PlateSpin and LeoStream step up with stable products that ease the migration
Conversion applications move a working host into its new virtual environment along with its stored application images in tow.
Unless VM guests are installed from ‘bare metal’ operating systems and application media, guest instances are converted from an existing stable VM that has a specific foot print that regards its memory needs, disk displacement plus future storage needs, network I/O requirements and settings, and needs for raw CPU power.
Converting that stable platform requires preserving many characteristics, including operating system-specific files, settings, and user and group attributes. On top of that, the servers must maintain access to external resources, such as DNS servers, certificate authorities, directory services, external databases, and the links that are components/resources to a working server.
Keeping all of these components in tact with causing as little server downtime as possible is a tall order that both PlateSpin and LeoStream take pretty successful stabs at it.
PlateSpin PowerConvert
PlateSpin’s PowerConvert aims to be an egalitarian provisioning product that seemingly doesn’t care which virtual host platform you are working with. Its main goal is to help move a server process from point A to point B. In PlateSpin’s realm, these points could exist in either the virtual or physical worlds, or have a foot in both. PlateSpin supports P2V, V2V, V2P, P2P conversions, P2I (physical to image), V2I, I2V and I2P deployment.
The key to PlateSpin’s proposition is that it uses a procedure of conversion that’s designed to provide at most, a single re-boot. This is obviously a beneficial trait in any conversion product because it minimizes server downtime. Likewise PlateSpin’s product assists in hot failover and disaster recovery scenarios as essential services and other mission-critical systems can be mirrored in live or standby modes on physical or virtual backup resources. Of course, the one reboot claim comes with the caveat that you’d likely have to run several test conversions before attempting one on a live, mission-critical server.
PowerConvert analyzes candidate servers to take an inventory code and correlate services running on them. Once the systems are fully understood, the task of dragging a working system that may consist of just a few files and service profiles, or very many can proceed in various ways. The server’s workload can be consolidated onto a larger server, converted in its entirety to a candidate virtual or physical resource that can serve as its replacement or run as a standby backup or archived into a image repository for backup or other purposes.
Installation of the PowerConvert server components requires the Microsoft .Net 2.0 framework running on the Windows 2003 Enterprise Server platform, which isn’t included in the distribution. We had Windows 2003 running on a VM in our lab, but had to manually upgrade the system with thee required .Net components. We used both a CentOS 5 and Windows 2003 Enterprise Server to produce image archives, gold masters or build-groups of server/client operating systems that were typical configurations for running production servers, which we wanted to convert with the PlateSpin product.
We tested PowerConvert on both Virtual Iron and VMware virtualization platforms. It analyzed software running on both platforms quite well and understood the small differences in formatting between the two platforms perfectly.
The conversion process, whether scheduled or live, proceeds through a detailed step-by-step checklist that helps to pinpoint exactly where any conversion may have failed. If, during the conversion processes, there is a failure at any step, a helpful error code is registered. PlateSpin provides a searchable knowledge base of possible errors that can occur during any conversion and provides guidelines to remediate the issue.
We used PowerConvert to run the gamut of migrations, checking P2V (Windows and Linux) as well as V2V, I2V and V2I images of running platforms on the target platform. Processes included in these images are DNS, file/print services, networking configurations, and in the case of Linux, sendmail/procmail. The only glitch we encountered was occasionally, and not repeatedly, PowerConvert would be unable to stop services on Windows servers during P2V conversions.
However, simply fixing the error and restarting the process where it left off never worked for us; instead, we had to restart any failed conversion from the beginning, a measure that resulted in a complete conversion. An administrator needs to monitor and shepherd the process closely as SMTP alerts are the only method of knowing something’s gone wrong in an unattended conversion. An upside is that the conversion(s) can be scheduled to run during quiet network time if the conversion process has been stabilized.
PowerConvert was very fast in its ability to take our test P2V and V2V conversions, especially on Linux. Taking a running physical server and converting it to a VM the first time is fast; rejuvenating from an image is very fast — as fast as the server can boot (of course, as it’s a native image). It’s a far faster process than a manual program load, plus an operating system load plus configuration. Our Linux CentOS load with a running Apache server with Tomcat support took 52 minutes on a P2V conversion and a V2V transfer took 9 minutes. The Windows 2003 Enterprise Server conversion took longer, 94 minutes for a P2V, and 19 minutes for V2V.
The PowerConvert application server, does limited administration — just things like stopping and starting services both pre- and post- conversion. There is, unfortunately, no image authentication or image ‘proofing’ methods available. User security and other controls are also missing as well — these functions are left to the tools provided with either the target platform virtualization management platform or the target guest operating system to handle.
As PlateSpin licenses PowerConvert by number of conversions and features such as incremental conversion, and workload (operating system, application and data snapshots) recovery and movement, it’s a tool that does a versatile and egalitarian job of converting from numerous physical server platforms to virtual.
Furthermore it can be used for virtual-virtual ‘mirroring’, or to spawn numerous copies of ‘perfected’ or ‘generic’ images to virtual or physical hosts in PowerConvert advanced editions (such as the Enterprise Edition we tested). We were only mildly disturbed by the bugs we found in the conversion processes we tested (and only in P2V Windows, and then, it was tough to repeat). What PlateSpin does is a basic – but not easy – job of converting system images with no compatibility issues, with little fuss and no muss.
LeoStream P>V Direct 3.0
Leostream’s P>V is a physical to virtual conversion tool that competes with PlateSpin. It consists of a Proxy Server that resides in its own preconfigured Debian virtual machine on the virtualization host and an executable wizard program that sits on the machine being converted and walks an administrator through the P2V process. The host proxy controls the P2V conversion process and can create a new virtual machine or use an existing preconfigured VM.
In turn, the wizard integrates virtualized images into VMware ESX or Microsoft Virtual Server 2005 virtual environments. Leostream’s P2V product targets mid-market customers that may have significantly less source machines to convert than larger corporate customers by its packaging of five conversions for $600.
The P2V wizard is intended to be run interactively (one conversion at a time), but it’s also scriptable for those that want to automate the process of P2V conversion for like-type physical hosts. Unfortunately, Leostream doesn’t make use of any native scheduling concepts (it is conceivable that the native windows scheduler could be used, but we did not test that theory) and therefore you have to monitoring the conversions while they’re happening.
Each conversion is licensed separately by entering a unique key per source physical machine obtained from the wizard and subsequently logged into the company’s Web site to obtain a conversion key. Once a source machine is licensed it can be converted multiple times free of charge for a limited sort of backup/disaster-recovery scenario.
We found Leostream’s P>V capabilities to be useful for small batches of conversions or ‘one off’ conversions of about 10 to 20 machines. A system administrator with reasonable skills should make short work of conversions in the physical host systems we tested. Overall, we encountered no glitches in our admittedly ‘plain vanilla’ host to VM conversions.
In terms of conversion speed, Leostream was faster than average. LeoStream conversions took about 38 minutes for VMWare and 39 for XenEnterprise. Its disk imaging process utilizes a raw disk block level transfer rather than a file-by-file copy and therefore moves very quickly. These times beat PlateStream by 16 minutes, VMware’s best conversion time 76 minutes and XenEnterprise 4.0 two-hour conversion time.
Leostream’s Web interface is both understandable and rapidly productive in terms of identifying physical hosts, conversion steps, and controls in the target virtualized host environment. One small limitation is that there’s no way to resize the disk allotment until the conversion is done; it can’t be changed prior to or on the fly — only afterwards.
In a few of our conversions, when we didn’t use forethought in allocating physical space to cover the image, the processes failed. Pre-conversion software sanity checks would be helpful. Although all processes are logged for both the physical and virtual sides of the conversion, which can then be sent to Leostream support for guidance when things go awry.
Overall, we were pleased by the straight-forward use of P>V through its various stages. Leostream P>V adheres to the KISS principle (keep it simple stupid) in that it does what it’s supposed to do; nothing more, nothing less. What’s missing is a nice library of images, the ability to authenticate images in subsequent use, and a lot more management of images and policy/user administration would make this product more useful than it is now as a one-off conversion tool.
(Compare physical and virtual server management tools in our revamped Server Management Buyer’s Guide.)
Henderson is principal researcher and Dvorak is a researcher for ExtremeLabs in Indianapolis. They can be reached at thenderson@extremelabs.com.
Henderson is also a member of the Network World Lab Alliance, a cooperative of the premier reviewers in the network industry each bringing to bear years of practical experience on every review. For more Lab Alliance information, including what it takes to become a member, go to www.networkworld.com/alliance.
