Announcement

Collapse
No announcement yet.

The Internet

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • The Internet

    I figured I'd post my ramblings on this subject in The Back Porch, as the Internet affects everyone's daily live anymore - not just the techies and nerds at this point. I thought it was interesting, but I'm weird, so you may not think it is interesting at all.

    I work in IT and sometimes it baffles me how little I know about how the Internet actually works.

    Three days ago (8/12/2014), you might have noticed that some of your Internet connections were slow or intermittent; either from your house, your place of business, or wherever you were that day. If you didn't, then whoever planned your work-place infrastructure is doing a good job. Even then, though, it was probably still slow or intermittent, and it wasn't just you. It was the whole world.

    Here's why:



    Basically, if you don't understand what those links are saying, Internet-facing edge-device routers around the world started running out of memory. The route table each router holds (think of this as a contact list that also includes mapped-out directions on how to get to that specific contact) that tells them exactly how to get anywhere else on the Internet grew to a size that some routers in poorly planned network infrastructures could no longer handle. These router tables are constantly growing. Major edge-facing devices have RAM (memory) to process these routes, and a lot of them only have 512 MB (half a gig) of memory. So, when the route tables settled in at over half a gig's worth of data on Tuesday, routers started flipping their shit and dropping traffic or taking a long time to process requests.

    This would not affect something like your in-home router - at least not directly. This would only affect large companies and Internet Service Providers directly, but the connection issues would have trickled down to everyone.

  • #2
    Originally posted by Ratt View Post
    If you didn't, then whoever planned your work-place infrastructure is doing a good job.
    Or at least has somewhat of a clue. If you were running 6509's you should have known this has been an issue for while now. I really hate those routers. Some other business units in my company can't get the bean counters to release funds for an upgrade on a few lingering devices, hopefully they just got incintivised!
    DamonH

    Comment


    • #3
      Originally posted by DamonH View Post
      Or at least has somewhat of a clue. If you were running 6509's you should have known this has been an issue for while now. I really hate those routers. Some other business units in my company can't get the bean counters to release funds for an upgrade on a few lingering devices, hopefully they just got incintivised!
      Yeah. All of our routers are variations of Cisco stuff. ASA 5505, 5510, 5515, etc. Our data center in Houston is impressive, lol.

      Comment


      • #4
        Technically speaking neither 5500s nor 6500s are designed for full Inet routing tables. ASAs are not good edge devices unless only a single ISP. If utilizing full routes with BGP you need a device designed for that.

        Typically we position ASRs/7600s or in the past 7200s (NPE-G2 EOL) for edge Inet routers, beefed up with lotsa RAM if utilizing full routing tables.

        Internal routes should be summarized if at all possible, which is easy if architecture was correct from day one. I see this fail when companies grow from one site to 10 to whatever and never planned properly.

        Can other devices sort of work? Yes, but typically you run into limitations with configuration ability and/or performance.
        Originally posted by MR EDD
        U defend him who use's racial slurs like hes drinking water.

        Comment


        • #5
          I'm trying to remember how long 6500's have been EOL.

          Comment


          • #6
            Originally posted by Ratt View Post
            Our data center in Houston
            Need a new one built? I'm your huckleberry

            Comment


            • #7
              Originally posted by ceyko View Post
              Technically speaking neither 5500s nor 6500s are designed for full Inet routing tables. ASAs are not good edge devices unless only a single ISP. If utilizing full routes with BGP you need a device designed for that.

              Typically we position ASRs/7600s or in the past 7200s (NPE-G2 EOL) for edge Inet routers, beefed up with lotsa RAM if utilizing full routing tables.

              Internal routes should be summarized if at all possible, which is easy if architecture was correct from day one. I see this fail when companies grow from one site to 10 to whatever and never planned properly.

              Can other devices sort of work? Yes, but typically you run into limitations with configuration ability and/or performance.

              Comment


              • #8
                I didnt read the links what caused the spike?

                Comment


                • #9
                  So where is the internetz located?

                  Comment


                  • #10
                    Originally posted by John -- '02 HAWK View Post
                    I didnt read the links what caused the spike?
                    There was no spike, it's just a steady growth over time that finally reached a size that some routers couldn't handle.

                    Comment


                    • #11
                      We have datacenters in downtown FTW, Richardson, Dallas, Houston and and a few others scattered around the globe, and some pretty smart people designing and implementing things. Didn't really notice anything at all. I spend most of my time ironing out issues with the VDI that our legacy desktop teams just don't seem to grasp the concept of, so I wouldn't notice if the Web shut down completely most of the time.

                      Comment


                      • #12
                        Originally posted by Strychnine View Post
                        Nice! That about sums up what I think with your science stuff

                        Comment


                        • #13
                          Originally posted by Ratt View Post
                          There was no spike, it's just a steady growth over time that finally reached a size that some routers couldn't handle.
                          Actually it sounds like Verizon leaked around 15,000 routes. That being said, we're still pretty close to the limit organically for those routers.
                          DamonH

                          Comment

                          Working...
                          X