At a presentation to analysts this week Brocade talked about a growing Fibre Channel market, an expanded HBA product line, new fabric core capabilities such as encryption, replication and deduplication, and FCOE and services. There was no mention of FANs (File Area Networks) but file virtualization was discussed.
The presentation involved a stunning 131 slides, an 11MB PDF download, if you want to look at them.
It was split into four sections: data center infrastructure; server edge; file and data management; services and support. The trends and drivers on Brocade are the usual suspects: data growth; consolidation; virtualization; 'greening', etc. As a data center company primarily Brocade sees two kinds of networking: server-to-server and server-to-storage. Storage is block-level storage area network (SAN) storage with separate server links to content and file servers.
There is no mention of a file area networking anymore, the FAN concept having outlived its usefulness. There is also no mention of wide area data services (WAD) with that ground ceded to Riverbed, Cisco and Silver Peak.
Data Center Core
The Fibre Channel SAN market is healthy and will continue to grow. FCOE (Fibre Channel over Ethernet) will extend FC's presence and begin being elevant at the server edge in 2010. The market will evolve in a risk-averse manner (no disruption please.)
Brocade owns about 70% of the FC SAN market revenue. Virtualisation of servers means more servers need to connect to networked storage. It aims to develop new services in the network and tie its server edge products - HBAs and CNAs (Converged Network Adapters - for FCOE) into its core switches and directors to deliver these services.
Enterprise storage is expected to be predominantly Fibre Channel until at least 2014. IDC has recently cut its iSCSI forecast. FCOE storage-facing components are expected in 2010 and native FCOE-interface storage products are not expected until at least 2011-2012. 8Gbit/s FC is shipping with 16Gbit/s FC demos being expected in 2010, shipments in 2011, and 32GBit/s FC standards being developed in 2012 with a 2014 or so ship date.
There is no sense here that Ethernet will take over the data center anytime soon. Brocade is looking ahead to a multi-protocol world.
What products does it intend to develop?
1. Replication in the fabric - it has the 7600 and DCX (Data Center Switch) blade now and supports EMC RecoverPoint. There will be a next-generation platform in 2009 plus other partners and solutions.
2. Compression and dedupe for WAN replication - Brocade product H2 '09 - unique product - only FC solution - much higher speeds compared to classical dedupe approaches.
3. Fabric-based encryption and security - New Brocade products 2H '08: new encryption switch, DCX blade, 10x performance of the competition, partnerships with EMC/RSA and NetApp/Decru.
- and - Brocade security technology to be the basis for next-generation NetApp encryption solution. This goes much further than the announcement that Brocade will use NetApp key management technology. It implies that the Decru DataFort technology has reached the end of the road and will be replaced by a Brocade engine, at least for FC SAN data encryption.
Then there is FCOE.
FCOE removes FC ports from servers and converges server-to-server and server-to-(block) storage onto Ethernet, an enhanced Ethernet (CEE - Converged Enhanced Ethernet). The FCOE destination is FC storage, FCOE being an I/O consolidation play for servers, a first hop. Brocade reckons that the FC incumbent has a huge advantage in being the FCOE vendor of choice. Like today's FC networks, it does not expect mixed-vendor FCOE - FC networks.
There will be FCOE concept validation in 2008/9, greenfield server-edge deployments starting in 2010, second generation FCOE products in 2011/12 and new FCOE storage perhaps appearing in 2012/13.
Brocade will introduce its first a top-of-rack FCOE switch, by end-2008, and it will be a fabric gateway for server FCOE HBAs and also be a server-to-server and server-to-corporate network link product.
The second FCOE product will be an end-of-rack-row FCOE switch with 24 x 10GbE ports, produced as a DCX blade.
The Data Center Switch (DCX) is positioned as a FC SAN backbone now. It will help server consolidation wih FCOE in 2009 and provide server-to-server connectivity with CEE in 2009. This puts it up against InfiniBand as a server-to-server link with InfiniBand coming at this from the compute node-intensive HPC market. We might expect Brocade to say that the trend is for network consolidation and convergence. Why open the door to eject Fibre Channel if you are going to bring InfiniBand in through it?
Brocade will introduce a 10GbE core blade in 2H '09 with 36 or 48 ports, enabling a 384 x 10GbE port DCX. The blade will support FC, FCOE, IEEE CEE layer 2 and IPv4/IPv6 protocols.
The Server Edge
Brocade reckons server virtualisation changes the HBA game with greater host connectivity intelligence and greater performance required. End-to-end network automation is required to enable application mobility. This is why it is entering the HBA market.
It is providing 8 and 4Gbit/s HBAs for servers, in mezzanine format for blade servers. It will also provide 10GbE FCOE adapters, also in mezzanine format, in the first half of 2009. For some reason Brocade isn't using the Converged Network adapter (CNA) term in its marketng.
It is also now using its own ASIC design instead of the PMC-Sierra-supplied ASIC in its first generation HBAs (whose performance was trashed by Emulex' independent tester.)
It says its HBAs have a hardware FC stack whereas its competitors have firmware stacks, giving it 'the highest levels of performance in the industry,' mentioning 500,000 IOPS per port and N_Port trunking. These HBA's have an on-chip encryption engine for in-flight data protection. It also says they are greener, having an industry-leading power-performance ratio.
Brocade was emphatic that it has major OEM commitments for both FC HBAs and FCOE cards, no names though.
This will take place in Brocade's view across a 10Gbe CEE which is reliable (meaning loss-less), deterministic and has a low latency. It will deliver CEE blades for its DCX in 2H '09 and says it has secured tier 1 OEM design wins for its switch and adaptor with products appearing in 2009.
The applications for this networking are identified as the web, database, HPC - putting CEE firmly in contention with InfiniBand - and (server) tier-to-tier. Regarding InfiniBand, Brocade says it has poor manageability and thin industry support.
CEE is supported by 16 vendors committed to an open standard via the IEEE and IETF. It dismisses DCE (Data Center Ethernet) as a Cisco-trademarked term and expects it to be a proprietary implementation, not backwards-compatible either, meaning a fork-lift upgrade.
The CEE timetable is: 2010 - pre-standard CEE from start-ups; 2009 - standards completed and Brocade products introduced; 2010/11 - initial custmer trials and deployments.
Brocade's strategy is now to provide more capablities to better manage file data in mixed file server and SharePoint environments. It has its existing StorageX (remote office file management) and File Insight (enterprise-wide file discovery) software products and is going to introduce a File Management Engine. This will provide:
- A policy engine to automate file management tasks,
- Compatibility with Microsoft server and networking technologies.
- Content indexing and search powered by Microsoft's FAST
- Open File Migration technology to enable the movement of active files at any time.
Brocade has to find a subtle niche here because its FC OEMs and channel partners, with their own filer and scale-out file storage products will prefer to have such functions carried out by their products and not by a Brocade box interposed between them and accessing servers.
Brocade is trying to find a network or fabric-based file anagement/virtualization role that can be applied to its DCX and Data Center Fabric ideas where there is no help from WAD vendors or filer and scale-out filer vendors, hence the relationship with Microsoft.
F5's Acopia-based technology will be competition here.
The FME timeline is:
2008 - File virtualization - Open Fie Migration - Policy Engine - Tiered storage enablement 2009 - SharePoint file services - Disaster Recovery for file servers and SharePoint 2009/2010 - Content-driven migration policies - File deduplication
It has already mentioned dedulication in connection with WAN replication as a DCX service.
In short, Brocade wants to provide unified file management across file servers and SharePoint.
Fibre Channel rules okay, and iSCSI isn't ging to knock it off its perch. Brocade is going for a linked HBA/FCOE adapter-to switch play to edge out its HBA/CNA competitors, Emulex and QLogic. It's going for an Ethernet-based server-to-server link and competing against InfiniBand, borne into the data center on its HPC chariot.
Brocade isn't being turned over by Cisco in SANs, not at all, but it is finding it hard going in the file area with the FAN concept ditched and a File Management hardware product coming to try and combine file management for file servers and SharePoint.
Copyright © 2008, Blocks & Files.com