Open Compute Project- Taking out HP & Cisco from your network infrastructure

Replies

• 

  durga ch

  I am so glad you spoke about SDNs. It is such an intruguing topic to be discussing at this point of time. The concept they are trying to bring in is pretty simple. Seperate the control domain from the data domain thus making the hardware devices just dummy so it doesnot matter what plays below. There are already few companies who are going the SDN way, majority being GOOGLE (they seem to have already put SDN in place) and the main pusher of the technology here is NICIRA, Stanford.
  Interesting is that, big players are advocating SDNs aka cisco/juniper/huwaiee. They seem to have been coming up with their own SDN architectures . For starters, its the controller which will be pushing across forwarding entries into the switch tables , so I am so much intriguied by the routing part , will the controller make the routing decisions and then only push forth the forwarding device.
  Also, there is advocation of having open flow switches for this job, but I am not sure how existing infra will support Open flow?
  Also,any any entreprise opt SDN architecture??? I am not sure, its same as migrating from IPV4 to IPV6, and more over to support SDNs you need competent network engineers as you would need programming skills to manage a network and not just technical know how.
• 

  Abhishek Rawal

  Overall we're about to discuss whether Opensource hardware works or not , right ?
  
  Well, few days ago in phoronix forum, we had very good debate on Opensource GPU project. A team is trying to develop the Opensource GPU, they are willing to know the opinions of users whether people will back the Opensource GPU in kickstarter or not.
  
  In entire debate we mentioned few pros of Opensource GPU, but the cons were more than that. It takes more than two-three hundred millions for the project. And even after backing the project, we won't be sure that OS GPU can perform well or not. It will take plenty of developer's help from within the community to improve the hardware.
  I do think that opensource hardware can be possible & should be implemented, but there will be very few developers to back such projects, as rest will not even understand the hardcore process of developing a hardware. Developing hardware even though you have all resources open for you, requires skills; & the skills which are required for that comes with lot of experience in the hardware domain & practice, which is scarce & thus Opensource hardware project maybe or maynot work.
• 

  Ashraf HZ

  An opensource GPU development is quite deep. I'd guess you'd need an existing GPU designer to lend support, perhaps provide a legacy design to start of as a base.
  
  OpenSPARC (opensource CPU) was well developed because of strong support from Sun Microsystems.
  
  However, we can still focus on opensource hardware on an higher level, that is, to use existing COTS components (propriety or otherwise), use an open design to develop an open system. The open compute project works well because its practical to achieve and provides very strong incentive for businesses to support- lower costs for hardware and energy.
• 

  Ashraf HZ

  durga

  Also,any any entreprise opt SDN architecture??? I am not sure, its same as migrating from IPV4 to IPV6, and more over to support SDNs you need competent network engineers as you would need programming skills to manage a network and not just technical know how.

  Enterprises which have to manage large networks would certainly benefit from SDN. Well, they don't have to migrate everything to SDN, but at least the switches which are highly utilized and often required to configure frequently. I'm thinking network deployments and upgrading would also be a lot easier for enterprises. Just swap out the hardware. It could possibly make transition to IPv6 easier no?
  
  SDN is worth to study at this time 😁
• 

  durga ch

  @ ash : true that! There has been a lot of ruffle around SDN and one logical speculation was that SDN wont /cant overthrow cisco or anyone as such. Frankly, I have been studying about SDNs recently and have yet to come across details relating how controller will handle extensive requests/ or fails , back up controllers?? there seems to have been predecessors for SDN and they never saw light as they needed seperate hardware , hope SDN wont meet such fate.
  deploying SDN on huge enterprise also pose a security threat , don't you think? the controller can anytime be compromised or any rogue dummy devices playing man in middle stuff?? too much to take care of! so it might be possible tht SDN mght not really overtake anyone, but simply easy the job of network management guys you see..
• 

  durga ch

  I am not really sure of IPV6 though, people are so happy using IPv4 , and all those stoneage devices which dont understand IPv6, they just make things worse and i personally don't really like the way IPv6 address is written for matter of fact 😛.
  yes, but in case SDN comes into play may be deployment of IPV6 will be lot easier, given that the switches wont really know for that matter what they are dealing with. I thnk latest version of openflow supports ipv6, nt sure though
• 

  Nayan Goenka

  Talking on security point of view, Cisco has gifted us what network is today and their products are completely standardized and reliable. Opensource hardware may be cheaper and efficient but for conglomerate enterprises viz mega companies, security matters a lot and clients are ready to invest millions of dollars for security implication. And if we start suggesting opensource environment which does not appeal them, they wont just accept it. Unless some brand backs it. On the other hand, all these companies use branded tailor made softwares which on the other hand would have to be tested every instance. Cisco product has a goodwill and in order to beat it, opensource has to go a long way.
• 

  durga ch

  @ Nayan, understandly cisco/HP or for that matter any networking gaint invests a lot of resources into reserach work , hence the by product is of quality, but also note that, these very gaints have a history of buying anything new which might pose them a threat, remember meraki systems which was getting noticed for its work related to cloud ? now its owned by cisco.
  also, security is not something owned by a single or group of companies, its rather a functionality which again might be programmed in the controller , like the controller can command the switch to act as a firewall or a ASA / IDS .
  and as pointed out in previous post, these companies are not oblivious of opensource presence. Infact they themselves are trying t make their presence in this niche area
• 

  Nayan Goenka

  True but brands have goodwill n trustability. No matter how good new tech is, clients won't accept unless its a big name. I work in this field daily, I knw it
• 

  Ashraf HZ

  Nayan Goenka

  True but brands have goodwill n trustability. No matter how good new tech is, clients won't accept unless its a big name. I work in this field daily, I knw it

  Not all the time 😁 Clients also need to think about costs. A lot of businesses wouldn't mind swapping out Cisco with something that's far cheaper and has almost similar functionality. Not many are happy with vendor lock-ins. They buy not because it is good, but because they are forced to. That's how Cisco controls their market.
  
  So in a way, its good that Facebook is leading this initiative. It should instil some confidence in enterprises to know big network centric companies can run happily on open network products.
• 

  durga ch

  Just sharing the interview of Martin Casado - One of the Inventors of the OpenFlow and founder of NICIRA