Diggin' technology every day

December 13, 2011

Extreme Grey

Filed under: Networking — Nate @ 10:51 am

I was on Netgear‘s site earlier this morning planning on filing a support request for one of my home switches, when I managed to resolve the problem myself, at least for the moment, I’m half expecting the problem to return. Over the past 12-13 years or so I’ve never had even one issue with the small (8 ports or less) metal-enclosed Netgear switches so have stuck to them they have worked very well for me (one company I was at bought a single netgear 48-port gig switch which didn’t work very well by contrast).

I remember reading a while ago how Netgear teamed up with Extreme to re-sell their Black Diamond series of switches.

I didn’t think too much of it till I was on Netgear’s site today so I decided to try to poke around and see if I could find the product(s) that were being resold or OEM’d, and I found them, here is one.

The Netgear 8800

When I saw that it just looked so strange! It’s HP-grey in color, not the usual purple I’m used to seeing. Speaking of HP and purple someone at HP recently speculated to me that the 3PAR arrays will likely stick to being yellow instead of HP-grey because it makes them stand out in the data center.

Tangent comin’ hold onto your butts…

While troubleshooting my home network this morning I think I let some of the smoke out of my HP workstation. Which reminds me of this quote I came across on slashdot years ago

 There is no such thing as a "safe" capacitor! They are filled with SMOKE and that smoke is DEADLY. ALWAYS let the smoke out of the capacitors before attempting to handle them! This should only be done by PROFESSIONALS. Do NOT try this at home.

 Always assume a CAPACITOR is holding a charge. And: Capacitors don't kill people, it's the circuit of which the person is a part that is dangerous...

I thought the networking issue may of been somehow caused by the HP box, so I rebooted it, while it was in the midst of rebooting(middle of POST before the screen came up), I powered it off(by holding down the power button), to reset the network chipset entirely. When I did that I heard a weird clicking sound coming from either the HP box(I think so) or my Cyberpower UPS which was right next to it. Within about 10 seconds I swear a little puff of smoke came out of the HP box(I think), there’s a remote chance it was just dust but I don’t think so. I unplugged the HP box and the clicking stopped. Then I plugged it back in about 30 seconds later, which caused it to turn on automatically, it booted like a champ, no errors, the UPS event log reported nothing. So I don’t know what inside the HP box released the smoke but I guess it was not vital?

Back on topic..

Anyways I poked around in the user manuals and they did a pretty good job of replacing all references of the original product and making it look like a Netgear product through and through (with a couple minor exceptions in diagrams).

I remember about 11 years ago now when I was shopping for a Summit 48 on Ebay for my company(this product wasn’t known for quality at the time though I didn’t know it at the time), I came across some Compaq OEM’d Summit 48s that I think were white in color.

If I was building a bigger network I really would be tempted to opt for this Netgear product if for nothing else to see the expression on people’s faces when I tell them I’m using Netgear, not a brand that comes to most people’s minds when it comes to data center networks! Speaking of data center it looks like Extreme’s 40GbE offerings are leading the market pretty good, I’m so proud of them! Hopefully they can sustain the execution and gain market share. They’ve had some missteps in the past which has knocked them back a few notches(at the time), but they certainly have another opportunity here.

I remember when HP used to OEM/re-sell Foundry Networks chassis switches though I seem to recall HP not making any modifications to the chassis itself(at least according to the pictures on the website, I don’t think it even had an HP logo on the thing). The product at the time was the MG8 which I was entertaining for a data center build out back in 2004/2005. I wasn’t going to buy from HP but was just one of those days that I was poking around and came across it on HP’s site.

Oh and in case your wondering my home network used to be powered by Extreme, I had a trusty Summit 48 for many years, which I eventually upgraded to a Summit 48si (which I still have now). I stopped using it many years ago because I just didn’t have enough ports at home to justify the power usage or more importantly the noise, 1U data center switches are so noisy for home use! I went so far as to replace all of the fans in the 48si (I believe I used Sunon Maglev fans) with quieter ones which reduced the noise by at least half but it was still really loud.

The patented MagLev design is based on magnetic principles and forces that not only propel the fan but also ensure stable rotation over its entire 360 degrees of movement. Utilizing the attraction of the magnetic levitation force, MagLev eliminates the wobbling and shaking problems of traditional motor fans. With this new technology, the MagLev fan propeller is suspended in air during rotation so that the shaft and bearing do not come into direct contact with each other to create friction.

(I dig technology even when it comes to fans!)

The Summit 48 by contrast was 2U and had 80mm fans which spun slower and were quieter. At the moment I have 9 devices wired into my netgear-powered home switching network (one 8-port switch and one 5-port). I used to have a couple Foundry load balancers, and a Cisco switch and a couple other things I think but I recycled them with my Summit 48 years ago, was too lazy to try to re-sell them.

I just saw that picture and was just fascinated by it. It also gives me another opportunity to add more color onto this blog.


  1. Wow, crazy rebrand. But then i know some shops who think consumer Netgear stuff is AOK for production systems, so you never know, it might fly . . . just train all the – ahem – tech support staff to push the Extreme tin.

    Thank you for your Sunon link. Very cool. Spookily timely. I want to play with OpenFlow, only hardware implemented, as if it were production, not software emu, at home, because this scratches a personal itch. Most affordable route is a 1U HP 3500 because the needed firmware is GA. This might save me from trying to soundproof the cable closet. (and i distinctly recall the sound of the predecessor 1U HP switch i had there originally.)

    OpenFlow being new to me (and only just arriving in standard SKUs) it certainly does things like L3 software defined VLANs and much more. Looks like time to brush up on Java to write rules, looking at the controllers. Oh, err, it presently seems you need a HA box to run the controller, unless you can afford one of the just released rack – head or core switches (NEC, IBM, Ciena) which run it native. Finally, i get to use the Cat6 i ran about the house, but entirely understandable if I’m being just a little crackpot. ‘Least the upshot will be i loose some rust on my networking knowledge.

    Comment by John (other John) — December 14, 2011 @ 2:55 pm

  2. I just browsed through the Openflow data sheet, I was thinking it was something similar to NetFlow or sFlow but it seems really different — what sorts of experiments are you interested in trying with Openflow ? It sounds like a neat idea I just can’t think of ever being in a situation myself where I would be able to leverage it. I guess because I’m not a researcher ๐Ÿ™‚

    The Sunon fans are pretty cool, though the ones I got at least did not have the wire to measure the RPM, so the switch complained all the fans had failed. Fortunately it didn’t cause any issues with the switch other than alerts in the user interface.

    Comment by Nate — December 14, 2011 @ 10:37 pm

  3. Well, it’s a bit difficult to call my plan pure research, because I’ve found something not – quite – production but commercial which could tolerate quite a high screwup rate. This must be my fourth attempt to reply coherently, because I’m finding it hard to delineate which bits of the stack constitute research without giving too much out publicly. (and on the other hand, bootstrap style, i am continuously re rolling my plans) Super niche / vertical photo hosting for ad network, is about the best I can do.

    On the one side, I need a lot of transaction monitoring, and the other initially at least robust replication of images. This is not scale out noSQL or distributed SAN stuff. A hair vest strength / thoughtfully optimized Postgres box can handle much of it. But the transactional side is something which also has a lot of front end logic (feeds logs and so on back to the DB, not necessarily old school stored procs or warehouse analysis.)

    Now the thing which we all know that sucks about ad delivery, is distribution latency. Hardly as if even the big boys build out Tier 0 networks, but cost of anything half decent is a big barrier to entry. All the stuff i have been tinkering with, likes to think it is connected to a well managed private network. Okay, this is truly a pet project, so even say 3 geo diverse racks, heads, routers and transit, is beyond where I want to start. Because statistically even survival, a outlier data point, would require a lot more kit.

    OpenFlow stands a albeit small chance to allow me to use colo + VMs to control a much bigger network. Moreover, it places load balancing as a first class citizen, provided you can write that. Now, where you want to load balance, in a latency priority network, is right at the head. It’s not strict latency, it is removing the outliers and what I’ll call the ugly tail.

    In an image based game, if you do not replicate everything, might be a good idea to have one bit of software with the latency metric, as well as what was called for, and make a copy request. From nearest neighbor. Moreover, there are calculations which affect what advert is needed which should be a product of what sees the traffic, as well as back end analysis. If, as Robin Harris advocates, there was a standard API for storage, much of this would then be already possible with off the shelf kit.

    My objective is to use the minimum rackspace, using highly reliable kit at core, against as much data as I can coming from distributed VMs. But the OpenFlow angle is only in part how much sFlow type monitoring I can get from those VMs, (and note I need protocol insight for the transaction stuff, so that might be a win for software defined or controlled, as the overhead may level out) without a five figure hard install for each. If you want to consider failure rates, in jest i offer up the age old “half of advertising is wasted”, and loss of impressions is something this model can tolerate.

    I did indeed mistake OpenFlow for a new-fangled sFlow! You’re not alone there. That’s how I found it.

    To a certain extent, this would not be research at all, if – as noted above – I could afford really high end switches. There’s a research / experiment borderline here.

    I just read your post on StorageMojo, describing 3PAR. Thank you so much. Freaking amazing. To the point I wish vendorland would allow me not to go through all the usual dance to get some of that kit. I hate this, – and admire you for your perseverance with attempting secret handshakes to get information. What you wrote nails it in one why the industry is a mess. The salesbods are forever scared to commoditize, or give up their deductible lunches, but i owe some steaks still to a (now in name only) three letter.

    On that theme, because i like VMS things, take a look sometime at ZeroMQ. Then take another, considering ZFS is a bit flaky (sorry to the Illumos / Nexenta boys, I am not right up to date with your stuff) and how transactions – even for little old photo hosting – might want to be first class considerations. There, you have the tip of the iceberg I am scraping broadside.

    So I think the only way to leverage OpenFlow is at application layer. There must surely be a penalty for running it versus hardware direct packetflow. Maybe this is something i can report on in the near future, because I shall be pumping some storage targets on WAN. Also, I don’t see the attraction to write to a JVM as opposed to a instruct an Extreme prompt for usual datacenter. But the difference is it is programmable. I can take data from flows and in theory redirect connexions without drops. The obstacle I see to this being widely used, is the controller / HA side. None of that kit from NEC or Ciena is affordable for now, unless I reckon you are Fortune 500 or have a big angle on what to do with it. Moreover, it is another layer of abstraction, surely a performance hit (no info yet if they implement in silicon or FPGA) and looking at how VMs and storage has gone, it may be so much staring into pea soup hoping for a elegant and unattainable vista. Does it suck to have a cheap outer sphere yet a multi megabuck core? Depends, also on which gets built first. I think – but not tested – that the controller side really wants to be on hardwired networks. So, might be back to buying the top kit all over. But as for staring into pea soup goes, Nate, you’re the man to clarify the consommรƒยฉ.

    to be continued, I hope!


    – john

    Comment by John (other John) — December 17, 2011 @ 2:27 pm

  4. I missed a joke, maybe which I hope is taken nicely as entirely at my expense: If all I was after was pure research, to a extent, good method would not have me trying to seek a result or proof of anything. In the sense that preconception can be a bad thing. Every now and then, i find some benefit in being pure dumb. What I do intend to do though, is put up some machines, try to see if they can behave well as a L3/4 network, all soft, and try to break them by trying to get the system to find some better way to get lots of small files close to a local web server. (this is not a static set of files, which encourages me to think a bit more) But, making no bones about it, I am not super confident I shall be able to write the software to do that well, or even at all. What I do hope to do, though, is find some boundaries as to the gaping holes in my knowledge. Which, for me personally, is research. I would love some day to explain what I practically want to do. I got that down to simple short sentences. I just face the fact that no way do i have the depth of knowledge or talent and learning I would wish for. My practical aim is rather to show up what I do not know, in some greater but hopefully more accurate depth of unknowing. In fact, my aim is to find out a bit more the kind of question I do need to ask of true domain experts. No guarantees, but i have been suggested this is a legitimate deductible, so long as I document the process. To me, real research would be trying to create something as useful as open path shortest, or a routing algo of any kind which could be used broadly. I am simply not in that league of trying to discover things which are widely helpful. In other words, i consider anything I do applied, and 99% at least received knowledge.

    I’m sorry Nate, this is something of a rejoinder which original I never tried properly to explain, but for some years now I have been glued to the business end of life, not the technological one. All I do see, is that there is to my mind the start of a break in how people can set up networks. So I really want to explore this. However, in mind of the Netgear name head of your post, I see little way any of this OpenFlow thing will ever trickle down to your usual retail CPE. Maybe it offers a theoretical way to more cheaply transition to things like v6 and multicast. Give the customers a processor with ports, and everything down to the home could be split as MPLS. That might even be a way to regulate / bill for multicast television. Reconfigurable higher layers might even be a de fact toll / billing possibility. If OpenFlow does allow – as it is suggested – a controller to direct the packets and not mess with any other subnet, then the controller might be a kind of billing mechanism. In other words, cut the broadcast, but not the rest of the connexion. This might even possibly have some amelioration to the net neutrality debate. The whole SOPA thing, and other ideas are all about cutting off actual internet access for supposed infringers. But what instead if the movies and all that were on a v6 subnet, switchable by a controller, and you could drop the nice “satellite tv” and never hurt freedom of general access? I do not think that would impinge as I think Sweden has it, a human right to connectivity. Because this OpenFlow thing, if it can handle multicast and v6 and all that, starts to sound to me more like tuning an FM receiver. The reason I think that might work, is because this new stuff suggests to me that it can handle the idea of multihoming per NIC and so on. Almost like listening on ports, it might allow you to listen to announces, from multiple sources, and not get them confused. This last thought of mine is simply off the top of my head, and I have very low confidence I can argue through the details, without a lot more experience.

    Please take what I scribble as so much pure speculation. I am still naive in the sense that I think technological solutions may exist for intractable pursuits such as the media industry, which I find – working in that field – has Neanderthal leanings on the best of days. I hope it’s cool I just am thinking out loud. This is self serving, but I do believe my application can stretch some legs, and my experience tells me that the obstacles to that initial thing, getting a paying account e.g., are not very technological. So my dream is not so much to nail the business, or write an awesome clever large count LOC, but instead to find out on my own ticket, if I can articulate what I hope is a useful idea, from the top business level down to how it is all set up. (with help, sure) If I can do that, I have far better chance of meeting and engaging proper talent. On a personal level, I just want to write some lines which control something I can test to oblivion in real time. With a little bit of money at stake, but not the whole farm.

    Please forgive me my waffle. It is quite a few years since I ever had a clear undistracted day, pursuing just the code. I actually fear I am better at the business end than programming. That does not really say a lot in my favor, I am way happier dealing with abstracts than I am with a IDE or EMACS / SLIME. But that is really where I want to be again, with a bit of full circle maybe, until I learn what this new kit does well enough I can report back. Otherwise, I had better brush up my MBA jargon or something ๐Ÿ™‚

    If it’s cool with you, I shall try to write up what I find. I don’t think I’ll have the needed kit before mid Jan, however – totally missed any pre holiday timeframe.

    Happy Christmas Nate, and do please forgive me for being a pest. I am trying as much to provoke new thinking in me, from what I scribble, so all pie in the sky. Yours, – j

    Comment by John (Other John) — December 18, 2011 @ 10:01 pm

  5. love those big comments! If your interested in front page action let me know ๐Ÿ™‚

    I am in Atlanta installing my gear, 3par f200 going in as I type so I’m really busy and can’t read the big comments yet but wanted to thank you for them and I will read them as soon as I can!

    Comment by Nate — December 19, 2011 @ 11:34 am

  6. Hey Nate, my fave Cous was 30yrs in Atlanta, selling sweet fizzy stuff! As was my best mate in London, for some years – slightly different, ahem, bartender extraordinaire. Sorry for the wall of text, I promise you it is all flowcharted (affine transformations and angle of view depending) . . you had me trying to dink it to first base, so comes out a bit funny. Squint a bit more! ๐Ÿ™‚
    Must get back, thank you & Happy New Year to all your lot! – j

    Comment by John (other John) — December 26, 2011 @ 4:43 pm

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress