Network Unpleasantness Part II

Seek and Ye Shall Find! ... we found a volunteer with network infrastructure experience!

Terry volunteers on the Tech Arts side as one of the weekend Technical Directors ... and now that child checkin falls under my care we've made a connection over the past couple months.  I knew Terry was an IT guy at Bethel College, but I fault myself for not digging deeper into his job role and skill set ... I did invite him to join our IT Team, but he already volunteers heavily for Tech Arts.

Thursday afternoon I sent an email out to people I know are in the IT field, but not part of our IT volunteer team ... basically saying we need switch help and pointing them to my prior blog post and if they knew anyone to please contact me ASAP.  I shortly got an email from Terry saying he was available to help ... turns out Terry is a Network Engineer at Bethel College and handles their switch infrastructure among other things IT related :-)

Terry, Ed and I meet for several hours Thursday night and it was most helpful.  We identified some next steps and the start of a road map to get our infrastructure in better shape.  Basically our first step is to segment the network up so it's easier to find where our problem(s) are.  Step #1 is routable VLANs ... and thus we need a layer 3 router.  Terry has dealt with Cisco and HP gear ... like many, he says Cisco is not worth the cost and thus Bethel is all HP ProCurve gear.

After discussing options and future growth plans we're looking very hard at getting a ProCurve 5406zl-48G and an additional 20+4 GBIC module to comprise our new big daddy core switch ... big unplanned purchases hurt :-(  We'll also move our "core location" from the MDF/DMARK to our newest IDF which will also house our server racks down the road.  While a good chunk of change, the 5406 would give us 60 ports freeing up 2 managed Dell 5324's ... and we free up another 5324 by yanking it from the server rack and running cat6 directly from the servers downstairs to the 5406.  These would replace the remaining 3 non-managed switches of the 13 switches we have in our IDFs which is also a high priority on our road map.  Someone asked in the comments if we're running all gig yet ... yes, were gig to the desktop everywhere.

So no decisions yet, but we've now got a direction in mind.  I'm sure a good portion of Monday will be spent yakin about this :-)

Comments

[hackathology]

Why is cisco not a good choice besides the cost factor?

[Greg]

We use a mixture of Dell 6024s and Cisco 3750s for our layer 3 switches. They seem to work reasonably well. You do get some extras with Cisco - the extra price isn't just the name. You get extra buffering per switch port for example, which is good if you're streaming data at close to wire speed. We tend to use the Cisco (with a stacking cable) for our core switches, and trunk them to the 6024s for our edge switches.

Don't forget you can get refurb Cisco's with a support contract. I can't think of any reason you need to buy new ones. The switches will last forever.

[jason lee]

J, why no fiber between your exisitng server stack and your future location... you could tie in your office to the stack later on with that.

[Jason Powell]

We're not saying Cisco is bad ... in my former job I bought ALL Cisco because I could afford it and our consultants recommended it (that was 5-6 years ago).
My budget is vastly different where I'm at now ... cost is a huge consideration. I just can't justify the cost difference to get Cisco gear ... and our volunteer is very familiar with HP gear which is also key.
I'm also liking HP's lifetime warranty ... and NOT having to pay for smartnet every year.

Jason - we want to eliminate the need for a switch in the server rack so we're doing a couple straight cat6 runs right from the server NICs down to their future home. We eliminate a hop and it frees up a managed switch.

[Jason Powell]

Jason - I should also note that we're only talking 6 cat6 runs ... which is 2 more than we need currently (4 physical servers).

[Clif Guy]

We were on 3Com until three years ago when we started buying HP to outfit our new building. At that time a vendor proposed Cisco but it would have been more than double the cost for no real benefit. We don't have any truly high-demand applications on our network so there is absolutely no reason to pay extra for Cisco.

Our decision to go with HP was wise. We use a 4100 as the core in each of our two main buildings and 2600s in the IDFs. We haven't had a single failure in three years. I can recommend them without reservation. HP is a great company with quality products.

[Dustin]

We put a managed switch in each rack. From a monitoring and traffic control standpoint it lets you limit server to server traffic from hitting other switches and know exactly what servers could be causing a traffic issue. We also don't use Cisco due to price except in one area. iSCSI. There is a performance difference when pushing storage traffic over a switch and Cisco's IOS knows how to handle Jumbo Frames very well. I can't say the same for Dell.

[Scott Reichling]

We started using HP managed switches 4 years ago, on a vendor recommendation. We have a 4104GL as our core switch and 5 2xxx series switches connected via 3 fiber, 1 CAT 5e, and 1 Proxim Wireless Bridge. Our new EqualLogic SAN will run on two HP 2824 switches. We have had an HP 2650 switch fail on us, actually due to a lightning strike, that HP replaced a.m. the next business day. I highly recommend HP’s Procurve gear, based on my experiences, the price point, and the manageability of their gear. Also keep in mind the lifetime warranty and the no cost firmware upgrades.

All of our servers connect directly to the 4104 via a 24 port patch in the server rack that terminates in the MDF.

[Jason Lomonaco]

You could use an opensource solution for routing, like Quagga or Vyatta. They just run on any old server you have laying around.

If you want a full hardware-based router Adtran makes very good ones for a better price than a Cisco.