Network performance measurement start-up Virtual Instruments (VI) has built some neat new functionality to extract extra VMware virtual machine (VM) input-output (I-O) performance metrics into release 3.0 of its flagship NetWisdom product.
ProbeVMW has been added to its network probes and analysers collection; this gathers information every 20 seconds on VM switches and APIs utilisation—and covers CPU, memory, and both LAN and SAN I-O traffic.
NetWisdom can then pass any information to VMware's vCenter, for instance to enable a VMotion trigger to move a VM application as a result of a SAN I-O performance threshold being breached. This can be used, for instance, to assist in ensuring SLAs are maintained.
"Only NetWisdom provides insight into the data traffic and flows inside the network," VI's VP of marketing Len Rosenthal told me. "VMware is blind to I-O performance."
The sorts of challenges organisations are being confronted with include sustaining compliance and SLAs which demand consistently high performance, and control of costs in storage management despite data growing up to 50% annually.
Network problems that increase costs and hinder performance can manifest in many ways. For instance, there may be FC scaling problems or lack of routing, SAN over- or under-provisioning and more general difficulties in SAN troubleshooting. Overlaying VMware can exacerbate these because it has very limited visibility into the SAN infrastructure.
Yet companies want to deploy virtualisation more widely as a way to reduce capital expense and operating costs—reducing power, cooling, floor space and (ideally) their maintenance bills.
For its part, VMware is very conscious that mission-critical applications are not yet being widely deployed on VMs and would very much like to change that. At the same time, enterprises, in particular, want to move away from direct attached to SAN-attached storage but are concerned about I-O (including for VLUNs) and lack visibility into I-O loading and characteristics.
I-O performance may have a major impact on a VM's overall throughput. Users need real confidence that they will always be able to maintain optimum performance, so this could be a key plank in VMware getting where it wants to be. As Rosenthal said: "You can't optimise what you can't measure" and pointed out that some enterprise SANs nowadays contain a large number (even thousands) of ports to SANs with edge, core and director switches.
So increasing SAN I-O visibility should ease broad VMware deployments and improve virtual to physical server ratios. Much work has been done on improving ease-of-use so that NetWisdom can more easily audit and report on traffic in real time and capture trends.
The NetWisdom Portal Server collects the key metrics, aggregating and correlating traffic data every minute. Users can, optionally, define and monitor all components—including switches, LUNs and HBAs—that make up the I-O path for a specific application. Using user-defined correlations, NetWisdom can test and report on, for instance, new VM to ESX server mappings including historical data.
NetWisdom Views provides customisable GUI dashboards enabling reports, graphs and charts, and which integrates with the VMware dashboard. A GUI can, for instance, monitor I-O for a specific VM and application.
The company is also launching a virtual infrastructure healthcheck service. NetWisdom consultants will install and use the capabilities of NetWisdom and its monitoring devices as a service to quickly identify performance problems and what it calls "behaviour anomalies."
According to VI the service will be able to point out potential as well as actual problems, for instance where performance thresholds are nearly exceeded, potential SAN infrastructure cost savings and future recommendations. Obviously, using this service this will also be a way for an organisation to evaluate the likely ongoing worth of installing NetWisdom.
Other network information gathering involves installing, where needed on the physical network, VI's existing range of devices. These include: ProbeV, a software probe that monitors SAN switch metrics via SNMP and runs on a Windows server, Xgig Analyzer which collects detailed FC traces, ProbeFCX, which extracts header information (taking measurements once a second), and Traffic Analysis Points (TAPs) devices which provide non-obtrusive FC access.
Sadly, you cannot virtualise network performance measurement but VI's devices are at least network component vendor independent.
Meanwhile, VMware may not be able to be VI-independent in its quest to bring mission-critical systems under the VM umbrella. Tools designed for a virtualised infrastructure which still maintain deep visibility into the SAN—helping to optimise SAN utilisation and speed problem resolution—are a must for a mission-critical application, not a nice to have.
VI is a brand new company (although NetWisdom has a longer pedigree); it was officially launched only in June last year having been spun off by Finisar (which retains almost a fifth share and remains a technology supplier and partner). It now has some 40 employees and, Rosenthal said: "The company is growing like crazy; it has doubled in the last three months."
In good economic times—let alone the credit crunch—that would be very impressive. As well as many major company users, of which about 50% are in Europe, it has an impressive list of vendor partnerships including the likes of EMC, HP, IBM, HDS and, especially, VMware, with whom it has a new source code and API partnership. Yet VI's first two European sales offices only opened last week (in London and Munich).
In preparing for this release, VI had a mission to enhance its tools to offer a solution for organisations to full realise the potential economic benefits of a virtualised IT infrastructure. It has now plugged a measurement gap which virtualisation created. VMware should also be happy.