Steve Whiting, Brocade VP EMEA and Latin America was asked how Brocade views the director market? Whats happening it it? What do customers want?
The main engine for growth in this space is consolidation of existing SAN islands. SANs grew in islands using pizza box-style switches. Growth was mushrooming with issues of cabling and complexity. Solving this through storage and SAN consolidation, continues to be a major engine for growth in the storage market place. Whiting considers this requirement will remain a strong one for some time.
These now bigger SANs need more flexibility, scalability, manageability, etc. because there are more users, more ports, more storage arrays and more administration effort. Directors need to meet this newer set of customer requirements. The standard ones are port density; we need more ports, more connectivity into the boxes, and I think youll see bigger and bigger boxes appear.
Brocade has just introduced its 24000 director switch which supports 128 ports in a single domain, up from 64 with the 12000. Whiting says, Well have 256 ports in the future. (Presumably in a 48000 director if the numbering logic continues.)
Directors are bought today with a requirement that they will be able to adopt new technologies. The director chassis will stay in place for some time but the technologies used inside it may change. This removes the need to rip and replace the big box when you move, for example, from 2Gbi/s Fibre Channel to 4Gbit/s or more.
Whiting says, I may want to connect those stranded remote Wintel servers at some point, over IP at some point, I may decide to bring my mainframe storage into my open systems world at some point, so having the ability to mix and match protocols, is important.
Customers are not just buying vanilla devices. They want flexibility and the ability to take advantage of new technology. The 24000 is aimed squarely at these customers. It and the 12000 have the ability to run Fibre Channel, IP and FICON the mainframe storage protocol in the same chassis. Whiting says, All you are doing is mixing and matching blade technology. We need to be protocol agnostic.
"There is a passive backplane in these director switches. As and when the industry moves to 4Gbit/s Fibre Channel or straight to 10Gbit, well be able to accommodate these high speeds within the same chassis." Director switches are expensive and this ability to refresh the technology inside them lengthens their active life.
Faster Fibre Channel
Whiting is not expecting Fibre Channel speeds to increase at the expense of port counts. A 128-port director operating at 2Gbit/s wont become a 64-port director to accommodate 4Gbit/s. In fact, at the same time as the speed increases we are expecting to be able to offer more port density.
The move to 4Gbit/s or 10Gbit/s blades is some way out. Since 10Gbit/s is not backwards-compatible with 4 or 2Gbit/s you will have to have assigned ports; there wont be auto-sensing ports able to operate both at 10Gbi/s and 4 or 2 Gbit/s. Theres a different physical interface. Its expected that the first implementation of 10Gbit/s would be blades with 2 slash 4 Gbit/s ports and a limited number of 10Gbit/s ports.
Whiting sees 10 gig being used for inter-switch linking and high bandwidth applications. But its some way off. Meanwhile; We can trunk four 2gig connections together to make an 8gig pipe and theres almost nothing out there that will stress that in terms of performance. Yes, it does mean you lose four ports but in a 128-port chassis thats a reasonable trade-off.
When will 4Gig happen? I think youll see some moves this year with 4Gig host bus adapter vendors and the SAN vendors ourselves with announcements. THis suggests that in a few months time, perhaps next year, Brocade will anounce a higher port count for the 24000.
What about 10 gig? The situation is different. Whiting things the media will be expensive. It looks like $25,000 a port at the moment, pretty restrictive. He thinks it might appear in the 2005/2006 period as the storage arrays start to require that speed.
Whiting emphasises that, for Brocade, its not about sheer port density; its about networking. This doesnt mean that all the Brocade switch products will run all the protocols. The directors can and will. He expects that mid-range products such as the Fibre Channel router will also have a multi-protocol capability; The Fibre Channel router product is also a switch that can run iSCSI or Fibre Channel over IP. We are moving in that direction as an industry to bring that multi-protocol capability to the mid-range. Entry-level and Edge switches though can be expected to be dedicated to particular protocols.
Fibre Channel and iSCSI
What does Whiting think about the iSCSI vs Fibre Channel debate for entry-level storage networking? He mentions that views have been polarised I the recent past; iSCSI will knock out Fibre Channel or iSCSI, in a countervailing approach, is dead. He doesnt see it as a technology choice. Its really about price and performance. What can I afford and what kind of performance do I needed. Fibre Channel until recently has been prohibitively expensive for the entry level market until recently.
As Brocade has now partnered with Dell and EMC to bring out a sub £6,000 SAN that argument appears to have been routed.
Certainly Brocades view, expressed by Whiting, is that, We need to get Fibre Channel pricing down to a point where it is accessible to smaller businesses. He also says, We need to be able to offer Fibre Channel or iSCSI or both. What this means that Brocade does not see itself as a Fibre Channel switch vendor. It is or is becoming a storage networking vendor, one that wants to be protocol agnostic. For example, Brocade will have iSCSI solutions available at that end of the market.