What's In Store 2004: Storage Network Infrastructure

What's New in 2004?
In 2003 we saw a lot of activity in the storage network infrastructure space. We saw Cisco make a big threat to the incumbent switch/director vendors, yet, by the end of the year their footprint wasn't anywhere near as large as they envisioned. We saw port prices decrease either through OEM pressure or competitive posturing. A number of new and established companies announced products aimed at implementing storage services in the fabric. Port counts are increasing, SANs are being implemented in larger numbers and iSCSI is finally ready for prime time.

So the question is will there be anything new and different in 2004? We believe we will see quite a bit of activity in the space, although we are not expecting too much in the way of technological innovation. Instead, look for products to move from power point and beta to GA, vendors will finally be shipping the products they have been developing for years. We believe the real changes will be in how the switch vendors position and sell their solutions to end users. We expect a big push of low cost SAN and iSCSI technologies into the SMB/SME and mid-tier markets. We also expect that the large vendors will be pushing end to end integration, where one vendor's solutions can cover all of an organizations storage networking requirements, from data center to edge to remote office and DR sites. Intelligence in the fabric will become a more important purchasing factor; however we don't expect that many organizations will implement these technologies in 2004. Overall, expect brutal competition between switch vendors, more acquisitions (yes there are more companies that can be acquired), and perhaps a few surprise entries into the storage market by "traditional" LAN and WAN switch vendors.

Top Trends
Intelligence a must- but not deployed this year
Seeing intelligence in the fabric was all the rage last year, let's start with this trend. First, we must define what intelligence in the fabric really means. Most every vendor that touts intelligence in the fabric (Cisco, Brocade, McDATA, CNT, Maranti, Maxxan, Candera, Troika etc) is actually referring to the ability to run storage services in the fabric. Various storage services software, such as volume management, replication, mirroring, and snap shot solutions sit on the switch/platform, and those services run in the fabric. There are a number of benefits to running these services in the fabric as opposed to in the host or on the array. Customers can realize significant cost savings in management, licensing and capital costs when using these technologies.

Even though we believe the fabric is truly the best place to implement certain services (replication, mirroring, storage pooling), we don't believe the adoption rate of the services in the fabric will be significant this year. However we DO believe that users will make current purchasing decisions based on the vendor's ability to support intelligence in the fabric in the future. Many organizations are looking for longer product lifecycles, and the expectation is that they will eventually run (some) storage services on a fabric based platform, so they are looking for solutions that will meet future requirements. Six months ago, ESG did quick survey of end users, asking how many felt that having intelligence in the switch was a key criterion to their purchasing decisions; less than 20% of respondents felt it was important.

Today, similar surveys are showing that over 30% of users say it is necessary for purchase decision, and another 30+% believe it will be a requirement in the next 2-3 years. However, most users we talk with are only experimenting with fabric based services, implementing them in test scenarios or in non-customer facing applications such as backup and DR. Users see the potential benefits, but are still moving cautiously with the new technologies.

Expect that the ability to support intelligent services in the fabric will be a purchasing requisite for director class switches in 2004; however implementation of intelligent services in the fabric will not move to production until early 2005. Look for a greater adoption rate (vs. director class intelligent switches) of smaller port count switches with intelligent services (Maranti, Brocade), or network storage services platforms (Troika, Candera FalconStor/Network Engines) which enable users to implement fabric based services without the need to change out their current infrastructure or buy into a new director platform. We expect that the need for these "stop gap" solutions will diminish as fabric based services become more of the norm; however that will not be for another 2-4 years.

One other note, there is a different type of intelligent switch hitting the market, we would classify as an "Intelligent Network Switch". What's the difference between this and Intelligent Services Platforms/Switches? Network intelligence is about sensing traffic patterns and monitoring the performance of the fabric itself, and intelligently and dynamically reconfiguring the network, distributing the bandwidth to meet QoS levels. Think of this as bandwidth on demand, which telecommunications vendors have strived to achieve for years. Currently Sandial is the only vendor on the market with this type of solution; however we are hearing rumblings of other vendors trying to provide some similar type features.

End-to-End Architectures, Moving out of the Data Center
Two years ago, most every one knew the statistic by heart: less than 10% of all available storage was networked. That is no longer the case as adoption rates of SAN technologies have shot up in the past few years. In a recent ESG research study, over 70% of the respondents had networked storage of some form (SAN, NAS, file servers) in their environment. ESG estimates that over 30% of available storage is currently networked via FC SANs, and this will increase by 10% in 2004. What has caused the great jump in implementations? Bottom line, end users are seeing the light, and understand that there are significant benefits to networking storage (and yes, NAS implementations are networked storage, but because there is no need for special switches on the front end we won't cover that trend here).

This year will be the year of consolidation, and by this we mean consolidating storage resources onto SANs and consolidating existing SAN's. We expect organization that have multiple SAN islands to network those islands together; remote SANs will be connected via FC/IP gateways, and multiple data center SANs "islands" will be consolidated onto director class solutions that support autonomy (either with VSAN's, hard partitioning or SAN routing).

We expect the entry level SAN market to increase significantly this year as companies like HP, Dell and even Microsoft (with Q-logic) promote low cost, "pre-configured" SANs to the SMB and SME markets. McData will be making a strong push into this space, as has already been witnessed with their partnership with IBM on the X-series platform. Of course vendors will also be pushing iSCSI to this market sector, however we expect a much slower adoption rate of iSCSI (vs low cost FC SANs) due to the maturity of the FC market. All of the major switch vendors have some form of IP (iSCSI, FC/IP) offering as part of their portfolio, which enables them to offer multiple networking options to customers.

One thing we can be sure of, the marketing messages from each of the "portfolio" switch vendors will be focused on end- to end infrastructure support and integration. ESG surveys indicate that over 70% of end users would prefer a single vendor for their end-to-end storage switch infrastructure (so much for interoperability). For the most part, the technology from each vendor is on par (that is if Brocade actually gets a new high end solution out the door this year), or at least it's close enough that enhanced feature set and completeness of the portfolio will win customers- not throughput and number of ports. Fabric management applications, security features and diagnostic capabilities will all be important factors for customers choosing an end-to-end storage network infrastructure supplier.

Of course Cisco has the broadest portfolio of enterprise switches to complement their storage offerings, and they leverage the single vendor card in their massive installed base. However, McData has perhaps the strongest end-to end storage infrastructure portfolio, and surprise, CNT is not too far behind. Brocade of course has a wide breadth of products, but needs to enhance their high end, and they still do partner for much of their distance interconnections. All offer fabric management and security features; again each solution has strengths and weaknesses. The point is customers should now be evaluating not just the port count and performance of switches, but should also focus on how well the fabric management solution fits into their enterprise management platforms and if vendors can securely enable BC and DR schemas.

Expect to see quite a bit of your sales representative this year. Switch vendors are going to aggressively sell "in front of" the OEMs, ensuring that the customer request their solutions from their storage array vendor. The message will be all about end- to end solutions, and single vendor support- we believe this is exactly what the customer is looking for and should result in increased revenue across the board for the portfolio vendors.

How does IP fit in?
Is this the Year of iSCSI? Wasn't that last year? We'll admit it, we have been big proponents of iSCSI from the start, and really did believe it would have taken off by now. We're still bullish on iSCSI, but we're not predicting that it will take over the world anytime soon. Certainly, a number of milestones occurred in 2003 that will make this a better year for iSCSI. The standard was ratified, Microsoft is supporting a driver, and a number of vendors have introduced target devices. Performance has always been a question, however ESG labs has run a number of tests on iSCSI based solutions, and found that performance is on par with FC for most applications (sequential streaming caused some performance degradation). We do expect that iSCSI adoption will increase, but it will amount to less than 8% of total available networked storage in 2004.

However, iSCSI is not the only IP based technology that is important to storage administrators. FC/IP will be implemented more often as remote replication and DR schemas are implemented over IP based networks. We expect a significant increase in remote replications due to lower costs and more accessible solutions, which will drive the need for FC/IP based switches. We also believe SAN consolidation over WAN will increase in 2004, driving the need for FC/IP routers/gateways. We believe the FC/IP router/gateway market will increase by 20% in 2004.

Price wars?
Prices on FC ports are coming down, there's no question about that. Late last year, Q-Logic introduced an entry level switch with a price point of $500 per 2G FC port! Brocade lowered prices to their OEMs, which in turn forced McData to do the same. And Cisco realized that even though they were Cisco, they couldn't command a 2x premium, so they brought their price per port in line with the competition. So these days we are seeing prices range from $500- $1800 per port (street), depending on the type of switch (entry level, workgroup, or director). Although we believe we will see some natural reduction in cost, we are not expecting all out price wars, except on entry level switches.

All of the vendors will have opportunity to lower prices as they move to new silicon, which can support more ports per chip. As we move to 4G (maybe) and 10G technologies, expect that 2G prices will decline. But this is what we would consider natural decline, and one vendor may be able to get there faster based on what chip vendor they work with. Q-Logic can afford their $500 price as they are their own chip supplier (and these are not director class solutions). Further, there is no imminent threat from iSCSI this year, so FC vendors are not being forced to lower prices due to a cheaper alternative cutting into their market (this will not be the case in 2005). Overall, we are expect some pricing battles in the entry level switch market, OEM's could be paying $350 per port by year end. We expect that prices for fabric/workgroup and director switches will see minimal decline in 2004.

Speeds and Feeds
Speaking of 4G, we really aren't seeing that much demand at an end user level. Many users are still migrating over to 2G, and we don't hear too much about performance issues at 2G. So what would it take to get users to upgrade their 2G systems? An upgrade to 4G that is so seamless they would never notice, both from a cost and implementation perspective. If there is any major overhaul required (software upgrades, full HBA and board upgrades, design changes) they aren't terribly interested. Overwhelmingly, users tell us they would prefer to wait for 10G FC before they make any major changes, and even then 10G will be implemented in the data center for high end applications.

So is 4G dead? Not at all, but we expect that most 4G implementations this year will be internal to arrays, where array vendors are moving toward FC switching (as opposed to FC-AL) between controllers and disks. The bottom line is 4G is more of a "gimmee" for any vendor developing FC chips. The new silicon they are using gives them the 4G speed without much additional cost, but they will have to create a market that just isn't there now. Eventually, we do believe there will be a wholesale migration from 2G to 4G, but not for years to come- when there is real demand for it.

Who will acquire whom?
Despite what so many people think, vendors do not share their acquisition plans with us. However, we do follow the market close enough to know what makes sense, and what doesn't. But even then, we don't always see acquisitions that make sense, at least when we look at some prices vendors pay. But we speculate anyway- its fun.

Rumors abound that Brocade has been or still is up for sale. So (hypothetically speaking), if that were true, there are only a handful of vendors that would be able to buy them. Large storage vendors have gotten away from owning switches, they prefer to OEM- so they are out. Two players that come to mind that could be interested are Cisco and McDATA, but both have just made sizable acquisitions (if you want to consider Andiamo a "captive" yet not yet completed acquisition), and those acquisitions have given them technology that Brocade does not have. So then, why else would they buy them- footprint of course, Brocade does have the largest installed base of SAN switches on the market. But does either Cisco or McData need that? At the price they would have to pay for Brocade, we don't think so ¿but we do think there is an opportunity for a new entrant to pick them up. Who could that be? We are seeing Nortel, Cienna, Extreme and Juniper show up at storage trade shows these days. Cienna did just buy a FC/SONET vendor- what would stop them from moving into the FC SAN market. Any of those players could follow Cisco's trend of buying their way into storage (first with NuSpeed, then Andiamo). Again, this is pure speculation¿.

What other companies look like good acquisition targets? Maranti has (hands down) the smartest intelligent services switch on the market. Other players have very strong technology (Maxxan, Troika); while they are taking advantage of 3rd party vendor services solutions, Maranti has developed all of their own. An all inclusive solution may be a more attractive acquisition candidate to a vendor looking to enter this space, but that is only true if the acquiring company has no experience in partnering in the storage market. Sandial has unique technology that no other vendor has, so it's possible that Cisco or McData could see this as a way to complete the portfolio. There are a few other players not yet announced (sorry) that have some interesting technologies that might make them good acquisition targets. We believe there will be a number of acquisitions this year, Brocade will be bought or merged with another switch vendor, and "traditional" network vendors will enter this space though acquisitions.

The Bottom Line
Used to be that storage switching was right up there with tape as far as the excitement the market generated. However, with much recent advancement in technology and the addition of intelligence into the fabric, this has become an interesting space to watch. We are particularly interested to see how the selling model changes as each vendor fights for attention from the OEMs and customers. It used to be that OEMs sold Brocade for fabric switching, McData for directors and CNT for interconnect. That has all changed, each vendor is vying for the others space, Cisco has entered the market, Q-Logic has a great new stackable solution for the low end and mid-tier markets, and a handful of newcomers are giving all of these guys a (technological) run for their money. With so many vendors offering so many good solutions, how do they get the attention of the customer?

These vendors all have good products, each has its pluses and minuses; but we don't believe this is about speeds and feeds any longer. As with many other sectors of the storage market, selling infrastructure products is no longer a valid strategy. The winners will be the vendor that can provide the best overall solution, pushing customers to see a switch purchase as a long term strategic decision as opposed to "that thing in the middle that hooks my arrays and HBA's together. The bottom line is, this year it's going to be about sales and marketing, the technology folks will continue to innovate, but it's up to the rest of the team to help users understand the value of the integrated infrastructure.

Author: Nancy Marrone- Hurley

This article was originally published on February 19, 2004