Investing in diagnostic tools for networked storage has several benefits. With the proper tools, you can easily shorten the time it takes to get to the root of performance problems in a complex storage infrastructure. Moreover, even when nothing is broken, a good diagnostic tool can provide a better understanding of your SAN’s inner workings, knowledge that can help you improve the performance and resilience of your storage networks over the long haul.
Late last year, Finisar released Xgig Analyzer 1.5, a new version of its popular SAN analysis tool. Xgig 1.5 maintains the capability of previous versions to do flexible and accurate diagnosis on Fibre Channel (FC) networks, and adds similar features for iSCSI and Fibre Channel over IP (FCIP) protocols, extending its support to the quickly growing area of IP-based storage.
Finisar offers different models of Xgig tailored to different analysis requirements. They range from a portable single-blade model to a rack-mountable four-blade version, all sharing the same analysis software. Finisar sent me a desktop unit which I’ve been using in my tests of storage networking equipment. It’s about the size of a shoebox, and includes two blades for analysing FC and GbE networks, each hosting four 2Gbps FC ports.
Tap and trace
The first step to setting up the Xgig is to invoke the Xgig character-based management application via serial connection and set an IP address.
After that you use a Web-based management tool to configure other settings, and to update the Xgig’s software via FTP if necessary.
The Xgig software works in synergy with a suite of applications that includes Expert Viewer, Performance Monitor, TraceControl, and TraceView. Their self-descriptive names give away their functions, but I should clarify that they run on a separate Windows console and interact with the Xgig to capture and analyse traces or monitor performance.
The Xgig architecture gives you great flexibility. You can position a unit to monitor a single device or plug it behind a switch to monitor the traffic from multiple devices. Or you can position several Xgig blades at different locations in your SAN and monitor them from the same workstation.
Moreover, your applications won’t miss a beat while you monitor or analyse their traffic; the Xgig is completely transparent to the storage network.
For this review, my goal was to analyse the traffic flowing to and from an iSCSI array in my test network, namely the Overland Storage RA2000 I reviewed in February. After connecting the Xgig to my SAN then back to the machine hosting Xgig’s software suite, I launched the main application, TraceControl.
It auto-discovers Xgigs on the LAN and remembers findings in future runs. The first time you use it, you simply type the IP address of the Xgig, press the “discover” button, and TraceControl opens a window showing the Xgig and the ports available on each blade.
Filters and frames
TraceControl doesn’t analyse data, but collects information for later analysis with TraceView or Expert Viewer. However, you can program Trace-Control to collect only selected data, filtering according to a protocol, to specific frames, or just for errors, which is extremely helpful because it narrows the capture to the specific information you need.
With a few mouse clicks, I captured only iSCSI frames sent on port 3260 from an initiator, building that filter from a comprehensive template that covers the most common protocols. For frequently used and complex filters, you can also create custom templates by dragging and dropping entries from the master template.
After you save the trace, you close TraceControl and launch TraceView. You’ll typically use TraceView to analyse a file saved with TraceControl, but it can also analyse data straight from the tap, which is useful to chase errors interactively.
It’s impossible to mention the numerous features of TraceView in the space allotted here. One that captured my attention is a narrow bar chart at the bottom of the screen that provides an immediate view of the trace file, listing peaks such as a surge of similar frames. That chart is very useful for spotting possible problem areas.
Clicking on a peak brings the details into the main pane for further analysis.
To isolate a problem, you can define additional filters using the same comprehensive template seen in TraceControl. You can also filter only frames related to the same port. Moreover, TraceView can automatically color-code entries to mark different protocols or frames.
To facilitate the analysis, TraceView opens a separate pane for details such as source, destination, length, and time of each packet, as well as additional and more detailed information, down to the bit content. Those details are a useful complement to the summary frame data listed in the main pane.
I liked TraceView. If I had to spend many hours analysing traces, TraceView would be one of the first choices for my toolbox. However, if I had to spend many hours analysing traces, using Expert Viewer instead would probably cut that time to minutes.
Using Expert Viewer is like having a skilled co-worker point out errors in a trace. In fact, in contrast to TraceView, Expert tries to make sense of all frames in the trace, verifying not only that each frame is correctly formatted, but also that their behaviour is consistent within the context of the stream.
Expert Viewer did not detect any error during my regular tests, indicating that the dialogue between the software initiators on my server and the Overland Storage was spotless. However, I triggered quite a few false positives working with badly filtered traffic. For best results, Expert Viewer should have access to the full trace.
After using Xgig for a few weeks, mostly positiveremarks populate my test log. Setup can be challenging, but the reward is much needed flexibility and flawless operations.
A last minute check with Finisar silenced one of my few criticisms: Support for 4Gbps FC should be available by the time you read this. If your SAN is getting complicated, you should make room in your budget for Finisar’s Xgig Analyzer. It’s not cheap, but it’s an investment that should quickly produce a sizeable return.
Matrium’s sales and technical manager, Dominic Fitzgibbon, said the product required technical know-how in the IP-based storage arena.
The Xgig 1.5 software application sat on the Xgig platform, and the price ranged from $45,000 to $250,000, depending on the interfaces and chassis included, he said.
When pitching the top three features, the reseller needed to look at line rate traffic (or wire rate traffic).
“If there’s an aging link — you can capture a gig worth of traffic and therr’s no lost packets,” Fitzgibbon said.
Another top selling feature was a flexible platform, and the ability to mix Fibre channel and Ethernet modules, he said.
The technology also covers all speeds.
Fitzgibbon said version 2.0 has just been launched. Enhancements include: new multi-blade configuration on Xgig platform; new scalable multi-chassis configurations (2-64 ports); and the latest feature-rich GUI with updated status view, performance monitor and trace view.
The product is distributed in Australia by Baulkham Hills, NSW-based Matrium Technologies. (www.matrium.com.au)