Storage Peer Incite: Notes from Wikibon’s May 29, 2007 Research Meeting
Wikibon presents iSCSI: DAS is odd man out. Growth in iSCSI has been impressive, prompting some to forecast impending doom for Fibre Channel-based SAN. The Wikibon community forecasts a big uptake for iSCSI but the near-term target will be existing DAS installations. Users shouldn't expect to change Fibre Channel strategies for quite some time.
iSCSI growth has been impressive over the last few years, based on its ability to provide key capabilities like block I/O. Some, including IDC have forecast that iSCSI will take over 30% of the SAN market by 2010. Its growth in fact has encouraged some to regard iSCSI as the ubiquitous winner in an ongoing market confrontation between iSCSI, network attached storage (NAS), direct attached storage (DAS) and Fibre Channel.
Our view is that users should avoid being wrapped up in the type of artificial crises that have traditionally characterized these types of technological discussions and instead focus their adoption practices on the realities of technology capabilities, costs and application requirements constrained by their own personnel capacities. Indeed as iSCSI continues to grow in the market, it will attract significant invention dollars which will improve its performance characteristics dramatically over the next few years. The effects of this will be a relatively quick transition at the low end of the market from a direct attached and even NAS-attached storage model moving more towards iSCSI SANs.
However, the differential at the high end between iSCSI and Fibre Channel that has Fibre Channel generally providing a 25% throughput advantage relative to iSCSI solutions is not going to abate any time soon. In fact, the runway for invention in the Fibre Channel market is at least 3-5 years given the projected growth rates of Fibre Channel. In the intermediate term we see greater iSCSI implementations for increasingly high end applications that can take advantage of improved performance in the iSCSI world with reduced costs. At the same time, we see Fibre Channel suppliers continue to work to drive down the cost of their solutions, highlighting an area of overlap that should be a planning concern for users. However the approach users should take is to focus their time and attention not on an artificially driven consolidation strategy but rather on an appropriate mixing and matching of storage technologies especially driven by new capabilities in iSCSI over the next few years.
Action Item: The growth in the iSCSI market is attracting significant investment from multiple new players. Users should rejoice in the opportunity to utilize new classes of storage technologies driven by an iSCSI enabled explosion in SAN capabilities. However users should follow the market rate of invention and not force a consolidation in their environment from Fibre Channel to iSCSI and should plan on Fibre Channel being the preferred solution for certain application types for the foreseeable the future.
IT organizations have three main choices when it comes to disk storage, including direct attached storage (DAS), networked attached storage (NAS) and Fibre Channel-based storage area networks (SAN). iSCSI adds to the mix by offering another choice for block-based SAN.
Despite the simplicity and lower costs of iSCSI, Fibre Channel (FC) SAN is not going to vaporize anytime soon. Users have major investments in infrastructure ($25B by Wikibon estimates) and skill sets built around FC and there is no compelling business case for converting in the near and mid-term. Indeed FC will continue to evolve as an independent protocol and as FCoE; and be the primary choice to deliver performance and reliability levels required by many applications.
DAS, however remains a multi-billion dollar marketplace today and will be affected by iSCSI. The 1:1 relationship between servers and storage within collocated and rack mounted server environments is ripe for change. iSCSI will allow small and medium-sized operations to improve boot management, simplify backup, automate recovery and enable a variety of 'light' SAN services. This is good news for several reasons, not the least which is it will allow smaller organizations to develop SAN-oriented best practices while larger organizations can support the evolution of iSCSI-based services. This will pressure FC-based infrastructure suppliers to continue to drive cost out of solutions.
Action Item: Users should aim iSCSI at DAS to improve the automation of existing collocated and distributed storage infrastructure. As well, organizations with enough critical mass should consider iSCSI for blade server environments where often the cost of SAN is prohibitive.
One of the organizational issues that will need to be addressed over time is the management of iSCSI SANs. Network administrators will have a great deal of skill and experience to offer, especially if they have already successfully employed Voice-over-IP (both iSCSI and VoIP are very sensitive to quality of service).
It is sensible to restrict the scope of iSCSI SANs initially, and limit them to being within the data center and on a separate network. These limitations are advisable for security and performance reasons (all block based protocols are 'chatty' and do not like long distances). It makes sense for the initial management of iSCSI SAN to be led by the storage group, with significant assistance from the network group.
Long term, iSCSI SANs will operate over longer distances, especially if distance replication works technically and makes sense from a business perspective for an organization. At that point, it will be advisable for the Network staff to treat the storage network just like any other network, and take over the management and monitoring of the network component of the SAN. The management of storage logical and physical volumes should remain with the storage group, as should the decisions of what is the appropriate network and protocol to support applications.
Action Item: IT executive management should allow the storage group to take the lead in implementing iSCSI. Longer term, the storage group should manage storage, and the network group should manage the iSCSI SAN network, ensuring that they have the tools and processes to provide effective control over quality of service.
One of the trade-offs that NAS users had to make was the less than perfect implementation of database and email applications in return for greater connectivity at lower cost. For example, Microsoft introduced what can at best be called tepid support for NAS with Exchange 2003. For databases, Oracle and Microsoft have always shied away from NAS implementations. Microsoft “generally recommends that you use a Storage Area Network (SAN) or locally attached disk for the storage of your Microsoft SQL Server database files because this configuration optimizes SQL Server performance and reliability.”
The ubiquitous availability of iSCSI and its strong support by all major software vendors (and Microsoft and Oracle in particular) means that the main driver for the dubious use of NAS, lower connectivity costs, has disappeared. IT organizations should take advantage of this to ensure that the right technology is supporting these applications.
NAS will still grow strongly; the systems that use NAS the wrong way are a minority. There is a rich set of use cases for NAS users and applications addressing the explosion in non-formatted data.
Action Item: IT management should ensure future implementations of "state" applications (such as database applications) use iSCSI or FC SANs, according to performance requirements. Installed applications of this type should be migrated off NAS when low cost opportunities occur, such as the implementation of a major release level.
While iSCSI shouldn't change FC strategies for users any time soon, suppliers need to start placing bets today. iSCSI significantly increases the market for SAN, especially SAN management software which now becomes economically feasible for small and medium business users. Suppliers are looking at a substantial opportunity to develop low cost management solutions riding on top of the iSCSI wave where simple, cheap and automated are becoming the defining attributes in the market space.
Vendors should not ignore iSCSI, especially given Microsoft support for the standard. Rather suppliers should develop 'everything included' and true out-of-the-box iSCSI products bundling management software with hardware as turnkey solutions. As the market explodes, this will put serious pressure on the high end to respond, especially as it relates to SAN software, which continues to increase as a percentage of overall storage acquisition costs.
For established vendors like Dell, HP, NetApp and newer players such as EqualLogic and LeftHand Networks, this becomes a game of distribution channels, bundled software function and simplicity of solution all aimed at commoditizing the pricing umbrella set by the likes of EMC, HDS, IBM and Symantec. Vendors ignoring iSCSI and focusing only on Fibre Channel solutions risk substantial margin pressure down the road.
Action Item: iSCSI is a multi-year trend that necessitates serious planning today on the part of vendors. Storage and storage software companies need to develop volume strategies and prepare for the coming iSCSI onslaught. Sacrificing some near-term margin to get in the game is always a hard choice but in this case it may be the right one.
The iSCSI v. FC debate will benefit users in three ways. First, it will tease out real insight into capabilities of each technology to support real applications, and ultimately complement each other. Second, it will be marginally entertaining (and good for advertising revenues). Thirdly, in lower-end, block-based application environments, it finally will take storage technology decisions out of the hands of server product folk – both on the supply and buy side – whom traditionally have utilized direct-attach storage (DAS) designs that optimize server product competitiveness at the expense of storage manageability. The clear and immediate effect of rapidly maturing iSCSI technology is to wipe out any modest DAS price/performance advantage with dramatically improved low-end storage flexibility.
Action Item: Across the board, DAS-based server designs should be avoided in favor of NAS and iSCSI SAN (lower end) or FC SAN (higher end) designs.