With a growing NAS customer base and major upgrades to Windows Storage Server on the way, Microsoft is serious about becoming a major player in storage. But can it secure a place in your data center?
IT'S CERTAINLY NOT NEWS THAT MICROSOFT CORP. wants to become a big player in storage. But can the Redmond, WA, behemoth go toe to toe with companies such as EMC Corp., Hewlett-Packard (HP) Co., IBM Corp., Network Appliance (NetApp) Inc. and Symantec Corp.? Because Microsoft has established hegemony in other IT markets, it's prudent to scrutinize its storage roadmap to determine what a mature Microsoft storage strategy may look like and what technologies, partnerships and strategies it will leverage along the way. While its success is far from assured, if Microsoft can extend its dominance to storage, end users will undoubtedly feel the effects.
Microsoft storage technologies are broadly improving, albeit with varying rates of maturity. Looking past the market hype, it's apparent that there are four major areas where Microsoft will have an enduring impact on the storage industry:
- File serving and NAS
- Data protection
- Remote-office support
- SAN and storage management
|Microsoft Storage Timeline|
|Click here for a look at Microsoft's Storage Timeline (PDF).|
Storage professionals should evaluate Microsoft's progress along these distinct functional lines, rather than try to make sense of the numerous individual component technologies the firm announces with its platforms. By tracking Microsoft's strategies in these four areas, storage managers will be better equipped to make solid decisions about when and where to consider Microsoft products for their storage environments.
|What's new in R2?|
The R2 release of Microsoft Windows Server and Windows Storage Server includes several critical additions that expand storage-related capabilities.
File serving and NAS
The backbone of Microsoft's storage strategy has been file serving on Windows Server, which has evolved into a mature NAS offering with Windows Storage Server (WSS) 2003. In April, Microsoft plans to ship the next version, Windows Storage Server 2003 R2. Available only through designated OEMs (now numbering more than 50), WSS has quickly become the NAS platform of choice for low-end enterprise file serving. Microsoft claims 54% of the NAS market, based on total unit shipments. It should be noted that Microsoft-based NAS is predominately in the lower end of the NAS market. WSS is cheap to acquire and deploy, receives consistently good reliability marks from users and has easy-to-use management tools. WSS demonstrates that Microsoft can "do storage."
However, since WSS emerged in 2003, enterprise storage professionals have been wrestling with how many "heavy-duty" storage applications they could safely move to the platform. To name just a few of its shortcomings, NFS performance was abysmal, Unix clients were treated as second-class citizens and WSS lacked built-in quota management. While WSS has a good track record in small- to medium-sized businesses and departmental environments, its capabilities as an enterprise-class NAS offering have remained largely untested. Major OEMs like HP have made hay with the WSS platform, but in the $50,000 and lower price category. However, key improvements in the upcoming R2 release may broaden WSS' appeal.
The platform will be faster. Microsoft hasn't made much noise about it, but it has allowed its OEMs to optimize the registry settings in WSS for NAS workloads. The result is a 10% to 25% file-serving performance gain over the standard Windows Server platform, depending on the workload. To provide some context, Windows Server 2003 R2 on an HP ProLiant 585 scored 4.1Gb/sec, as tested by NetBench and using its benchmark.
R2 will also significantly improve NFS performance, delivering 22,000 IOPS on a single node vs. the 10,000 IOPS to 12,000 IOPS previously delivered (SPECsfs testing on an HP ProLiant 585). Microsoft also says it will support multinode clustering on NFS in 2006, with up to eight nodes and linear scalability. This level of NFS performance would be appropriate for many heterogeneous NAS environments up to the enterprise midrange.
Other WSS R2 enhancements include single-instance storage at the file level, and full-content index searching of the data store. On the file-server management front, the File Server Resource Manager (FSRM) also brings some much needed file-level controls regarding quota management, capacity reports by group, file screening and duplication controls. While FSRM in R2 is solid, it doesn't support sophisticated policy automation like competing products from Acopia Networks, NeoPath Networks Inc. or NuView Inc. However, as a feature of a server platform, it's impressive.
Despite the R2 release improvements, it's still questionable if Microsoft NAS can compete with the likes of NetApp or EMC. The issues are performance and throughout. With R2, Microsoft is taking on established high-performance NAS players for the first time. It's doing this by coupling Windows Server 2003 R2 with the Windows version of PolyServe Inc.'s Matrix Server, a cluster file system (CFS). The combined solution creates the first symmetric, scalable CFS-based NAS cluster based entirely on a Microsoft platform, which can compete effectively on a price/ performance basis with high-end NAS.
Given these advances, 2006 should be a good year for Microsoft in the file serving and NAS markets. Storage professionals should evaluate Microsoft-based systems for more demanding data center tasks. Consider the R2 release as a harbinger of things to come from Microsoft in the world of enterprise NAS.
Microsoft released System Center Data Protection Manager (DPM) software last year, a disk-based software recovery application. As of this writing, there have been more than 200,000 beta downloads and trials. While groundbreaking for its ease of use and bringing disk-based recovery to small file-serving environments, the current DPM release has its limitations (see "Pros and cons of Data Protection Manager, Version 1.0."). However, priced at $995 per license for three servers ($150 per additional host), Microsoft is changing users' expectations about what advanced data protection should cost.
|Pros and cons of Data Protection Manager, Version 1.0|
DPM 1.0 has few pretensions of grandeur. The easy-to-manage, disk-based backup application resides on a dedicated server running Windows Server 2003, and supports NAS and SAN storage. DPM requires deploying agents on each of the file servers it protects. It uses Active Directory for discovery and wizard-driven installation, as well as for its security controls. DPM creates recovery objects through a combination of byte-level replication and Microsoft's snapshot technology, Volume Shadow Copy Services (VSS). It can create up to eight user-selected recovery images a day (with a limit of 64 per volume).
DPM isn't a "continuous" any-point-in-time recovery product. The logs DPM uses to synchronize data changes against the VSS replica are created once an hour, so there's a potential for some data loss; however, recovery efficiency is improved by the checkpoint capabilities that automate recovery to the last viable replica. API integration is supported with traditional tape backup applications from all major vendors. When combined with a tape solution for long-term archival, DPM is a low-cost, data-protection product for a small- or medium-sized Windows file-server deployment.
Expect DPM's playing field to broaden as Microsoft says future releases (V2.0 is expected by the end of this year) will support direct application integration with SQL Server, Exchange, SharePoint Server and the 2007 Longhorn Server release. Other planned improvements include bare-metal recovery, high availability, SAN integration and a centralized management interface. Within two years, DPM might begin to look much more like a true enterprise-caliber data-protection offering--albeit, Windows only--that could compete with the likes of EMC, IBM, Symantec and the range of startups that comprise today's midmarket.
With regards to data protection, Microsoft is using its massive installed base to drive familiarity with a disruptively priced offering in the hope that users will ante up for the real deal in V2.0. Microsoft has been successful with similar tactics before, so it's safe to assume it will emerge as a formidable competitor in data-protection software within three to four years.
Remote- and branch-office IT issues are top of mind at many companies, often accounting for 30% or more of a company's entire IT budget. Storage-related efforts are key factors, including data set consolidation across multiple sites, data replication over the WAN and remote-office file service delivery. Microsoft has talked a lot about branch-office support, but has released few products that address the issue. Past releases of Windows Server have fallen short of providing the features that would make it a truly viable platform for managing remote offices, including inadequate replication capabilities and data compression over the WAN. Most of these shortcomings have been addressed in the R2 release, with a range of new features that make Windows a more attractive platform to the distributed enterprise.
Most notably, the Windows Server 2003 R2 release includes substantial improvements to the Microsoft Distributed File System (DFS), which will enable it to provide centralized namespace management across file servers in distributed geographies. It can support user-defined policies to publish, collect or replicate content across approximately 1,000 connections per implementation. The namespace management features are also greatly improved, with controls for server/site failback, closest site referrals, administration delegation, bandwidth throttling and targeting priorities. IT managers who have grappled with a multisite Windows Server 2000 or 2003 DFS implementation will welcome the new release; it simply looks and feels much more like an enterprise-class distributed content management tool.
Many users will be overjoyed to discover that Microsoft has replaced its nearly unworkable File Replication Service engine with a new DFS Replication Engine, which optimizes Windows data transfers between servers and their clients over the WAN for the first time. The new DFS Replication Engine uses compression algorithms to reduce the size of data transfers, making replication feasible for enterprise workloads.
Microsoft is also partnering with Tacit Networks Inc., a Windows-based provider of WAFS technology. WAFS products have become increasingly attractive because they enable IT teams to remove storage from remote offices and replace it with remote gateways that use the WAN to link back to centralized data stores.
Microsoft's intentions to be a player in the remote-office arena are clear, but it faces significant challenges from established players like Cisco Systems Inc., IBM and Novell Inc. By making the Windows platform friendly to IT managers, Microsoft achieves its main objective of protecting the platform from competitive encroachment. From a user perspective, this means more built-in management flexibility for distributed Windows deployments and the ability to blend remote-office approaches.
SAN and storage management
Microsoft is dead serious about enterprise SAN, although it may not seem so today. While never explicitly stated, Microsoft appears to realize that it must ultimately control the trajectory of the entry-level SAN and IP SAN markets. Losing control of the storage data path--from creation to archiving--could threaten the dominance of the entire Windows platform.
One of the most strategic moves Microsoft made was to bake-in the Storage Manager for SANs software with the Windows Server 2003 R2 release. The functionality is still basic by enterprise standards, as it focuses on the immediate issues of Windows-based SAN provisioning, configuration and deployment. Still, it works like a Windows product and, most importantly, it repositions SAN management as an expected feature of the server platform. Storage Manager for SANs uses Microsoft's Virtual Disk Service via an API to support Fibre Channel, iSCSI or DAS. The software enables autodiscovery of storage objects, wizard-based provisioning and iSCSI security configuration. But limitations abound. There's no hint from Microsoft of SMI-S/CIM support, it has limited heterogeneous host support (host bus adapter discovery of Unix/Linux hosts only) and no support for array-based replication. Thus, Storage Manager for SANs will only be useful for entry-level SAN and small business IP SAN users. Of course, that's precisely the market Microsoft is targeting with its storage management energies. Because the majority of all SANs have yet to be assembled, this is a smart move.
Additional SAN news will begin rolling out of Redmond throughout 2006; expect iSCSI remote boot support in the spring, as well as distribution partner announcements with IP SAN vendors such as EqualLogic Inc., LeftHand Networks Inc. and others. And it's not far-fetched to expect to see Microsoft extend its focus to include iSCSI targets in 2007.
Ultimately, it won't be until 2009 that a mature Microsoft SAN and storage management threat could encroach on enterprise-class deployments. Among the factors that will determine if Microsoft will find success with larger enterprises are its willingness to embrace heterogeneous data center storage infrastructures, and an ability to get above the Windows Server platform for advanced storage resource management.