Emerging storage protocols such as InfiniBand that once captured the attention of both the IT press and the user...
community have been hard hit by the realities of the current economy. That's not to say that InfiniBand is dead, it's not, but the momentum behind its adoption has stalled as storage managers are being forced to turn their attention to more pressing issues (should I say fires?) of the day such as continually growing storage capacity requirements and shrinking storage budgets.
These nearer term issues are making new technology acquisition programs highly conditional on cost savings and on leveraging existing systems and talent to build larger and more cost-effective storage systems. Witness the renewed focus on storage management and the need for content aware storage systems and the interest in iSCSI. In the back end of storage, behind the controller or NAS head, back end switching and denser drives are also enabling new capacity levels and lowering the cost per megabyte of storage. The current high interest in serial ATA (SATA) punctuates the economics of the day.
The one common thread running through all the technologies that now constitute "next generation storage" is cost savings. Forklift upgrades during a capacity explosion attach such a serious cost barrier to some of these new cost-effective technologies that leveraging existing systems and capacity will be a key element of any transitional strategy. This is where emerging connectivity technologies can help.
Avoiding forklift upgrades
From the perspective of an IT manager trying to avoid a forklift upgrade as new storage technologies do emerge, some very practical implementation questions arise. How do I leverage my Fibre Channel (FC) front end when iSCSI has finally proven itself? Will iSCSI affect my FC back end? I really like the promise of dramatically reduced costs with SATA, but where is the "gotchya"? How will SATA or serial-attached SCSI (SAS) affect my FC back end?
Interestingly, early adopters over the last few years chose FC at a time when IT budgets weren't as sensitive as today so these same early adopters are finding it hard to justify "playing" with a new technology when the old one (FC) works perfectly fine. This "raising of the bar" in justifications means two things. First, the most likely adopters of new technologies will be the same early adopters as the ones who have migrated or are migrating to FC and second, leveraging that existing FC infrastructure is key to lowering the justification bar to introducing new technologies.
Key issues in transitioning to new front-end protocols
On the front end of storage systems either a SAN or a NAS resides. The fundamental difference is whether block-based (SAN) or filed-based access (NAS) is used. Today, a SAN is used in conjunction with the servers connecting islands of information and providing scalable storage. The only battle tested and proven transport technology for SANs is currently FC. What this means to the prognosticator is the inertia behind changing from a large FC infrastructure to something different is quite high, therefore leveraging that base infrastructure is crucial. Part of the inertial resistance to change consists of the millions of person hours the industry has invested in the software and firmware to get the high reliability – availability – serviceability (RAS) that is a prerequisite to real storage solutions.
In the world of NAS, the front end is typically plain old Ethernet making IT managers who have a fear of the unknown quite happy. Theoretically, this is one of the key advantages of iSCSI. However, even with iSCSI the transport technology in the back end behind the controller is dominated by FC and in all probability will never be iSCSI.
The future protocols that may play in the front end are iSCSI and InfiniBand. With FC currently dominating large and mid-sized enterprises, InfiniBand continually rises from the grave then falls back into it. Even server blades, where InfiniBand was designed to dominate, will include FC as a fundamental portal to storage. iSCSI is making progress but it is still in boot camp, not having any real battle testing under its belt yet. As much as the protagonists of these new protocols would like to think otherwise, the great majority of new design cycles in the storage industry today are still going to FC and the rate of FC design wins is increasing, not decreasing. The key here for new technologies again, is leveraging existing systems and infrastructures as much as possible.
New back-end protocols -- confusion or clarity?
Behind all the RAID and NAS controllers is the area known as the back-end. It is in the back end where all of the data that is being moved around in the SAN or NAS is actually being stored. The back end contains all of the disk drives, tape systems and virtualization components that abstract the complexity of storage. Knowledge about the back end will gain in criticality as new protocols begin to appear. Key to these new protocols is leveraging existing infrastructures, firmware and software. Also key is the IT manager understanding how each protocol plays into specific applications in order to maximize the efficiency of their storage while minimizing the costs and risks.
The dominant protocols in the back end are SCSI and FC, but just to help add to the confusion why don't we throw in SATA and SAS? What's an early adopter wannabe to do? The storage industry is now holding out the carrot of cheap and flexible storage that can head towards the nirvana of common system components and infrastructure while ignoring the existing momentum of FC. SATA offers drives at fractional costs to FC or SCSI, but some realities are just now becoming apparent. Is the reliability there to allow use of SATA drives in storage arrays? What would be the effect of SAS on mid-end enterprise storage arrays?
Finding their place in the sun
As the storage industry continues to mature, content-aware data will become the new Holy Grail. A necessary component in achieving content-aware data will be the introduction of storage suited to the type of data stored. No one with a lick of sense will store credit card purchase information on a SATA array any time in the near future. Conversely, why store Suzie's instant messaging banter on expensive FC drives? How about reference data? Reference data doesn't change but relatively quick access to the information is needed – when needed.
Where data is stored, how long data is stored, how quickly access must be gained, how safe the data must be, how much the data changes and a myriad of other factors all play into what type of storage an end user requires. This translates into storage systems that are heterogeneous rather than homogeneous and this is the great opportunity for new and emerging storage technologies. Rather than pure FC in the front end or the back end, the heterogeneity of FC mixed with other protocols, like iSCSI and SATA, leads to a storage solution that still provides the requisite RAS while being more efficient on budgets. Realistic solutions solve the budget limitations while building off of existing infrastructure keeping the forklift out of the data center.
Why emerging connectivity technologies offer a solution
Emerging connectivity solutions will eventually help enable more cost-effective content aware storage, break scalability barriers and help users integrate existing and cost- effective next generation systems. New back-end switching technologies are already enabling the implementation of better storage system management and improved RAS by bringing FC point-to-point switched connectivity down to the drive level where it can used to monitor and feedback performance continuity information. Similarly, by developing back-end switching technologies to make FC- and SATA-based storage play together in the same SAN, content aware storage can be realized, existing capacity can be leveraged and the implementation barriers to new cost-effective storage technologies like SATA can be overcome.
About the author
Tom Hammond-Doel, technical marketing director, The Vixel Corporation
After a seventeen-year stint in the aerospace industry, Tom Hammond-Doel joined Vixel Corporation in 1997. He has played a pivotal role in virtually all of Vixel's product development initiatives, including its award-wining embedded storage switching technologies. In January of 2001, Tom was elevated to the position of Technical Marketing Director. In this capacity, he presents regularly at storage industry events, has published numerous storage technology periodicals and opinions, and has served on the Board of the Fibre Channel Industry Association (FCIA).