HighGround Stakes Out NT Storage

HighGround Stakes Out NT Storage

Put all of the storage information you might ever need automatically on your web.

By Jack Fegreus

You can never have enough up-to-the-minute--or at least up-to-the-hour--information about the state of all on-line, fixed-media storage in a large enterprise environment. That is the philosophy behind Storage Resource Manager (SRM) from HighGround Systems.

For any size site, version 2.0 of Storage Resource Manager promises to provide both tactical and strategic information to better manage this critical resource on Windows NT servers. IT\IS Labs examined an early beta copy of the new version of this software, which HighGround Systems released last month.

On the tactical side, SRM enables a systems administrator to set soft disk quotas on storage across an entire enterprise, to better manage the consumption of disk resources. Unlike hard quotas, which lock users out of a storage subsystem when they exceed a preset quota, soft quotas simply trigger the issuance of an alert notification, which typically takes the form of an e-mail message.

The new version of SRM increases the tactical usefulness of the product. Resources can be organized into logical groups upon which thresholds can be set for conditions and events that will trigger an alert. Under version 2.0, SRM extends the range of alert handling to Win NT Event Logging, SNMP trapping, and the sending of e-mail.

With SRM, a systems administrator can now effectively manage storage on a "by exception" basis. The systems administrator simply sets event thresholds on conditions such as the introduction of a new computer, disk, or disk partition on a scheduled resource scan; the failure to find an old computer, disk, or disk partition on a scheduled resource scan; or the addition of new blocks in a SCSI disk`s table of bad blocks.

A site has a great deal of control over the scheduling of this data collection. At the top level, there is a scan of all of the computers in a domain, which can be set down to a granularity of six hours. At a much finer granularity--one hour to be exact--systems managers can separately schedule scans of disks on managed computers and scans of the logical partitions on managed computers based on membership in a named logical group of systems. In our test of the SRM beta, we created a server group for our Southborough Labs cluster dubbed "Sboro Labs."

Like omnis Gallia, Storage Resource Manager is divided into three parts: the SRM Server, SRM collection agents, and a Web Browser shortcut. HighGround has designed SRM to take maximal advantage of everything currently available within Microsoft`s DNA architecture.

The SRM Server currently uses the Jet database as a data store. A future release will add support for SQL Server. The server uses the TCP/IP and HTTP protocols, along with COM, Jscript, and Active Server Pages (ASP), to create an intranet web site to host the analysis and display of the data. Prerequisites for the installation of SRM Server are Windows NT 4.0 sp3 and IIS 4.0. HighGround recommends that the server have at least 64MB of RAM and 1GB of disk space available for database growth.

SRM Agents handle the task of data collection. These agents run on Windows NT servers and workstations under both version 4.0 and version 3.51 of Windows NT. Systems running Windows 95 or 98 are captured in the summary information of unmanaged systems collected on scans of the entire domain. Similarly, any virtual nodes created within MSCS clusters also appear as unmanaged systems captured in the full domain scan. Agents are currently tied to the physical node on which they are installed.

The presentation of all SRM data is handled through hyperlinked web pages. Active Server Pages tied together with JScript present the latest data collected by the agents and stored in the SRM repository. This greatly facilitates managerial ad hoc queries that meander along lines of thought, which only become obvious through the results of sequences of queries. This capability differentiates SRM from standard systems administration tools. Yes, SRM can be deployed as a tool to help the systems administrator, but SRM really takes on value as a strategic management tool.

By collecting a wealth of data on disk and file activity down to the logical disk partition, SRM can be used by managers as a critical planning tool. Using the pie and line graphs generated on the fly via ASP, short- and long-term trends can be easily recognized and acted upon.

An analysis of disk activity in terms of access rates can be used to balance the load on network servers. An analysis of the Sboro Labs group of servers identified several volume hotspots, including our intranet home page and SQL Server master database file. Using this information, we moved the virtual SQL Server in our cluster to a different node to minimize I/O queue depth.

We repeated that hotspot analysis, this time looking at the size of the files being accessed. Nightly incremental file backups of changed files should be part of any good backup plan. With all of that high-access activity occurring on our intranet web volume, it would not be unreasonable to anticipate a substantial nightly incremental backup load. Analysis with SRM, however, painted a very different picture. With SRM, we immediately verified that all of that activity was occurring on a tiny portion of the data residing on the disk.

SRM can be used to track long-term resource-utilization trends. SRM can display trend lines in the utilization of disk partitions that can be used to justify the purchase of new servers, robotic libraries, or new drives. Long-term budget proposals can be backed up with concrete data on resource utilization based on any number of criteria. Charge-back systems can also be validated by further dissecting this data on a per-user or per-group basis.

SRM can collect data on the physical health of SCSI disks by monitoring the grown defects list. As a SCSI disk encounters bad blocks, it replaces the bad block by remapping that physical address to a spare sector. SRM can track the number of blocks each SCSI disk remaps. Any disk that shows a growing trend in this defect list is a sure candidate for a catastrophic hard-disk failure.

This article is reprinted with permission from BackOffice magazine, a sister publication of InfoStor. For more information, or to subscribe, visit www.backofficemag.com.

In Summary

Product : Storage Resource Manager, V2 beta

Company: HighGround Systems

(800) 395-9385


Price as tested: $2,192 ($995 for SRM Server and $399 per SRM agent on a Win NT server)

Bottom line:

The new version of Storage Resource Manager should prove to be a highly-valuable analysis and planning tool for top IT management, which also doubles as a tactical administration tool. The use of hyperlinked web pages as an interface makes ad hoc query analysis a very intuitive process.

Click here to enlarge image

The presentation of all SRM data is done via a web-site paradigm. In the Managed Computers summary report, hyperlinks to computer groups or individual computers (1) enable an extended discovery process. Active Server Pages and JScript are used to provide pie and line charts (2) to analyze data on the fly.

Click here to enlarge image

FrontPage 98 can be used to examine the structure of SRM as a web site. With this web-site tool, the hyperlink structure of SRM framework is clearly visible.

Click here to enlarge image

For load balancing and capacity planning, the SRM report on file distribution by access date is invaluable for IT decision makers. The data provides both the access rate for disk hotspots (1) and the size of the files involved (2). The "E:" partition on node Kanga in the Sboro Labs group is the home of an intranet web site.

Click here to enlarge image

SRM can automatically tabulates data as a pie chart or a line chart. Here, daily file-access data is graphically displayed for all managed servers in the Sboro Labs group. The intranet web-page hotspot (1) represents a major portion of all data-access activity, but represents only a small portion (2) of data stored on servers in the group.

This article was originally published on June 01, 1998