Original URL: http://www.theregister.co.uk/2010/06/24/virtual_instruments_vw2/

Virtual Instruments gets hooks into SANs

From hardware to VMware

By Chris Mellor

Posted in Virtualization, 24th June 2010 09:27 GMT

Virtual Instruments (VI) reckons it has the broadest and deepest way to monitor and manage end-to-end SAN traffic, from the application in a virtual machine right down to Fibre Channel cable into an array.

It's announced VirtualWisdom 2.0 and says the product is all about optimising virtual infrastructures and enabling the Storage Area Network's (SAN) resources to be balanced with those of virtualised applications. That way there's no need for the SAN to be over-provisioned with both disk and bandwidth, VI says, and users save both operating costs (OPEX) and capital expenditure (CAPEX).

Len Rosenthal, VI's marketing VP, said: "We re-architected the whole thing, particularly around ease-of-use. Virtual Wisdom collects over 100 metrics and there is a new GUI designed to give users a customised dashboard. They can do what-if modelling on their infrastructure."

VI is the result of testing company Finisar spinning off its Fibre Channel testing operation a couple of years ago. It has grown 100 per cent annually since then, with lots of blue-chip, big corporate customers, and attracted John Thompson, the ex-Symantec CEO to come out of retirement and be first a board director and then its CEO. Why should Thompson be interested in a nuts and bolts Fibre Channel SAN component testing business?

He isn't of course; he's interested in what it's becoming, which is a bottom-to-top virtualised infrastructure testing and monitoring company. Products like SANScreen and StorageConsole have a partial view into SAN operations in that they don't have direct access into the wires; they don't have I/O transaction-level visibility. VI does because Finisar developed taps which connected into the wire and copied the traffic to a data recorder for playback and analysis.

VirtualWisdom turned this into its Traffic Access Point (TAP) product which produces an out-of-band copy of Fibre Channel traffic. Data from the TAPS goes to a SANInsight product running in ProbeFCX hardware. This can show real-time traffic latency via the Fibre Channel frame headers.

Fibre Channel SAN customers are facing huge growth rates in traffic as applications running in virtual machines (VMs) in multi-socket, multi-core servers talk through virtualised SAN infrastructures to target storage arrays which are thinly-provisioned and virtualised themselves at many levels.

An app's disk I/O has become almost frighteningly complex in terms of the software abstraction layers it passes through on its way to the wire and then further abstraction layers en route through the fabric, into the array, and through yet more layers of abstraction before it actually hits the disk blocks.

Thompson says that this complexity and the uncertainty it engenders is stopping customers putting mission-critical apps into virtualised servers because the risk of them performing slowly and being unable to fix it is so high. Provide a fix for that, means F1000 corporations could get vastly more use out of their servers and make bucketloads of CAPEX and OPEX Savings.

What VI is doing is adding upper level capabilities to Virtual Wisdom and adding ProbeV SNMP data from Fibre Channel switches to the ProbeFCX data, and then adding in ProbeVM data from VMware's vCenter for data relating to virtual machines and their apps. Data from the three sources is presented through a dashboard which is very detailed in what it can display.

ProbeVM collects ESX and guest metrics via VMware vCenterServer. It enables the correlation of server metrics with SAN metrics by ESX server and cluster, by LUN, whether shared or specific to VM, and by unique VM WWN (world wide name) if using NPIV. It's scalable to thousands of ESX instances and thousands of VMS.

You can look at traffic performance by LUN, by HBA, at a read level, at a write level and at application level. You could display charts showing a MB/sec data transmitted view for Exchange, Oracle and DAP concurrently, with such correlations being user-selectable. Users could answer “what if” questions by simulating different scenarios using historical production data recordings. They could ask questions like what happens to my read/write performance if I consolidate the LUNs on two arrays into one?

With VirtualWisdom 2.0 a single portal server can scale to 50,000 or so ports, and reporting can run concurrently with data acquisition. Think of it as a network operations centre for a SAN. VI says that the product tracks every transaction from VM to LUN, measures SAN latency per app, and can prove whether app performance problems are in the SAN or not.

Thompson said he expects VI to grow at a 100 per cent a year rate for the next three years. There are 65 to 70 people in the firm and it has operations in the USA, UK, Germany, France and Canada. VI is looking to get more traction and build its presence in those geographies with VirtualWisdom 2.0.

The key to this is the instrumentation, from taps into the Fibre Channel hardware to probes into VMware, with the dashboard being able to show and replay events, a TIVO for the SAN. The key appeal for customers is the ability to better match their SAN infrastructure to the needs of their virtualised server applications, meaning potentially lower purchases of SAN infrastructure and storage.

Thompson said most sales are direct with channel fulfilment: "The hope is we will be able to focus more on the OEM side and use a few OEMs with deep relationships. A potential issue is that we have the potential to impact long-term sales for the storage vendors."

Both Rosenthal and Thompson are emphatic that Virtual Wisdom 2 is a key product for enterprises. They have maybe 30 per cent of their servers virtualised with a focus on test and dev' environments.

Getting to have 70 per cent server virtualisation with the mission-critical apps running in VMs means they must have confidence that they can design, detect, diagnose and fix changes in the virtualised infrastructure, from app in the VM to disks in the end-point arrays. That's what Virtual Wisdom gives enterprises, they claim, and nothing else has the same depth and breadth of instrumentation and dashboard smarts. ®