Year: 2004

They are not only different products laying on different hardware handlers (operating systems as we like to call them).
The most important fact is that they handle system resources in a total different way: GSX follows standard allocation resources rules (in other words acts like a standard application) while ESX reallocates memory and CPU when needed among VMs.
This implies that ESX requires much less base memory to run, and a small footprint custom kernel helps this more and more.

Another important implication is that, IMHO, running a performant GSX server is much more difficult than running a performant ESX server: in GSX there are so much factors linked to OS and applications handling that could negatively impact on VMs performances. Perfectly knowing how to light and speed up your hostOS is THE task before any other optimization you can do on GSX itself. The study of the best hostOS (between Windows and Linux many variants) would merit so many pages.
In ESX everything is in VMware engineers hands. Yes, you can hack some small aspect but the gain doesn’t do the difference. In a sense you have no choice, so there are less probabilities to go wrong.

If you look at GSX vs ESX in a how-many-features sense, then is simple: GSX is unbeatable, and 3.0 edition will demostrate this at the best.
I’m a big fan of GSX and always suggest it to my customers for many reason: first of all I don’t like “appliance phylosophy” where you haven’t full control on underlaying software, then ’cause today, with Windows 2003 hostOS, I can achieve tasks hard to achieve with Linux hostOS or ESX.
But I perfectly realize that ESX performances are superior, and less prone to negative events.

Both GSX and ESX are going to raise level to same features pool, and you will notice better within some weeks. Probably ESX will insist on performances and super enterprise features (multiprocessor support, native SAN support, VMs fail over and load balancing, and so on), while GSX will insist on flexibility in any scenario (I expect a further improvement about this aspect when virtual driver SDK will be released).

Count Susquehanna Financial Group among the doubters of EMC’s $635 million bid, which it says is expensive considering VMware’s 2003 sales.

While encouraged by EMC’s earlier purchases of Legato Systems and Documentum to bolster the company’s information lifecycle management (ILM) strategy, Susquehanna views EMC’s interest in VMware as largely impractical.

“VMware provides little synergy with storage, whereas Legato and Documentum are building on EMC’s Information Lifecycle Management strategy,” a Susquehanna report said. “VMware, which makes server provisioning software for Windows and Linux environments, is in a different space altogether.”

Susquehanna analysts Kaushik Roy and Phillip Rowe also voiced concern that VMware’s revenues are from a low price-point software product with limited growth potential.

VMware offers a number of products, but most revenue comes from software licenses of VMotion, ESX server, and GSX server. VMotion helps customers consolidate servers by migrating applications from smaller servers to new, higher-end servers with no downtime for the applications.

“Although this is a valuable tool, we believe these migrations are relatively infrequent (perhaps several times a year),” Roy and Rowe said in their report.

The ESX and GSX server products allow users to partition their Intel servers for the usage of multiple operating systems and applications simultaneously.

“With already about 2 million users and less than $50 million in revenue last year, we can conclude that these products have a low price point,” Roy and Rowe said.

But Mark Stahlman, who covers EMC for American Technology Research, feels VMware could prove very valuable for EMC.

He sees what other analysts from companies such as Sageza Research or Gartner see: a company undergoing a metamorphosis as a leading purveyor of storage hardware to a software provider of so-called on-demand, or utility, computing.

Stahlman dismisses the term utility computing as unclear marketing hype, preferring the phrase “virtual computing” to describe a software-driven strategy that configures and reconfigures computing and storage systems with no downtime.

Stahlman told internetnews.com EMC’s recent purchases are evidence that the company is trying to improve its position in storage at the same time it is building a broader strategy in virtual computing.

The increased interest in virtualization as a viable computing platform technology is paving the way for an expansion into servers and networks, Stahlman said. IBM, and HP already have placement in this arena.

“EMC’s benefit from the deal is likely to be an opportunity to integrate its own R&D with the products of VMware as well as gain access to the new company’s considerable x86 expertise — crucial now that the Intel architecture is the largest volume for EMC and now that the 64-bit versions of x86 are getting ready to expand rapidly,” Stahlman said.

While he wouldn’t make any predictions outright, Stahlman said EMC repeatedly referenced network virtualization in a conference call, which he said could be an area EMC is targeting while it digests VMware.

Among the players in the networking virtualization is Inkra Networks. The Fremont, Calif.-based company integrates multiple services, such as firewall, VPN, intrusion detection, SSL, and load balancing in hardware platforms.

Meanwhile, Gartner analysts believe an application monitoring and management or server provisioning company is on tap for EMC. And Sageza Research Director Charles King recently told internetnews.com he expects EMC to acquire database and directory components.