SANs for SMBs: iSCSI and virtual servers
By Dave Simpson
In an online poll of InfoStor’s readers who had deployed virtual servers, only 6% said that direct-attached storage (DAS) was the primary storage architecture behind their virtual server infrastructures. While 10% of our readers with virtual servers rely primarily on NAS, a whopping 56% rely on Fibre Channel SANs and a surprising 28% said that iSCSI-based IP SANs were their primary architecture (see figure, above).
A survey of 706 IT decision-makers, conducted by Enterprise Strategy Group (ESG) Research, reveals similar trends. For example, in the ESG survey of virtual-server users, 86% of the respondents were using networked storage, while only 14% were still relying exclusively on DAS (see figure, top right). In the ESG survey, the specific breakout (allowing for multiple responses because about half of the companies use more than one technology) was 54% with Fibre Channel SANs, 47% with DAS, 44% with NAS, and 21% with iSCSI (see figure, below), although iSCSI is on a rapid rise in virtual server environments.
As might be expected, Fibre Channel SANs are preferred by larger organizations, while DAS is often the preferred storage architecture for virtual servers among small to medium-sized businesses (SMBs). Surprisingly, however, the adoption rates for iSCSI in virtual server environments are about the same across all sizes of organizations.
Whether it’s Fibre Channel or iSCSI SANs, the most commonly cited benefits of networked storage in the context of virtual servers are: better mobility of virtual machines across physical servers (cited by 66% of the ESG survey respondents), easier and more cost-effective disaster recovery, increased uptime and availability, more efficient upgrades of physical servers, and high-availability storage of multiple copies of virtual-machine images (cited by 54% of the survey respondents). (For more information about ESG’s The Impact of Server Virtualization on Storage report, visit enterprisestrategygroup.com.)
iSCSI can be used to provide all of those benefits and, because of its additional low-cost advantages, many observers have referred to server virtualization as the “killer app” for iSCSI. For more on this, read on…
Benefits of iSCSI and virtual servers
iSCSI SANs provide a solid foundation for virtual servers.
By David Dale
Server virtualization is one of the hottest trends to emerge in the computer industry over the past few years, at both ends of the server spectrum–from mainframes to high-volume x86-based servers.
This article focuses on the volume server sector, where analysts estimate that 7% to 10% of x86 servers have already been virtualized, and the rate of adoption is likely to accelerate when Microsoft enters the market later in the year. We’ll briefly discuss the benefits of virtual server environments, describe the storage capabilities required to optimally support virtual servers, and explain how iSCSI-based SANs are providing those capabilities today.
Server proliferation problems
Judging from the rapid adoption of server virtualization, there must be a strong value proposition driving the trend. Quite simply, it’s about reducing the costs of server proliferation.
Today more than ever, companies are dependent on the constant availability of business applications and the rapidly growing amount of data they create. This drives increasingly stringent requirements for IT organizations as they struggle to maintain and improve quality of service (QoS) while operating with tight budgets, an already-lean IT staff, and limited time available to perform data-management functions.
Typical problem areas include
- Poor scalability and capacity utilization–Administrators face a frequent need to increase capacity to accommodate rapid data growth, new applications, and increased I/O requirements–hence, server proliferation. But the resulting storage utilization rate is often as low as 30%, or less;
- Physical infrastructure costs–Besides the physical hardware costs, server proliferation leads to significant floor space, power, and cooling expenses;
- Increased administrative overhead–Server and application provisioning and maintenance become increasingly onerous;
- Application availability issues–Server proliferation increases the amount of planned downtime, and makes fast recovery from unplanned downtime increasingly challenging; and
- Data availability and business continuity challenges–Mostly a consequence of direct-attached storage (DAS), timely backup and restore, and reliable and affordable disaster recovery, become increasingly difficult with server proliferation.
Server virtualization can provide relief for all of these issues.
Virtual server benefits
Wikipedia defines server virtualization as “A method of partitioning a physical server into multiple servers that each have the appearance and capabilities of running on their own dedicated machine. Each virtual server can run its own full-fledged operating system, and each server can be independently rebooted.”
In other words, you can consolidate n physical servers onto n virtual servers residing on a single machine. Typically, in the x86 server space, we typically see between eight and twelve virtual servers deployed on a single physical server–although numbers up to 30 are not unusual, particularly when consolidating many older systems onto a new, and more powerful, x86 system.
Potential benefits include the following:
- Significant reduction in the number of servers required (10:1 is common), as well as associated savings in power, cooling, floor space, and administrative overhead;
- 60% to 80% server utilization rates (up from typical rates of <10%);
- Application provisioning time measured in seconds, not days;
- Response times for change requests measured in minutes; and
- Zero-downtime hardware maintenance.
Although networked storage is not a requirement for virtual server environments (DAS works in some cases), the majority of installations do deploy networked storage to meet the data availability and business continuance requirements of their organization, and to realize the application availability and provisioning benefits enabled by their virtual server environment.
Clearly, if you consolidate the application workload and data that was formerly running on, for example, 10 separate servers, there are some implications for the supporting storage infrastructure.
First, you need very reliable storage, because a storage failure can impact many more applications. This would typically mean enterprise-class networked storage instead of internal (or direct-attached) storage.
The second issue is backup. In a consolidated virtual server environment, you are likely to have 10 or more times the backup data to somehow accommodate in the backup window of a single physical server. So, backup has to be faster. Most IT shops solve this by offloading the backup processing overhead to storage by implementing array-based disk-to-disk backup and restore (snapshots) and archiving to tape from the offline snapshots.
The third issue is disaster recovery. A consolidated virtual server environment with networked storage enables relatively simple DR options via built-in capabilities in the virtual server software and the storage devices.
Finally, to realize the rapid application provisioning capabilities of virtual server environments and complement them with rapid and flexible storage provisioning, you need the built-in features usually associated with enterprise-level networked storage arrays (see figure, above).
iSCSI for virtual servers
iSCSI-based IP SANs have become a popular storage consolidation solution for business-critical Windows environments. This popularity has been driven not only by the fact that iSCSI provides affordable SAN storage, which is particularly well-matched to environments running on x86-based servers, but also by the fact that even entry-level iSCSI disk arrays typically include sophisticated data-management capabilities such as point-in-time copy, remote copy, LUN cloning, iSCSI boot, and asynchronous mirroring. These features enable IT organizations to improve data and application availability and recoverability, while reducing infrastructure and administrative costs.
Over the past year there has been a significant uptake of iSCSI-based storage solutions in support of virtual server environments. The prerequisite, of course, was robust iSCSI support in virtual server software environments, including support for enterprise-level security options and integration with the virtual server storage administration environment. These capabilities have all emerged over the past year, making iSCSI a viable choice for consolidating servers and providing storage services for virtual server software supporting business-critical applications.
Besides the feature set and operational capabilities, an important issue in these environments is that server and application administrators can configure and maintain an iSCSI-based SAN. They all know how to configure TCP/IP ports and security, and array management such as LUN masking is typically simple.
In an iSCSI-based SAN, host systems are provisioned with iSCSI initiators. Typically, software initiators are chosen. However, with virtual server environments–particularly when a single physical server is running a significant number of virtual server instances and CPU utilization goes above 70%–iSCSI host bus adapters (HBAs) or TCP/IP offload engine (TOE) cards are recommended. Also, a separate IP network or VLAN for storage traffic is recommended, configured with redundant connections supporting host multi-pathing.
LUN provisioning is straightforward: LUNs and virtual server instances are provisioned, and LUNs are mapped to individual virtual servers. Often, a secondary remote environment is set up, and array-based mirroring is configured to support data recovery and virtual server fail-over and recovery in the case of a failure at the primary site.
A variety of tools are available for data migration to the new virtual server environment. In addition, either host- or array-based snapshots integrated with the host environment can be used for fast backup and recovery of data and/or virtual servers. In addition, LUN cloning enables rapid provisioning of virtual servers.
The resulting environment provides improved storage and server utilization, a reduction in administrative overhead, fast backup and recovery, fail-over and continuous operation in the case of a disaster, rapid recovery in the case of application failure, and fast application provisioning for test and development and software upgrades. In addition, advanced capabilities such as thin provisioning are now available on both the host virtual server and on iSCSI arrays–further enhancing utilization and recovery. Combined with the affordability and relative simplicity of the infrastructure, this makes iSCSI-based SANs an attractive option for server and storage consolidation projects using virtual server software.
The figure on p. 25 illustrates how the potential backup problem identified earlier is typically solved.
The issue is how to minimize backup downtime and accomplish greater backup processing on a fully utilized CPU. The answer is to offload it to storage.
The use of array-based snapshot technology addresses these issues by allowing administrators to effectively freeze the state of the data on disk at a particular point-in-time. Since this is not a “copy” per se–the operation is almost instantaneous, regardless of the amount of data involved–snapshots can eliminate the traditional concept of a “backup window.” Snapshots can be kept on the primary array, or copied off to secondary storage (usually a lower-cost SATA-based system) for backup consolidation or tape archiving.
Data recovery is also extremely fast, since reverting to an earlier snapshot is effectively just a mount operation. Recovery from tape becomes unnecessary, saving untold operator hours and downtime.
The figure, top, left, illustrates how the potential disaster-recovery issue identified earlier is usually addressed. The solution here, again, is to offload the problem to storage. Virtual server software does not natively provide the capability to directly copy virtual servers to a remote location, so storage-based replication for both virtual servers and data is the usual solution.
The availability of native asynchronous mirroring options on iSCSI arrays provides a flexible, affordable way to perform remote mirroring of virtual server instances and application data to other systems over an existing IP-based LAN, MAN, or WAN. Since iSCSI is a TCP/IP protocol, this solution is quick to deploy, relatively simple, and cost-effective
The figure, below, shows how the potential of fast application provisioning can be realized, again by leveraging capabilities that are supported by many iSCSI arrays.
iSCSI boot (or “boot from SAN”), in combination with the availability of array-based LUN cloning, means that operating system and application environments can be rapidly provisioned. This approach affords near-instantaneous image provisioning. In addition, the increasing availability of data de-duplication solutions keeps provisioning efficiency optimized.
Many thousands of IT organizations have been able to reap the benefits of storage consolidation in Windows application environments using iSCSI-based SANs. In addition, we are now seeing significant uptake of iSCSI SANs for server consolidation projects using virtual servers, particularly in x86 server environments where the economics, skill sets, and application workloads are an excellent match for iSCSI SANs.
By leveraging affordable IP SAN technologies and the sophisticated built-in features of iSCSI disk systems, companies can transform their server and storage infrastructure into a simple, powerful, consolidated environment that meets the increasing demands of their business applications.
David Dale is chair of the SNIA IP Storage Forum and director of industry standards at NetApp.