UNSOLVED

jedijeff

updated

15 years ago

J

jedijeff

1 Rookie

•

31 Posts

0

16108

November 29th, 2011 07:00

Moving from FC to iScsi?

We currently have a Compellent fully loaded with 8gb FC cards. We route through Brocade 48k Directors. This Compellent is only used for ESX storage. 

Our next Compellent purchase we are considering moving to iScsi(10gb). I would like to investigate FCoE, but that does not show up on the compatibility list vmware maintains for compellent.

Does anyone have any experience moving from FC to iScsi on their compellent,,,primarily with ESX? We are not close to approaching any bandwith limitations on our current setup,,,but i worry a bit about increased latency and it also seems every Compellent firmware upgrade(and ESX patches) are always fixing serious iscsi issues. 

Obviously, every environment, traffic patterns are different,,,,so I am not looking for a go with isciso or not,,,I am just looking for any problems or issues people may have experienced in their shop if they made the switch..

thanks!!

  • trichert

    15 Posts

    1184

    0

    Posted February 15th, 2012 15:00

    If you are going to do FCoE you are all set, but you are going to spend more for 10G FCoE gear then you did for 8G FC gear and have 0 performance improvement to show for it.  

    You do NOT want to do traditional iSCSI in a data center/enterprise environment.  Despite all the hype of iSCSI it has all the limitations of the OSI model.  Lots of overhead, collisions, retransmits, it just not a good idea.  FCoE removes the OSI model from the picture and resolves all those issues .. but .. is horridly expensive

  • hallidayr

    48 Posts

    1184

    0

    Posted February 16th, 2012 07:00

    I'm not sure why you'd migrate from 8 Gb FC to 10 Gb iSCSI. At the end of the day the speed is about the same due to overhead. I have no idea what other problems trichert is going on about - as long as your equipment is set up properly it's a competitive solution. Do you have some requirement that all SAN traffic must be IP? I've only heard of this decision being made by folks when they're planning a new deployment where cost, flexibility, and upgrade paths are concerns.

  • trichert

    15 Posts

    1184

    0

    Posted February 28th, 2012 11:00

    Hallidayr,

     

    iSCSI is not, and never will be a competittive solution to FC.  I would take 8GB FC over 100gb iSCSI any day, now FCoE is a different ballgame, and truly a level playing field over IP equipment.

    iSCSI is a TCP protocol, it plays within the OSI stack this means it is subject to collisions, retransmits, encapsulation/decapsulation delay etc, all adding precious ms to the access time of your SAN.  Ultimately access time is more important for most functions then throughput.

  • hallidayr

    48 Posts

    1184

    0

    Posted February 28th, 2012 14:00

    "competitive" doesn't necessarily imply that two technologies are equal, just that there are arguments in favor of each one. iSCSI vs. Fiber Channel is not a black-and-white issue. Fiber channel is absolutely the king of performance, and like you said FCoE looks pretty awesome. However iSCSI can be done quite cheaply and offers more flexibility. For some folks this is a priority, especially under different use cases. A heavy-hitting database is definitely going to need low latency that FC(oE) can offer, but if your SAN is used for something such as a render farm or backups that are not concerned with latency than iSCSI is a competitive solution. Problems you describe with iSCSI seem like symptoms of low-end switches or poor configuration