|
falz posted:J MX-80, with software after August to prevent.. yesterday. Haha yesterday was awesome what are you taaaaaaaaaaaaaaaaaaaaaalking about.
|
# ¿ Nov 9, 2011 05:09 |
|
|
# ¿ May 5, 2024 22:32 |
|
nzspambot posted:Ahh Telstra not filtering it's customers, fun stuff I got a notice about that yesterday morning and had a good laugh. "Root Cause: Telstra blackholed everyone" hehe
|
# ¿ Feb 25, 2012 07:09 |
|
Anjow posted:I don't know how much you'd get things for, but for cheap 10GE setups we've been using Dell 8024Fs, 24 SFP+ ports and 16MB buffers. Performance wise they do as well for access stuff as Brocade TurboIrons (cost us around 1.5x as much) but the interface is a tad funky. By which I mean unintuitive. I thought they just have re-branded equipment that wasn't a server for the most part?
|
# ¿ Mar 7, 2012 14:56 |
|
SLAAC is an amazing piece of standard.
|
# ¿ Jul 3, 2012 04:30 |
|
I find MTR better than traceroute with figuring out network issues. Mostly because it'll give you packet drop % on each hop so you at least know where there may be congestion.
|
# ¿ Sep 20, 2012 02:17 |
|
Yo. What model router do people use for internet routing? I would need something that is able to hold full BGP tables. I was looking at the ASR 1004 and maybe the 7603-S as something of a possibility with the correct route processor cards. Anything else in the Cisco catalog that would fit the bill? I'm not familiar with Cisco product families so not sure if I'm missing some obvious solution without reading every data sheet.
|
# ¿ Sep 21, 2012 21:52 |
|
ragzilla posted:ASR1k with an ESP10 or greater I think. Thanks, looks like I was looking around the right products.
|
# ¿ Sep 21, 2012 22:25 |
|
Bluecobra posted:We have two Cisco 3845's each with 1Gb of memory and this has been pretty solid for years. I just checked and saw that we have ~420K IPv4 routes and ~10K IPv6 routes. Yeah that is about what the internet looks like for us. I checked out the 39xx chassis and it looks like it only can go up to 350Mbps. Yikes! Am I reading that wrong or do the the expansion slots matter? It looks like you'll only get line rate on-card with those expansion and as soon as it needs to be routed you run into the 350Mbps issue. For comparison the ASR 9001 can do 120Gbps and the smallest router we're buying now can do 20Gbps. I suppose I should've mentioned it being able to route probably at least 10 Gbps.
|
# ¿ Sep 22, 2012 05:43 |
|
BurgerQuest posted:You're reading it right. Any old router can in theory receive full routes with enough memory, but your actual load requirements may dictate spending more on beefier equipment. Yeah, the ASRlk2 looks pretty sweet since you can software upgrade to meet your needs. Though I'm trying to figure out why someone would get the dual height 10 port SPA over two 8 port SPAs which are single price. I'm going to assume price or niche requirement.
|
# ¿ Sep 22, 2012 23:47 |
|
Powercrazy posted:You also realize that running BGP over a dedicated circuit with a carrier is much more expensive than the usual "internet access" residential ISPs have right? With all them v6 address we should be able to
|
# ¿ Oct 4, 2012 17:05 |
|
Mierdaan posted:This is what I do right now, but it's a pain when new devices are ordered and you don't know the start/end dates for SMARTnet, only that you bought it. Doesn't seem like I should have to harass a rep a few times a year to check on this stuff. They don't give you a spreadsheet with that info on it that you can refer back to?
|
# ¿ Nov 27, 2012 04:14 |
|
Ideally we wouldn't need DHCP but they hosed that up too.
|
# ¿ Dec 18, 2012 02:11 |
|
To be fair on that your RA's will have your default route back out so DHCPv6 would not need that option. What RA doesn't have is such useful information like nameserver information (since you don't really need DNS right?) so to actually complete a full network set up you will need DHCPv6. But if you treat v6 like v4 then yeah you'll be confused as to why it doesn't work. We don't have such a large infrastructure where it isn't possible to handle everything with static assignment at least so some of those pain points aren't felt. Except when your providers forget to turn v6 BGP back on after maintenances doomisland fucked around with this message at 21:12 on Dec 18, 2012 |
# ¿ Dec 18, 2012 21:09 |
|
So for BGP how does everyone monitor their sessions/updates/etc. It's tough to search for some tool that will allow you to see information like total routes, routes per peer, active routes per peer, view BGP updates on X time segments, AS PATH per network and changes to it and other general BGP related stats.
|
# ¿ Jan 12, 2013 04:18 |
|
falz posted:Observium does most of that and a bunch of other stuff. And is pretty and free. Ah, thank you. This looks useful.
|
# ¿ Jan 12, 2013 05:11 |
|
Nitr0 posted:Anything that graphs snmp can do BGP stats. Yeah its more BGP historic type stuff I was looking for. So if a route gets announced and then deannouced I can easily view what changed in the route table. I wasn't aware of many things doing that in a monitoring package. Basically I'm looking for bgpplay but not a web app and with my own BGP feeds. The linked free version of that site doesn't work since the email address on the site to activate the product seems to go to /dev/null
|
# ¿ Jan 12, 2013 17:17 |
|
Anyone have any experience with ASR9K1?
|
# ¿ Jan 23, 2013 00:54 |
|
ragzilla posted:9001? Should be similar to a 9000 running an RSP440 and SE line cards capability wise. I have a couple of 9010s running the old RSPs in one of our DCs. What're you trying to do? Router on our edge that would be taking in at least 4 full BGP tables, probably at least a few more in the future. A 10g here and there and at least 10 1g SFP ports. It seemed of the models Cisco has it would be able to handle it and is in a small form factor. edit: I should note I'm more familiar with Juniper than Cisco so it would be something equal to a MX80 at least.
|
# ¿ Jan 23, 2013 02:50 |
|
teh z0rg posted:Arbor Peakflow. Wicked jealous.
|
# ¿ Mar 2, 2013 04:30 |
|
adorai posted:I thought all the big datacenter boys were moving to software routing these days anyway? If I was building a new datacenter I would run it off of force 10 48 port 10gbe top of rack switches with a core of 32 port 40gbe force 10 switches and vyatta routers. Distributed and cheap (or is it inexpensive?). Not sure what I would do for firewalls, but I would at least look at the virtual firewalls that are out there. If by big datacenter boys you mean Microsoft, Google, etc? Then yeah they don't want any human touching anything but the edge router.
|
# ¿ Mar 14, 2013 01:37 |
|
Bluecobra posted:Microsoft is doing BGP right to the ToR switch: Right, but the goal is to have it completely automated especially after installation. That talk was cool since he went into how they were more interested in hiring programmers than network admins and why. For most people though that is a lot of effort and resources they don't necessarily have. Also the scale is a bit different than most datacenter installs for a company I imagine. I think that NANOG also had some Google folks going into their OpenFlow set ups which is cool. They also have a lot of money to invest into this new tech. As a sort of side note did anyone go to the BAJUG 3 meeting? There was a presentation on puppet + JunOS but I can't find it anywhere online.
|
# ¿ Mar 14, 2013 04:25 |
|
Fatal posted:Do people actually do this? On purpose? Are you asking if people go about separating computers into different networks? As in not having everything in the same network/broadcast domain/vlan?
|
# ¿ Apr 26, 2013 19:52 |
|
We still have copper mostly except for 10g to switches. We have a central network rack in two of our datacenters and its gross. When we're able to do a redo of the DC we're going to go to ToR for sure. Hopefully with some 10g line cards in the routers so no need to break out to a aggregation switch between the ToR switch and the router.
|
# ¿ Apr 30, 2013 22:15 |
|
Bluecobra posted:I am a pretty big fan of Raritan PDUs. The only other smart PDU I have used have been from Server Technology and the Raritan's have a much nicer interface. They also have a serial port with a CLI for out of band management. Raritan also has horizontal PDUs which is good if you have some random cabinet in a colo somewhere and want consistency. We use the server tech PDUs everywhere and haven't had a problem. The CLI is a little rough but its not like you're going in it often. We probably have over 50 of them and only one had to be returned. Another had a fuse blown but it was an older horizontal PDUs had fuses inside that weren't accessible. So half the power ports were useless
|
# ¿ May 4, 2013 06:37 |
|
tortilla_chip posted:Mmmm vampire taps. Speaking of which does anyone have any setups where they do 10g line rate packet capture via optical TAPs?
|
# ¿ May 6, 2013 22:53 |
|
FatCow posted:Missed the cable porn discussion because I was too busy doing this. How do you like the brocade? I hear funny comments from europeans about them
|
# ¿ May 7, 2013 04:20 |
|
Bluecobra posted:Yes, we use Corvil appliances. Their prices for the 10Gb pretty high though due being geared towards the financial community. If you just want raw packet capture, these are the cards they use in their appliance. (Which also cost a fortune) Also, if you are looking for a reasonably priced 10Gb tap aggregation switch, you can turn an Arista 7150 into one. Interesting. I had thought about the aggregation switches but didn't think they really fit our model. Unfortunately the best option in terms of what we're capturing would be a span/packet mirror option from a router but then we could drop some packets. Optical TAP seemed the best route to take and just getting a capture box with a card like http://www.endace.com/endace-dag-high-speed-packet-capture-cards.html I'll have to take a look at the appliances you linked though. Thanks.
|
# ¿ May 7, 2013 04:23 |
|
FatCow posted:They are OK, I'm really not impressed with their 5.3 code so far. We've found a few bugs that *really* should have been caught in QA. Interesting. The impression I got was pretty much what you said. If you're not doing anything intelligent (MPLS etc) than they're fine. We're in a similar industry where PPS is preferred over *bps. Luckily it isn't that large so a router doing ~50mpps is more than enough and switches can do that easily. I know we looked at some of the NetIron products but never went with them in production so we have 3 lying around doing nothing.
|
# ¿ May 8, 2013 04:56 |
|
I think the solution is to move to ISIS
|
# ¿ Jun 20, 2013 03:56 |
|
Powercrazy posted:You just need long range SMF ethernet optics. Nothing super exotic required. Will work in any 10G capable switch from any vendor. I guarantee it's probably a lot cooler than NY7. Is it the Seacaucus one? Also the blue lights get annoying after 12 hours.
|
# ¿ Jun 28, 2013 01:41 |
|
Ninja Rope posted:Arista
|
# ¿ Aug 1, 2013 02:02 |
|
I'm thinking of using a SRX210 with the DOCSIS 3 module as a modem at home. Good idea or bad idea?
|
# ¿ Aug 22, 2013 03:06 |
|
Uhh who would use J Web in any serious fashion? We disable it on all our switches and routers as well.
|
# ¿ Aug 22, 2013 03:28 |
|
Also keep in mind that you do have Global unique addresses that you're assigned which is enough to work with (you requested the correct amount right?). Just because they're globally unique doesn't mean they need direct access to 'the internet'.
|
# ¿ Aug 22, 2013 23:40 |
|
We have a /32 and a /22 I think? Too lazy to look into the IPAM.
|
# ¿ Aug 23, 2013 02:25 |
|
Just have to remember you can only route a /48 on the internet which also means you get huge rear end space for each site. Used to be a /32 too which is how we got such a large allocation. I think now RIR's give out a /46?
|
# ¿ Aug 23, 2013 02:36 |
|
Internet routing owns and that looks like two different paths out of the Uruguay network.
doomisland fucked around with this message at 07:18 on Sep 5, 2013 |
# ¿ Sep 5, 2013 07:15 |
|
FatCow posted:Sounds like you collocate with Telx. That would be right out of their playbook. Or any of the "top 5". It's awesome when a technician mislabels cross connects so the one link thats down you're troubleshooting causes them to disconnect another cross connect since it was mislabeled. At 3am.
|
# ¿ Jan 12, 2014 07:49 |
|
Powercrazy posted:3am? Meh, that sounds fine, how about doing emergency troubleshooting/fixing during the production day. One of your 10G circuits is having an issue, so he unplugs it and "Well, I just lost access to the switch, and Nagios has Christmas-treed up, thanks man." Oh that happened to me in Hong Kong. The 3am was mostly me being awake at 3am. Hong Kong I was trying to tell them they cross wired two pairs of fiber between our cab and one of the patch panels but it's been figured out now and they just need to update their docs. This still didn't stop him from pulling one of the fiber's trying to troubleshoot one of the the circuits out of the patch panel for a live circuit with me 10 feet away. I guess another 15 minutes of explanation was necessary or a translator~
|
# ¿ Jan 14, 2014 06:38 |
|
|
# ¿ May 5, 2024 22:32 |
|
SFP+ ports should be backwards compatible with SFP modules anyways unless that's a Cisco thing where it isn't? e: Yeah just look at the data sheet: http://www.cisco.com/en/US/prod/collateral/switches/ps10902/ps12332/data_sheet_c78-696791.html It'll take SFP's which are like 4 dollars each. doomisland fucked around with this message at 05:37 on Jan 23, 2014 |
# ¿ Jan 23, 2014 05:35 |