Riverbed Technology's Cascade network monitoring appliances, which come in physical and virtual form, have been flying a bit blind on virtual desktop infrastructure (VDI) and software-defined networks (SDN), but with the Cascade 10.0 software stack, the network busybody is being tweaked so it can peer inside of the protocols that underlying VDI and SDN to help network admins prove that whatever the performance issue might be affecting virty desktops or networks, it's the server admin's fault.
Well, unless they screwed up some network setting somewhere, of course. But moving right along. . . .
Riverbed is also fleshing out the functionality of the virtual Cascade appliance family, adding profiler and gateway functions to the existing Shark network monitoring appliance, which was virtualized in February of this year.
The problems with VDI and SDN and how they affect network monitoring tools are similar.
In a less-cloudy IT environment, you have apps in the data center and PCs on the desktop, and Cascade appliances sit on the network and can see the bits passing back and forth over the network as users do their work. Cascade can see all of the printer traffic, every keystroke, all the mouse movements, all file transfers, and all screen refreshes as corporate (and usually distributed) applications are running. When something slows down, Cascade has a means of figuring out what it is and if it has anything to do with the network itself.
But once you move to VDI, you move the PC back into the data center and you encapsulate all of that data about the applications and end users in the PCoIP protocol if you are VMware or the ICA/HDX protocol if you are Citrix Systems – and no network monitor can see what is going on with individual virty PCs.
With the Cascade 10.0 software update, the Shark appliances can peel back the PCoIP and ICA/HDX encapsulation and expose individual user data again.
With software-defined networks that are using the VXLAN protocol developed by VMware and others, you have a similar problem. The reason VXLAN was created was to enable a virtual machine partition to hop from one data center on one virtual LAN out across Layer 3 of the network to another VLAN running in another physically distinct data center. Basically, VXLAN is a Layer 2 network overlay that rides on top of the Layer 3 part of the network, making all of the VLANs across Layer 3 look like a single Layer 2.
This is all well and good, but that VXLAN data is all encapsulated in a UDP tunnel, and once again packet capture and inspection tools such as the Cascade Shark can't peer through the walls of that tunnel and see what is going on inside. Well, not until Cascade 10.0, that is.
The Cascade software update will ship to current customers on maintenance, and will be bundled with new hardware and software appliances. Now the net admins will be able to cleanly point the finger back at server and application support teams when something goes wrong.
In addition to rolling out the new X-ray features of the Cascade Shark software, Riverbed is also rolling out Gateway Virtual Edition and Profiler Virtual Edition to complement the Shark Virtual Edition that was announced back in February.
Like the virtual Shark packet analyzer, the virtual gateway and profiler tools are packaged up to run inside of a VMware ESXi virtual machine. Cascade Gateway collects flow data telemetry from network infrastructure and pipes it up to Cascade Profiler, which is a network performance monitoring tool. The virtual appliances are supported on VMware ESXi 4.1 or 5.0, and will eventually be supported on the 5.1 release that was announced in August.
Shark, Gateway, and Profiler together give you the full suite of network monitoring, and Pilot is the console you use to bring it all together to help with troubleshooting network performance and configuration issues.
In addition, the Cascade products can feed into Steelhead WAN optimization appliances from Riverbed, helping to adjust networks so applications can cope with outages and other performance issues over the wide area network. Basically, Steelhead does the optimizations, and now Cascade can make sure those optimizations are performing as expected.
Cascade 10.0 also has tweaks to peer into file systems running the Common Internet File System (CIFS) from Microsoft or the open source Server Message Block (SMB, implemented in Samba) file-sharing protocol.
"As it turns out, poor CIFS performance is so slow that it is one of the main reasons why people buy Steelhead appliances," Dimitri Vlachos, senior director or marketing and products for the Cascade business unit at Riverbed, tells El Reg. Now the Cascade appliance will be able to see individual file requests, errors, and directory browsers over CIFS/Samba file shares to help Steelhead admins adjust their WAN optimizations.
The new Cascade 10.0 software will be available sometime in the fourth quarter. Pricing for the virtual versions of Profiler and Gateway have not been set yet, according to Vlachos.
The fully virtualized Cascade stack is aimed at service providers initially, giving them the virtual appliances they need to build scalable and cloudy infrastructure – and presumably with utility-based pricing.
Vlachos said that Riverbed is also looking to sell the new appliances to commercial enterprises for their internal use, and will be looking to go downmarket to SMB shops as well. It could support other hypervisors with the virty Cascade appliances, but is making no commitments at this point. ®