Mark Radabaugh [mailto:mark@amplex.net] Sent: Friday, July 10, 2009 9:42 AM To: nanog list Subject: BGP Growth projections I'm looking for new core routers for a small ISP and having a hard time finding something appropriate and reasonably priced. We don't have huge traffic levels (<1Gb) and are mostly running Ethernet interfaces to upstreams rather than legacy interfaces (when did OC3 become legacy?). Lot's of choices for routers that can handle the existing BGP tables - but not so much in small platforms (1-10Gb traffic) if you assume that IPv6 is going to explode the routing table in the next 5 years. The manufacturers still seem to think low traffic routers don't need much memory or CPU. What projections are you using regarding the default free zone over the next 5 years when picking new hardware? -- Mark Radabaugh Amplex 419.837.5015 x21 mark@amplex.net Checked by AVG - www.avg.com Version: 8.5.375 / Virus Database: 270.13.12/2233 - Release Date: 07/12/09 08:20:00
participants (11)
-
Adam Rothschild
-
Arie Vayner
-
Iljitsch van Beijnum
-
Ivan Pepelnjak
-
Joel Jaeggli
-
Jon Lewis
-
Justin Shore
-
Mark Radabaugh
-
Randy Bush
-
Ray Burkholder
-
Tomas L. Byrnes