SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : The New QLogic (ANCR)
QLGC 16.070.0%Aug 24 5:00 PM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: George Dawson who wrote (26636)5/1/2000 8:31:00 AM
From: KJ. Moy  Read Replies (1) of 29386
 
GE vs FC ( a reader's digest version)

There seems from time to time renewed interest of GE for storage using SCSI/IP. I have read some of the arguments and the latest was from Ashok Kumar and Brocade. The report is 16 pages long. I can't re-post the report. I will highlight some of the points and my
opinions on them.

1. GE may be suitable (good enough) for storage.

There is a proposed SCSI/IP standard on the table. Once approved, Kumar argued that GE vendors will start the ball rolling.

My take: FC was made into a standard since late 1980's. Storage and software vendors just embraced FC and start turning out FC drives and sub-systems. IMO it is about to reach critical mass. On the other hand, GE, fast
Ethernet has been around since 1995 (at least). Would these vendors abandon their mega efforts and go with another legacy type technology? Highly unlikely.

2. It is argumentatively SCSI/IP would require much more processor overhead in order to interpret the packet headers and determine where these packets are going. Error corrections for GE would be done at the
software (TCP) level (again, adds overhead to the server). There are talks for incorporating TCP
functionality at the hardware level(i.e. to reduce processor loads at the server level).

My take: If server/storage vendors are going to embrace GE/SCSI/IP, they will be going backwards. The whole argument of the needs for Infiniband is to solve these very limitations that SCSI has run into and to address a broader range of problems that FC didn't quite solve (such as channel speed). Compatibility with legacy networks will become a major issue if TCP is incorporated at the hardware level. It defeats the purpose of having
Ethernet as a single networking protocol. (i.e. GE will not work with old 10baseT or 100baseT devices if TCP is performed at the HW level for new devices)

3. It is perceived as a slight advantage that FC frame size is larger than GE's, thus enhance performance. Kumar argued that GE could propose to increase the frame size.

My take: again, compatibility becomes a major issue.

In conclusion: IMO FC will enjoy at least a two year incumbent status until Infiniband is well defined and well positioned. GE/SCSI/IP may come online then. Anytime a technology is required to support the old legacy stuff, it becomes a burden and will not work well.
GE is no exception. Will it challenge FC ? May be. But, they have many hurdles to overcome. Standard has to be in place. Server/storage vendors has to embrace it. There will be a big question mark when and if new products are actually made. Not only GE switches, but also
packetized SCSI drives and GE switches. The performance tests between these new GE switches and the new FC switches will be done if it gets that far. So far, the major companies have spoken. Remember, both FC
and GE were available when these companies were choosing a technology to relieve the limitations that SCSI has run into. They are embracing FC for the time being. They see FC SAN as a viable solution to relieve storage bottlenecks. I can't see they are moving to another technology soon. For what? To prove a point? To please GE switch makers? Customers just want a cost effective, reliable, robust solution. FC is at center stage now. It needs to take advantage of this window of opportunity and I think it will.

KJ
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext