Networked storage in the driver’s seat

Networked storage will be the most interesting technology trend over the next few years. The direct attached approach will take a back seat, and NAS, SAN or a NAS/SAN fusion will emerge as the leader in different forms, say Rahul Neel Mani and Gaurav Patra

According to P K Gupta, an organisation should continue to use whatever it has and slowly add more storage. It can add storage or build storage networks seamlessly

At the current rate of growth, enterprise data storage needs are doubling every 12-18 months, but as IT budgets decrease so do storage budgets. In this scenario, customers are growing serious about making better use of the storage they already have with them. It is estimated that most customers are currently using only between 30-40 percent of existing storage because a lot of information is located in islands that are not networked together. As part of the strategy for getting the most out of the least, enterprises are looking at how to get more out of their existing storage solutions. There is a gradual shift that’s happening from the direct attached to networked storage, and the growing thinking is that the information required should neither be delayed nor denied to the person seeking it.

Thus, the full force focus is on how to increase networked storage in the enterprise storage plan.

In a recent report (August 2003), IDC says that interest in SANs (Storage Area Networks) is high, with many organisations indicating they will add to the number of SANs over the next two years. Large organisations tend to have a number of isolated SAN islands, typically dedicated to a single application in a homogeneous SAN; these have usually not been linked with an organisation-wide networked storage resource.

Current situation

In the past two years, awareness regarding IT storage solutions has increased greatly in the country. Corporates have now started to understand the importance of information management, and have increased IT investments in implementing storage solutions, especially after 9/11 brought home the importance of storage. Networked storage was the flavour of the season in 2002-03, and helped notch a solid performance for the storage industry with a 9 percent growth. “Today, in India, enterprise technology buyers are embracing networked storage, be it SAN or NAS solutions,” says Shailesh Agarwal, country manager, storage solutions, IBM India.

However, there is still a long road ahead since Indian enterprises are very investment-sensitive and focused on securing a higher return on investment (RoI) while maintaining low total cost of ownership (TCO). “The Indian market is in the very early stages; to attain maturity it will take at least two years. Most enterprises are in the build-up stage (having deployed half their key applications) and are beginning to consider storage as an important aspect of their IT infrastructure,” says Manoj Chugh, president, India and SAARC, EMC. He says there is a move towards storage consolidation and networked storage. Industry analysts forecast that by 2005 nearly 70 percent of all information storage will be networked. This will allow an enterprise to cut costs and raise productivity by consolidating storage systems and servers. “Today there are islands of information in organisations, and managing them is a major problem. Networked storage can consolidate not only the storage infrastructure but also the entire IT infrastructure,” says Avijit Basu, marketing manager, NSSO, business customer sales organisation, Hewlett-Packard (HP) India.

Idris Vasi says it is difficult to calculate RoI for very small deployments because the up front costs are very high; he feels that when one goes up to 50 ports and 100 ports is when optimisation is really seen

Although Indian technology managers are well aware of the benefits of networked storage, many of them are still not committed to such an infrastructure—neither SAN nor NAS. Experts believe this is partly because of a lack of perceived need, lack of evaluation tools, and lack of confidence that the promised benefits will be realised in practice. They also feel that the storage industry went a little overboard in proposing SAN solutions, and that early adopters paid a hefty price for things such as the HBA (Host Bus Adapter) card—$2,500 each. Besides, there were inter-operability issues. Also, the new storage solutions came with management overheads (due to lack of software tools early on) which added to the cost of maintenance. Plus of course there was the initial high cost. Having put all their eggs in one basket, there was a need for an alternate solution in case the primary storage network went down. Earlier this was not very feasible. Now, with DR (Disaster Recovery) solutions available from OEM (original equipment manufacture) vendors and system integrators, the equation is changing and both costs and risks are reducing with time. Some standardisation is also coming in with the inter-operability issue going away. Which is why this is a good time to incorporate these solutions and build them as a practice within an organisation. One expects that technology centres and proof-of-concept centres will lead users to accept the credibility of such solutions that work for them and deliver value immediately.

There are many out-of-the-box/customised solutions available in the market for those who do not want to change their assets at all and yet want to adopt the networked storage environment in their IT infrastructure. “Storage requirements are constantly growing and information is increasing at an exponential rate, no matter what the size of the company. Networked storage solutions should address the issues of scalability, flexibility, inter-operability and management since each company’s IT environment is unique. Networked storage is not just about storage—it is about information management, accessibility and protection,” says Chugh. It is not just about tailoring the solution to each customer’s requirements, but also about allowing the organisation to address information storage issues—today’s and tomorrow’s.

Inter-operability is another issue. Its scope extends beyond mere hardware connectivity and encompasses switches, drivers, operating systems, and applications. While standards help create a foundation for inter-operability, they usually can’t address the entire problem.

Need for networked storage

According to IDC, many users are yet to understand and commit to a networked storage infrastructure, either SAN or NAS. So the question of whether SAN/NAS is required in the cash-strapped Indian economy has surfaced once again. On one hand we see the need for large investments in IT infrastructure; on the other enterprises are facing tight budgetary provisions. And it is a fact that many companies are not in favour of taking a decision for networked storage.

Unlike the situation a couple of years back, companies are now seeing a boom in business. Transactions are increasing, and the need to service these satisfactorily and comprehensively is also high. In addition there is a need to beef up the infrastructure to support such growing businesses. With storage practices being incorporated in enterprises, it not only creates room for expansion but also builds fault tolerance in networks—something required for a growing business. “If we do not respond to customer requirements proactively then it could affect the business and take us back to a position where we have to start rebuilding every bit. This should be the key positioning statement to enterprises when persuading them to take a serious look at networked storage,” says P K Gupta, director, strategic development, intercontinental operations, Legato. As it is, enterprises both big and small are facing increased pressure to enhance their business efficiency and ensure real-time business continuity. “Many businesses struggle with significant costs, inefficiencies and wastage associated with the traditional direct attached storage (DAS). It is this, along with the constant need to increase capacity, minimise time-to-market, and improve quality, competitiveness, and information sharing across communities which can convince Indian CIOs/CEOs to deploy networked storage in their organisations,” believes Chugh. He says that by doing so they reap significant benefits like lower costs, protection of their IT investments, reduction in back-up/restore time, and faster application testing and data migration. P P Subramanian, country manager, Hitachi Data Systems, believes that large organisations like telcos, banks, and manufacturing and software organisations have either already implemented networked storage or are in the process of implementing it, and he hopes that medium and small businesses will soon follow. “The networked storage benefits are the same for both large and small organisations, so implementation will happen in both segments,” he predicts.

Adds Idris Vasi, director, Asia Pacific, optical storage networking, Cisco, “Another generic example is the virtual SAN concept, where you partition your storage resources. By definition, virtual SAN is something that is implemented in the network. We basically have the VLAN concept—if you have multiple customers and a physical network, you have a single LAN and segment it using VLAN technology. We are taking this technology into the storage world and calling it VSAN. In this we are not competing with our partners but are giving more choice to our customers so that they can implement one of these features either on storage devices or servers, or they can implement it on the network.”

P P Subramanian hopes that SMEs will soon follow the lead of large organisations like telcos, banks, and manufacturing and software organisations in implementing networked storage

Deployment

When a company is considering the implementation of networked storage, it is not essential to replace the old infrastructure; rather, it is possible to utilise and protect the legacy investment in hardware and software and deploy inter-operable networked storage. This will give an organisation some cost benefits. “An organisation should continue to use whatever it has and slowly add more storage. It can add storage or build storage networks seamlessly, and does not require ‘fork lift’ upgrades,” says Gupta. Adds Vijay Pradhan, general manager, StorageTek, “When implementing a low-cost networked storage architecture, it is possible to use existing legacy storage nodes and yet reap the benefits of networked storage. This basically requires some specialised software and middleware.” However, it is important to protect the legacy investment only if it is useful or else some amount of scrapping is recommended. Since there is no standard infrastructure there can be no standard formula. But the usual safe model could be to analyse what is there and what is needed. It is also advisable that one should select a vendor based on openness as opposed to proprietary.

But when an organisation feels it is spending more time in planning storage needs than on the system’s management or performance issues within networks, it is time to rethink strategy and look at networked storage. Mathematical models can be devised but they would not hold good for every enterprise since the criticality of information differs from business to business for various data islands they maintain.

From an information infrastructure perspective, enterprises should also identify a storage solution based on how it can boost productivity and growth without discarding previous investments, and based on how it can lower TCO, improve utilisation of the network, and elevate service levels. “To help accomplish this, enterprises must focus on some critical qualities such as modularity and should invest in automated and open storage management software. They should also opt for storage that is not proprietary but is inter-operable with multiple vendors, and which allows seamless migration as the information needs of an enterprise grow,” says Basu. However Vasi of Cisco feels it’s the other way round. He says that whenever an enterprise is deploying any new technology it takes a while for the RoI to go positive. One of the challenges in the Indian market is the small size of SANs. “These are 8-port, 10-port, 12-port and 16-port. It is difficult to calculate RoI when you have such a small deployment because the up front costs are very high. It is a question of scale. When you go up to 50-port and 100-port then you really see optimisation,” explains Vasi.

iSCSI vs FCoIP

According to Manoj Chugh, the dominant channel network technology for storage access being used today is Fibre Channel

The debate over iSCSI and Fibre Channel over IP (FCoIP) seems to be heating up and will soon reach the level of a full-blown technology war. Although estimates suggest that iSCSI is still three to four years away from reality and FCoIP is the clear winner at this stage, there are heavy bets on the future of iSCSI as a medium to carry storage over networks. In such a situation, the choice of technology becomes difficult. It is worth mentioning here that the iSCSI specification has been frozen this year, and all vendors have been developing products around this technology.

But which of the two should customers chose? “The dominant channel network technology for storage access being used today is Fibre Channel. This is a high-performance transmission technology optimised for the block storage format that storage devices use,” says Chugh, while Pradhan opines, “It does not matter to a customer whether it is iSCSI or FC over IP. The customer is looking for operational and economic benefits, and he will choose the road that helps him achieve those benefits.”

The fact is that in both these technologies I/O is offloaded to the adapter in the server machine, and both use SCSI-3 commands. Again, storage is fast becoming a commodity with time, so one should adopt the technology that gives the fastest RoI. Viewed in this way, users will definitely choose the cheaper one.

However, IP storage is getting a second lease of life with new products and standardisation. Despite promising lower cost and easier configuration, IP storage has had difficulty gaining a foothold in the enterprise. But the tide may be turning since many vendors are ready with IP-based options that ease the transition to the fledgling protocol. For the enterprise, iSCSI is a natural choice for Web servers, e-mail, databases and business continuance applications since it mainly uses server-based replication software and takes advantage of the IP backbones and enterprise networks already in place with low-end and mid-range servers. The rise of iSCSI and IP storage coincides with the declining popularity of parallel interfaces (with their inherent physical limitations) and the increasing adoption of Serial ATA and other inexpensive serial interfaces. iSCSI and IP storage are also gaining popularity at a time when the trend is moving away from DAS and towards low-end devices that will allow smaller enterprises to implement an IP storage network without having to add dedicated support staff.

Networked storage can consolidate not only the storage infrastructure but also the entire IT infrastructure, says Avijit Basu

The Fibre Channel vs iSCSI debate is going on for quite some time. Experts point out that both technologies have their advantages and disadvantages. It is expected that iSCSI will become more popular in the SME (small and medium enterprise) segment in next two years, and that FC will continue to be the choice in the enterprise market since its speed, reliability and level of support is very much liked by its admirers; it is also a proven technology. Nevertheless, Gupta feels that since the cost of iSCSI is less, that will give it a winning edge over FC.

With all the big players recently announcing products supporting iSCSI and making alliances with others who are already big in the IP market, iSCSI is definitely gaining momentum. But experts remind us that iSCSI has some drawbacks. Current iSCSI technology is limited to 1 Gbps—the speed of Gigabit Ethernet—while FC solutions have a maximum transfer rate of 2 Gbps. (4 Gigabit FC is on the immediate horizon.) One expects that iSCSI will eventually be deployed on 10 Gigabit Ethernet, but right now that’s not possible—there isn’t a TCP/IP offload engine capable of handling those speeds. Another observation is that while iSCSI is better suited for small or mid-size companies, or branch offices within large companies, the more expensive Fibre Channel is suited for large enterprises.

Many factors go into deciding between the two storage options. Ultimately it comes down to the fact that FCoIP is simpler in terms of the number of devices as well as in the protocol itself. For a SAN with a core-edge design looking to mirror data to a SAN with a relatively fewer number of devices, FCoIP is the quickest, most economical solution available. On the other hand if one does not have a SAN yet has the leisure of being able to wait and see how iSCSI continues to fare in the marketplace before making investments, then iSCSI may be the best choice. “iSCSI won’t replace Fibre Channel completely in the near future. But enterprises will use iSCSI to consolidate back-up from remote sites and servers. iSCSI SANs are most suitable for organisations with a need for streamlining large amounts of data to store and transmit over the network,” explains Chugh.

As IDC has noted, iSCSI is still in the very early stages of the introduction cycle. Many users do not yet know whether it will be appropriate for their organisations. This cycle is likely to continue for another 18 months to two years before adoption is widespread.

Virtualisation is the reality

Today, in India, enterprise technology buyers are embracing networked storage, be it SAN or NAS solutions, says Shailesh Agarwal

One of the most important trends in the storage space is that customers will see more and more of their storage assets being virtualised. This will take many forms, such as virtualisation of device geometries, combination of storage devices into a single logical unit, and presentation of storage assets to servers in different ways such as file I/O devices, block I/O devices and streaming devices. This trend towards virtualisation will result in an overall lower cost of managed storage for the customer. “Everything will merge; not only the disk subsystems but the disks and tapes (libraries) would be a part of a large network,” forecasts Gupta, while Chugh predicts, “Virtualisation will exist to provide background data movement, background data transfers, data mobility across systems as organisations are doing upgrades and swap-outs and other things that help improve business continuity.” Says Vasi, “Virtualisation is of course the single largest pool of management. Basically, virtualisation is where one is moving intelligence to the network,” to which Basu adds, “It is also expected that large banking, financial, educational and research organisations will be the early adopters of virtualisation.”

Yet certain challenges remain for vendors as far as storage virtualisation is concerned. Being a concept that is being developed by various vendors at various levels (OS-level, SAN switch fabric-level and SAN storage-level), there is no standardisation—all vendors are presently offering their own features and functionalities. The challenge for the industry is therefore to offer the same level of virtualisation with multi-vendor storage subsystems.

gaurav@expresscomputeronline.com rahul@expresscomputeronline.com