The final installment of a three-part series on storage resource management takes an in-depth look at some of the leading vendors/products, based on hands-on experience.
By John Echaniz
Having examined the storage resource management (SRM) market in general, this article focuses on some of the leading products in the space and what differentiates them from their competitors. There are many products that perform portions of SRM-related functionality (e.g., Akorri’s focus on performance, Onaro’s focus on change management and root-cause analysis, etc.), but for the purposes of this article, coverage was limited to larger players in the market that offer end-to-end packaged solutions.
Of all the SRM applications available today, none is more firmly established than EMC ControlCenter (ECC). Largely due to the prevalence of EMC’s hardware, ECC has been the market-share leader since SRM tools began to pop up on the storage management landscape. It continues to lead the pack today, despite growing challenges from steadily maturing competitors.
ControlCenter was initially developed to support storage management operations only on EMC hardware; EMC concentrated its development efforts on deep support of its own gear first, with particular emphasis on Symmetrix arrays. As the market began to demand multi-vendor support, development focus shifted to broaden ECC’s management reach, while continuing to consolidate more and more advanced EMC storage management functions within the application’s GUI.
Although the majority of SRM customers do not rely on SRM GUIs for provisioning activities, ECC users are the most likely to use this functionality, because it is very sound, particularly on native EMC hardware. Fabric management features are straightforward, and most EMC-based replication functions are incorporated into the product in a manner that users-after they are trained-can handle for standard local and remote copy processes. ECC’s support of host operating systems, mainframes, Fibre Channel switches, and arrays is expansive, and the infrastructure used to run these operations scales to extremely large environments.
ECC’s reporting module, StorageScope, extracts a wealth of data from ECC that is readily available and customizable through a Web browser for pre-canned and some user-specific reporting. The data that resides within ControlCenter’s repository is also accessible through an SQL client and API, which opens the door for further data use and customization for more-advanced enterprises. Also, although not yet incorporated into the reporting interface, it is possible to set thresholds and report on key performance indicators, such as disk utilization, cache usage, and port throughput. The 6.0 release of ECC, released this summer, enhanced the look and feel of StorageScope and integrated the previously separate File Level Reporter functionality. In this product release, StorageScope will have its own Oracle-based repository for easier integration and better performance.
EMC’s acquisition of VMware and VMware’s general market success have pushed ECC’s support of VMware instances in the product’s 6.0 release. In this release, ESX servers can be fully discovered, providing views of all the virtualized guest hosts within them. While storage virtualization is not fully supported, ECC is able to discover Hitachi Data Systems’ USP arrays. However, the support for these systems is limited to basic array discovery and reporting, and because ECC is not aware that the array is in fact virtualized, there still remains the possibility of inaccurate capacity reporting if both the USP array and the virtualized arrays behind it are discovered.
ECC’s overall depth of functionality comes with a few drawbacks. To maintain its robust features as a SAN grows, ECC may require many servers to handle the agent traffic, as well as the increased number of objects comprising the storage infrastructure. Add to this the need for more servers to support heterogeneous arrays, and an ECC management infrastructure can begin to sprawl. Like most SRM tools, ECC installation is often performed through vendor services, but in larger environments, ongoing ECC management also will require full-time attention to keep the application performing well. For this reason, EMC often sells services around ECC management; ControlCenter is most beneficial to a customer when it is managed by someone with significant product expertise.
While the product’s heterogeneous reach is improving, multi-vendor disk array, NAS, and tape device support continue to be ECC’s primary weaknesses. Specifically, support for Network Appliance’s systems is a significant blind spot, especially given NetApp’s primacy in the NAS space. HDS and IBM array support is limited to little more than basic discovery and operation with the current release. With the closer adherence to SMI-S standardization in ECC 6.0 (and with the maturation of SMI-S itself), discovery and support of arrays from vendors such as IBM and HDS will improve gradually. That said, for shops where EMC hardware is prevalent, ControlCenter remains a solid choice for storage management.
HDS HiCommand Storage Services Manager
To address its own SRM functionality needs, Hitachi Data Systems began re-branding and selling the AppIQ StorageAuthority Suite in 2004. With Hewlett-Packard’s acquisition of AppIQ in 2005, this software package became the standard SRM tool for HDS storage environments. It is now offered through HDS (HiCommand Storage Services Manager, or HSSM), HP (Storage Essentials), and Sun (Sun/StorageTek Operations Manager), and also through vendors such as Engenio and SGI. Each vendor has its own packaging nuances, but the underlying product is the same.
Because of its independent origins, HSSM was better prepared for multi-vendor storage environments than some of its competitors. Array support began with the most prevalent vendors and models and expanded to complement App-IQ OEM partners. The software’s host, switch, and array support base is substantial, and on the whole, HSSM is not difficult to configure. In particular, HSSM’s support for NetApp devices is excellent; although users cannot actively manage NetApp filers from HSSM, the console provides extensive configuration and utilization data, down to minute details.
HSSM’s all-in-one server setup makes for a relatively painless configuration experience. The console is run entirely through a browser, so no client install is required across a distributed user base.
The facet of HSSM most appreciated by its users is the GUI. Well-organized and intuitive, the user interface is logically assembled and is able to handle large amounts of data in a manner that is easy to traverse. Specifically, the product’s topology views are very useful in tracking component connectivity without overwhelming the administrator.
HSSM also provides optional features for managing and monitoring specific applications, such as Microsoft Exchange and a variety of databases. Users can gain valuable insight into the performance characteristics of these applications and the components behind them. This can be accomplished through additional system configuration, without additional agents.
Despite these advantages, HSSM has some challenges to overcome. Primary among these is the scaling limitation presented by the all-in-one server infrastructure, which inherently limits the breadth that an installation can accommodate. A related concern is the time it takes the software to perform element discovery across a sizable installed base. Many operations are dependent on this discovery process, so such delays can affect users if the discovery process runs into hours of business operation.
The data made available by full discoveries is quite useful, but the host-discovery process can be onerous in some environments. For example, for full host discovery, administrators must install the CIM extension on each host and supply the host’s IP address, user name, and password to the HSSM server. On top of that, Windows hosts require an additional proxy system. Some users have opted to forego complete host discovery because of this process.
HSSM’s reporting component provides solid storage environment statistics, from high-level usage summaries to detailed component configurations. These reporting functions, however, require a regular data refresh, which can often be time- and processor-intensive.
The reporting component, especially the custom reporting capability, is the product’s one interface component that is not very easy to use and navigate.
Although it is the SRM product of choice for HDS storage environments, there are notable functional omissions in HSSM regarding HDS hardware. For example, HDS’s local and remote replication operations (ShadowImage and TrueCopy) are not available from the HSSM GUI. Also, HSSM does not fully support HDS’s array-based storage virtualization solution; it is not currently able to handle the discovery of a TagmaStore array and the arrays it virtualizes without double-counting the storage. Tighter integration with HDS’s storage virtualization functionality would help both solutions (HSSM and Tagma-Store) achieve greater market adoption. As it stands, HSSM does provide a very useful and functional tool to view, manage, and report on a storage network from end to end, and further market adoption will help HDS drive enhancements to meet customer needs.
IBM TotalStorage Productivity Center
IBM’s TotalStorage Productivity Center (TPC) is a modular approach to SRM consisting of TPC for Disk, Fabric, Data, and Replication, each of which was a separate Tivoli product in prior generations of the software. These elements have been integrated into a consolidated DB2 database and GUI front-end. Like HSSM, TPC relies on the SMI-S standard for heterogeneous device support and communication and as such provides a broad range of array, switch, and host coverage.
Numerous software vendors have demonstrated that field use will drive product improvements. If a growing customer base uses a product actively, the product is more likely to improve through feature requests and general feedback. IBM has wisely established TPC for Disk as a replacement for ESS Expert-the tool previously used to monitor performance on IBM’s older Shark arrays. This decision alone caused a significant increase in TPC usage. IBM is now also including the “limited” version of TPC with sales of its DS line of disk arrays. With TPC components already installed in the data center, it becomes considerably easier for users to explore other components of the product and ultimately increase their usage of it.
TPC is geared toward IBM hardware, while providing basic functionality for other vendors’ gear, which is to be expected of products that rely on the SMI-S standard. TPC for Replication provides extensive support for IBM’s local and remote replication operations (FlashCopy and Metro Mirror). While these capabilities do not yet extend to non-IBM arrays, the functionality offered through the GUI is deep and comprehensive. TPC’s support of NetApp hardware (which IBM re-brands) is also very good, providing useful insight into filer usage and quota levels, and even file-level data.
Users have found the TPC GUI intuitive and easy to use. Although it is not as slick as some of its competitors, it is relatively simple to use; most administrators never crack open the users’ manual beyond initial install because the tree-view layout is conducive to exploration. The topology viewer comes with a handy method for progressive drilldown, enabling users to “pin” certain levels of display while maintaining their overall context. In this way, the display can handle numerous components without becoming overwhelming to the user.
Another notable distinction for TPC is its support for IBM’s storage virtualization solution, the SAN Volume Controller (SVC). Although it only supports the IBM virtualization solution at this time, TPC is the only SRM tool in the market that can manage, monitor, and report on storage virtualization, with visibility into the SVC’s inner workings. TPC is able to monitor traffic through the SVC’s nodes, which was previously a significant blind spot for performance troubleshooting in a virtualized storage environment. Going forward, IBM will do well to provide greater visibility into virtual-disk to managed-disk relationships to facilitate more-extensive troubleshooting. TPC’s summer release (version 3.3) added change management functionality to the product in the form of wizards that help users plan and execute configuration changes and track their history. It also introduced functionality for analyzing SAN configurations against customizable best-practice policies. With these advances, TPC seems to be extending its reach into the change management space where vendors such as Onaro have found success.
Like its competitors’ products, TPC’s products also have some shortcomings. The product ships with literally hundreds of canned reports that cover a wide spectrum of information about the storage enterprise. That said, advanced users who wish to extract and manipulate the reporting data might have a difficult time digging into the underlying data; the database is open, but not clearly documented, which inhibits its use.
TPC agents introduce some drawbacks as well. These agents can be pushed remotely, but not “silently;” manual interaction is required for each remote agent installation. This is better than needing to make a physical visit to each server, but the ability to automate the agent push would make the overall agent rollout process simpler. TPC agents require a few more firewall ports to be opened than its competitors, which can serve as a significant impediment in environments with stricter firewall rules.
Finally, the fabric agents required by TPC for Fabric require undue proliferation. For proper switch management through TPC, a fabric agent must run on at least one host connected directly to each switch the user wants to manage, and basic redundancy rules call for two agents deployed per switch. Improving upon this condition will encourage more users to adopt TPC for Fabric as the switch management tool of choice, but until then users will continue to rely on native switch management tools.
The user community for the full TPC suite is still fairly limited in size, despite the product’s growing popularity. TPC for Disk is by far the most commonly used part of the suite, because it is the most immediately useful component to storage managers. TPC for Fabric, Data, and Replication remain relatively lightly used in comparison, but as storage managers become more familiar with the product’s benefits, the TPC suite will become more widely adopted.
Symantec CommandCentral Storage
Compared to its primary competitors, Symantec is at an inherent disadvantage when it is jockeying for position in the SRM space. Where other vendors have bundling opportunities for their storage management tools in hardware deals, Symantec must rely primarily on the comparative benefits of its SRM tool, CommandCentral Storage (CCS). Customers often use array purchases as negotiating leverage when they are considering array management software, and vendor-supplied SRM tools are usually more feature-rich when managing their vendors’ own hardware. These two market factors have worked against Symantec, but the CCS development team has been able to differentiate its product in an increasingly crowded market.
CommandCentral Storage uses a blend of SMI-S and vendor APIs to achieve broad component support while providing functional depth across a wide variety of hardware. For example, CCS uses the EMC Clariion API to provide advanced LUN binding functionality on Clariion arrays, where some of the other products are more functionally limited. Even in environments where CCS is used daily, it is not widely used for storage provisioning; that said, the product’s designers planned for it well by taking the hybrid approach to multi-vendor support.
One significant strength of CCS is its consolidated use of agents. Where other SRM products have a wide variety of agents used to manage different types of components in a storage infrastructure, CCS relies on one installable agent that is configurable to manage a variety of different components. The same agent- install package is used to manage hosts, arrays, switches, and specific applications throughout an environment. This makes environment tuning considerably easier: Administrators can configure different collection components (i.e., “explorers”) as an environment grows, without the need to install additional packages.
CCS has also differentiated itself through the ability to push agents to servers across an environment. Especially in shops where servers have a consistent maintenance login and password, CCS’s Agent Push Utility can load agents on multiple boxes through batch jobs that require no manual intervention. Symantec could further extend its advantage in this area by building more automation around the prerequisite-checking process. Host qualification still takes considerable time and effort, which slows down the overall deployment process significantly.
Database visibility within CCS is particularly helpful for storage managers and DBAs alike. Once the base agent is configured to collect database information, CCS provides a consolidated view of databases across a data center, including version information and tablespace utilization. CCS is able to display the entire data path, which is very helpful when troubleshooting.
CCS has also taken advantage of Symantec’s market share in the clustering and server volume management space. The software has excellent visibility into cluster utilization-another significant advantage in the SRM market-and provides similar insight into Veritas Volume Manager and File System configurations. With the prevalence of the Symantec Storage Foundation suite in the Unix world, perhaps Symantec will seize on this position in the future by bundling CCS agents with Storage Foundation. Such product convergence could help CCS gain considerable market share by removing one of the primary impediments to SRM adoption.
Symantec has incorporated the acquisition of Invio into the CCS mix in the form of Veritas Process Automation Manager, a workflow engine intended to drive collaborative tasks such as storage provisioning through a standardized, auditable process. It is a services-intensive effort to design and implement and has not yet been used widely; that said, this module has the potential to bring some order to chaotic storage environments, so it will be worth tracking how the SRM market reacts to this supplemental CCS offering going forward.
As mentioned previously, Symantec has some disadvantages when it is competing with SRM products from storage hardware vendors. These disadvantages have kept the CCS market share relatively small compared to SRM products from vendors such as IBM, HP, HDS, and EMC. Because of this smaller customer base, there have been fewer opportunities for CCS to be really challenged with field use, which means its general product stability may lag behind the other products. SRM tools interact with such diverse and numerous components that vendors cannot hope to test every combination of component, driver, and firmware on any significant scale. Constant field feedback is absolutely critical for SRM tool success.
CCS has experienced some performance issues in larger environments. In shops with server counts in the thousands, the managed object count may grow to a point where CCS’s GUI response times lag significantly. When this occurs, CCS may encounter component responsiveness problems, which will necessitate additional manual intervention to keep environment information current. How- ever, this summer’s release of CCS 5.0 provides a number of improvements in this area; for example, the new server architecture improved the product’s database usage, decreasing its reliance on memory. The 5.0 release also includes tunable element discovery, which will allow “delta” processing as opposed to full data pulls, which in turn, will make environment discovery more-efficient. These enhancements should alleviate many, if not all, of the performance challenges in the 4.3 version of CCS.
CCS has also had some challenges with reporting. One method for mitigating performance problems in larger environments is to break out environment management into multiple CCS servers. Global rollup of CCS data was intended to be accomplished by its Data Rollup Utility (DRU), but this process was originally intended for backup reporting and experienced scaling and stability problems on the storage side.
The 5.0 release will abandon the DRU process in favor of Enterprise Reporter, a Cognos-based reporting system that will allow for easier report customization, and global rollup of CCS instances for reporting purposes.
Canned reports in CCS are fairly limited, but useful. Again, in larger shops, data growth in the 4.3 version can become too much for CCS to handle, especially when running intensive reports such as trended utilization. CCS also comes with a separately installable Data Module, which can be used for file-level reporting. This produces valuable insight into aging files, and multiple file copies, but it must be used sparingly because of the data growth it can produce. The aforementioned enhancements in 5.0 will also enhance the reporting performance through improved use of the CCS database.
On the whole, Symantec CommandCentral Storage has many advantages, although to date it has suffered from lack of widespread use. However, as the product’s user base grows, and as Symantec establishes tighter integration with its other products, CCS will be in a position to make significant leaps forward in the SRM space.
Over the past several years, SRM tools have matured steadily and have seen more widespread use and adoption.They are increasingly considered to be a standard part of major storage sales and are being used as key points of data centralization and management. Ultimately, the value of SRM products must lie in IT budget savings; the more these tools allow IT shops to do more with less-to manage, report, plan, and troubleshoot effectively-the more these products will be integrated into standard operations.
For SRM to truly fulfill its industry promise, these tools must build on the progress they’ve made over the past few years. Among the more pressing challenges, they must incorporate deeper, more-meaningful cross-vendor support and seamless global aggregation, and they must improve in their ability to automate storage operations if they will ever gain acceptance by veteran storage administrators. Beyond these challenges, as SRM matures further to conquer archive management, backup integration, and data deduplication management- ultimately delving into the data itself-SRM will truly be in a position to provide that long-promised single viewpoint through which the storage world can be managed.
Parts 1 and 2 of this series of articles appeared in June 2007 and July 2007, on pages 31 and 33, respectively.
John Echaniz is director of client solutions at the NovusCG (www.novuscg.com). NovusCG’s Justin Schnauder, technologist, and David Askew, client technology executive, also contributed to this article.
Representative SRM vendors
- Hitachi Data Systems
- Network Appliance
- NTP Software