NetApp - Standards Watch

EMC & NetApp Donate Code for NDMP

You may have seen this press release last week: “SNIA Creates Technical Work Group to Enhance NDMP Standard Through Software Development” There are several interesting threads in this story, and a cool example of how head-to-head competitors (in this case, NetApp and EMC) get together in the Standards arena and contribute code to raise all boats and move the industry forward.

First, a little background on NDMP… 

NDMP is an acronym for Network Data Management Protocol and is an open standard protocol for enterprise-wide backup of heterogeneous network-attached storage. NDMP was co-invented by Network Appliance and PDC Software (acquired by Legato Systems, Inc., and now part of EMC.). NDMP is the de facto standard protocol for NAS backup.

The NDMP protocol is fundamentally about data movement .The NDMP Specification, currently at version 4 and available at NDMP.org, defines two services:   

• A DATA server – which either reads from disk and produces an NDMP data stream (in a specified format), or reads an NDMP data stream and writes to disk, depending upon whether a backup or restore is taking place.   
• A TAPE server – which reads an NDMP data stream and writes it to tape or reads from tape and writes an NDMP data stream, depending upon whether a backup or restore is taking place. All tape-handling functions, such as split-image issues, are dealt with by this service. 

Each service has a separate state diagram that dictates its behavior, e.g. the tape server can enter the pause state while tapes are being changed by the NDMP client. NDMP messages are categorized into distinct groups or NDMP interfaces, such as SCSI, CONFIG and TAPE, and can trigger state changes. 

So, what’s interesting with this announcement?  First of all it signals an expansion of SNIA’s portfolio of software development efforts, collectively targeted at expanding the universe of storage, storage management and data management interoperability tools. The new SNIA NDMP TWG is initially chartered to create a software development kit for NDMP v4.0. 

Secondly, it is a good example of leading storage vendors donating code to multi-vendor development efforts. It’s a common practice, even between fierce competitors like NetApp and EMC – though in this case I did hear a few comments along the lines of, “Wow, NetApp and EMC agreeing on something. That’s different.” 

But as NetApp’s CTO Brian Pawlowski says, NetApp has been a leading developer of NDMP since the late 90s; we have a strong history in contributing software in the FreeBSD and Linux communities; and we believe the SNIA contribution will increase vendor adoption of NDMP and provide customers with more choices in data protection solutions. 

So, it’s no-brainer, really.

FCoE - A Step Towards "Ethernet Everywhere"?

A major source of buzz at the fall storage industry events revolved around Fibre Channel over Ethernet (FCoE) – a new standards effort which was kicked off within the INCITS T11 technical workgroup earlier this year. Given the amount of hype and general interest, I thought it would be worth outlining what it is, how it relates to other protocols, what it means for customers, and when it might become a reality.

So what is FCoE? 

Well, it’s a new protocol specification which enables Fibre Channel commands to be transmitted natively over Ethernet by using an optional transport mechanism instead of TCP/IP. It is designed to maintain backward compatibility with existing Fibre Channel endpoint infrastructure, and it’s targeted at data center SANs using a "converged" (10 Gigabit) Ethernet network infrastructure.

For FCoE-only networks, it depends on existing optional Ethernet capabilities, such as Jumbo Frames and Pause.  For "converged" networks - i.e., also running TCP/IP, it depends on a new set of capabilities which have collectively become known as “Converged Enhanced Ethernet” (CCE) to provide a suitable transport environment. These capabilities (mostly congestion management and priority handling) are currently work items being considered within the IEEE 802 committees.

An obvious question is how does it differ from iSCSI?  Well, iSCSI is a standard SAN protocol designed to transport SCSI block storage commands over Ethernet using TCP/IP.  iSCSI was designed to take advantage of all the ease-of-use, availability and guaranteed delivery mechanisms provided by TCP/IP, while providing a seamless path from 1Gigabit Ethernet to 10 Gigabit Ethernet and beyond. 

What does FCoE mean for Fibre Channel customers?  Basically, it’s all good news. You can confidently buy Fibre Channel solutions today, and have the option in the future of either following the Fibre Channel bandwidth roadmap to 8Gbit and beyond, or of moving to a 10 Gigabit Ethernet “converged” fabric for the data center, with no disruption to your existing Fibre Channel hardware (HBAs and arrays), fabric management and storage admin knowledge base. There is also the obvious implication that 10Gigabit Ethernet is not only in your data communications future, but also a realistic option for a unified data center network.

One of the problems with standards is “when will it be real”. The FCoE T11 work is on a fast track schedule, and could become a ratified standard by the end of 2008. The CCE work could emerge from the IEEE in a similar timeframe. So there’s a good chance you’ll see real solutions in 2009, or maybe even before.

From a NetApp perspective, this is just another protocol. We support them all, and we have extensive experience in Fibre Channel and Ethernet environments. We are an active participant in T11, and we are involved in the promotion of the technology through the Fibre Channel Industry Association. We even participated in a cool FCoE proof of concept demo at Fall SNW in Dallas.