Storage Networks Start To Converge. NAS Gains Momentum And Respect.

The dynamics of the storage industry are changing at an unprecedented pace. A new millennium begins with annual spending on IT exceeding $1 trillion signaling to us all that change is (or should be) a core competency of the information age. Today, we see many new concepts unfold that were looked at quite differently by most of the industry less than a year ago. A year ago, would we have thought that NAS would be as attractive and scale as high as it now does? Would we have thought that xGigE from the bottom and the new Infiniband channel-based switched fabric technology from the top would squeeze Fibre Channel and offer legitimate alternatives? Could we have seen that new service providers would make the notion of outsourcing our most valuable IT asset, our data, a proposition that we could seriously entertain? Four new waves (at least) have captured significant momentum in a very short time.

NAS Is Gaining Momentum

The NAS evolution has shifted from being considered as a niche storage alternative for low-end networks to a serious storage solution for a much larger market. The differences between SAN and NAS stood out a year ago when the role of a NAS seemed secondary at best. NAS is now forecasted to grow at over 60% in revenue per year and approach $7B in annual revenue by yearend 2003.

The ever-increasing focus on storage has advanced the NAS industry significantly. NAS provides an increase in file serving performance by offloading much of the I/O traffic from the application server while offering easier administration of the storage pool. The opportunity for NAS improves further by offering a non-disruptive storage solution to the existing network with minimum staff training required. Storage devices can be added quickly without the presence of a host server. NAS clearly simplifies the IT manager's life considerably. Costly server upgrades to software and hardware occurs less frequently and can be avoided.

NAS should not be viewed as competition for SAN, but has now established itself as a complementary solution for block data. NAS also uses network interfaces such as Ethernet positioning it for the imminent xGigE bandwidth boom. NAS sales are expected to double in 2000 and the market may grow even faster as solutions scale further and add functionality.

File Management Systems And Software Are Keys

Software for network storage management is clearly the critical path item needed to enable the power of the network/storage cataclysm. As organizations turn to more open computing systems for business critical computing, achieving mainframe class reliability, availability, and performance becomes the grand prize. Standards are being pushed more quickly than before, though vendors still too often view open systems as a threat to their proprietary offerings. Distributed file management systems are gaining increased interest and visibility as they represent the building blocks for a more complete storage management solution. Sys-tems providing intelligent file I/O versus block I/O have gained considerable momentum. Domain fault-tolerance (spanning RAID across multiple arrays), dynamically spreading and balancing I/O across newly added disks, assigning RAID levels by object or volume, and adding storage non-disruptively represent required functionality going forward.

The true SAN market, in particular, is growing no faster than the pace that SAN management software becomes available. Whereas NAS is essentially a plug-and-play solution, the successful SAN requires access software and control software to function properly and fulfill its promise. SAN access software enables the application to use and manipulate the data on a SAN. LUN management, zoning, and locking are examples of access software. SAN control software provides storage resource management, data protection, device performance, and capacity monitoring. A great deal of work in the SAN software area remains to be completed. The location of the SAN software framework is one of the great debates today as the location of the "SAN brains" remains unresolved.

The SAN Fabric Changes

Fibre channel was the hot topic in 1997-1998, then SANs dominated industry talk in 1999. The discussion about the SAN brains aside, the recent activity on the future of I/O interfaces has risen to the top. A SAN over IP? IP is easily the dominant data network protocol and the basis of the global Intemet. Ethernet is the basis of roughly 90% of all LANs worldwide. What does this mean? Where does Infiniband fit in this picture? These rapidly emerging trends suggest a convergence or unification of the LAN, WAN, and SANs. Infiniband is a new bus architecture essentially planned to replace the NGIO, PCI, and PCI-X protocols most commonly used for server clustering, internal memory, and bus I/O transfers. It is not available today, though it has recently gained considerable support. Availability is a few years away. The goals of Infiniband (over fiber) include much higher reliability, availability, and scalability increasing cross-connected bandwidth and scaling faster than microprocessors are scaling. Infiniband will conceivably grow from a server interconnect technology and scale to the point where server architecture meets the switched fabric. The switched fabric looks more like an IP-based SAN every passing day. TCP is also the preferred protocol for most new applications. The simplest way to scale an enterprise backbone is to scale Ethernet.

The xGigE progression is gaining steam as it leverages the huge installed base of Ethernet. Ethernet first appeared in 1982 with a rated speed of 10Mbps. We will see 10Gbps in the late 2000-2001 timeframe. Discussions about OC-768 are underway. The momentum for xGigE and now Infiniband appear to be squeezing Fibre Channel from both ends. The debate around whether Fibre Channel over IP (the FC tunnel) will become the longterm SAN protocol will heat up as the year goes forward.

Storage Outsourcing Arrives

New service providers appeared in force in 1999. ISPs (Internet Service Providers), ASPs (Application Service Providers), and most recently the SSP (Storage Service Provider) are all forms of what has been called outsourcing. The SSP model treats storage (this means outsourcing your data) as a service based on the reality that the personnel pool of trained storage expertise is not growing and, therefore, storage is not or will not be a core competency for many businesses. Most companies rarely realize the impact IT talent shortages have on their IT function. Insufficient resources, as well as, (obviously) rapidly changing technologies make storage difficult to understand and manage. How many companies have a true storage talent pool? Why not consider outsourcing storage and have it backed by Service Level Agreements?

Eliminating storage from an organization can often allow more focus on the core competencies. The SSP energy comes from accelerating storage growth as annual growth estimates range from 60-100% with only guesses for the dot-com growth factor. The cost per megabyte of storage falls between 30% and 40% annually. The cost of managing disk storage today can be as much as ten times greater than disk hardware costs. The expenses associated with backup/recovery software, SRM (Storage Resource Management) software, tuning tools, and potentially more hardware add up quickly. Finally, the SAN model works particularly well for the SSP. The concept of a metro-SAN using a very wide bandwidth fiber backbone enables both low-latency data (online applications) and high-latency data (backup/archive) to be outsourced. Long distance optical fiber can extend the SAN beyond the walls of the business and offer lower management expenses due to consolidation and the associated economies of scale. No longer a future discussion, stor age outsourcing has arrived.

Converging To Zero...

The dynamic changing storage industry is actually part of a much bigger overall moving picture. In the next five years, we will see the "zero convergence factor" change the dimensions and the landscape again and again. We know that storage prices per unit are basically headed toward zero in the next few years. We also observe that CPU prices are headed to zero in a slightly longer timeframe. Global bandwidth, now increasing at a rate of 10x per year, is beginning a pricing slope that will have the price of broadband near zero per unit in a still slightly longer timeframe. Where will the IT focus be in five years when storage, compute power, and bandwidth all get to essentially zero cost? Where will the value be? Change is inevitable. Expect lots of it.