Citrix and VMware: The Future (Part 3 of 3)

Rating:
Votes: 75 rating(s),  Score: 322/375
22
comments
20878 views
This is the third article in the series about VMware and Citrix and Terminal Server technologies. In this article, I look ahead to what EMC's purchase of VMware could mean, and how future solutions could be relevant in the server-based computing world.
Written by:
Brian Madden
Publication Date:
March 29, 2004
Doc #Id: 152


In the previous parts of this series (part 1, part 2), I talked about how VMware got to where they are today and how their products are relevant in the Citrix and Terminal Server world.

For this final part looking to the future, I don’t want to talk about VMware versus Microsoft’s Virtual Server. Instead, I want to talk about how virtual servers will continue to become more and more relevant in the server-based computing world.

First of all, if you haven’t done so already, you should definitely read my article about how utility computing will shape Citrix and Terminal Server.

I’ll talk a lot about EMC in this article, since they’re the new owners of VMware. Personally, I never thought too much about EMC. I viewed them as a niche storage maker and nothing else. When they bought VMware I could definitely see how the two companies’ visions were able to line up, but I wasn’t sure how they would package VMware and EMC storage into a single customer solution.

That all changed when I learned about a company called Egenera. Egenera is a server hardware manufacturer that builds blade and 1U modules that work together to form single virtual servers. Each physical module contains processors and memory, nothing else.

For example, you could buy four Egenera blades, each with 4 processors and 8GB of RAM. These four blades would work together to form a virtual server. The modules plug into a standard 19" rack, and redundant control units house the networking, KVM, and SAN connections.

If you start to run out of processing power, no problem. Just plug in another Egenera blade to add more capacity. If a particular blade fails, no problem. You’d lose a bit of processing power but the “server” would still run on what’s available.

Egenera calls this a “PAN” (as in "Processing Area Network"). Each individual Egenera blade plugs into the frame with a custom connector (that has a whole lot of pins), so the blades are connected by a lot more than Ethernet.

To me, this is the future of virtual server / utility computing technology.

Image what you could do if you combined EMC storage, VMware virtual server software, and Egenera virtual server hardware into a single solution from a single company. You could build a datacenter that was 100% virtual. You would never have to buy “extra” capacity, because each virtual server would use exactly what it needs. If your data center needed more processing capacity or storage, you could add additional resources that would be used wherever they were needed.

Since EMC bought VMware, they should also buy Egenera.

EMC could be successful in this space because the other four major server vendors (HP, IBM, Dell, and Sun) all have a vested interest in selling “traditional” servers. While market pressure will force them to “talk the talk,” they’ll want to hold off on “walking the walk” as long as possible. In today’s world, the fact that someone is forced to buy a 2GHz server just to bring up some little web application is great for them.

In the future, a utility computing world based on EMC/VMware/Egenera would allow you to bring little servers online without an explicit separate hardware purchase, and it would allow the new “server” brought online to be more resilient and have more growth potential that the old model.

This model is also much more realistic for scaling (by creating lots of little virtual servers) than Microsoft’s efforts to build software that can run many different applications on the same logical server.

After reading this article and my other article about utility computing and Citrix, I think it's pretty clear to see how these technologies could work together to create a redundant Citrix- and Terminal Server-based system that could power all the applications for an enterprise.

Reader Comments
Cable system-like reliability
Sunday, December 12, 2004 1:21:37 PM

Guest
This message was originally posted by Gabe Knuth on March 29, 2004
Years ago we talked about servers being turned on, turned off, and generally available as if it was the cable system (without the 4hr service window). Someone could call up and say I need to bring 500 users online and it would be done in hours rather than days.

I think that "hours" bar can be raised now. Minutes instead of hours, and with next to no hardware acquisition time. With Egenera systems, just pop in a few new blades...no assembly, mounting, cabling, or headaches. And EMC + VMWare means very fast, centralized, redundant access to your virtual servers and data, all running on that 50GHz machine!

I agree, this could be it. And, whether it becomes "it" or not, it's what I'd like to see. Who'd have thought a datacenter could be a turnkey solution? With all these products working together, you could put together a "Datacenter-in-a-box" solution. Put a few of those together and you have your Grid/Utility/Distributed computing architecture.
Egenera
Sunday, December 12, 2004 1:21:37 PM

Guest
This message was originally posted by matt in london on March 31, 2004
hello brian,

we're currently investigating partnering with egenera - they have some very interesting technologies, but also some glaring SPOF's - another issue is the CAPEX: > £1million for a 'simple' setup ... then you have to add a SAN as well !

m@
Now we are talking...
Sunday, December 12, 2004 1:22:15 PM

Guest
This message was originally posted by Freddie on May 11, 2004
I feel that we SBC freaks will have a bright future...
This is why Vmware is a good solution in SBC
Sunday, December 12, 2004 1:22:17 PM

Guest
This message was originally posted by Freddie on May 26, 2004
Hi People…
Since Brian have talked allot of virtual servers in the last three articles I thought it was time to give you all an explanation of why VMware can be a real good investment.
First of all, the only real vmware product that should bee used in SBC production environment is the ESX version, the reason is simple, this is the only version that have it’s “own OS”, A small very stable Linux OS that do not affect the virtual servers, you do not need to be worried that you have to restart all the virtual servers just because you need to update the host server and restart that.
Now, the first thing. 1. You can scale and fine tune your virtual servers hardware to “perfection”, (yes running the virtual servers takes some overhead of the host server but this is marginal). By using vmware you can put the recourses where there are most needed, for example. If you have four virtual servers on a quad. CPU hostserver and one of the virtual servers (from here on called VS) hosts a very CPU demanding application and the rest just ordinary applications you configure vmware to allocate two of the CPUs to CPU demanding VS. Not just the CPUs can be allocated but disks RAM also. 2. You can by fewer but bigger servers, must of the time this would save money and must of all administration. 3. You have a separate remote administration GUI that works like the traditional remote admin mode, the good thing whit this GUI is that it uses other ports then the well known 1494 and 3389, this ads a small security strength. And you will by default use a null session on your servers 4. This is one of the best features of vmware when it comes to SBC, you can whit a “simple click” clone you servers and make an image of them, you can then mirror the terminal/Citrix server very fast and simple and you do not need any sysprep or nothing, the vmware have it’s own virtual drivers so you will never bother whit the hardware compatibility, just restore your image on as many servers as you like. 5. Now we will talk about real uptime functions, by using VMotion you literally move an up and going VS from one ESX server to an other, whit out users ever knowing it. 6. The last thing I will mention is the new function in ESX, were it is now a setting in vmware were you can fine tune the VS to be used for terminal/Citrix servers, there is a check box that puts the VS in a high number of processes environment, thus makes it more efficient in SBC environments.
Of course these features can be debated and dissected as any other product by I find them so good that I believe VMware is a good solution for SBC, and I believe there will be more of it in the feature, like the things that Brian takes up in his articles. See ya…
Re: This is why Vmware is a NOT good solution in SBC
Thursday, March 03, 2005 7:03:43 AM

Guest
Performance.
Simply, the performance of terminal servers running on a 8 Xeon IBM x440 and VMware ESX server was bad. I tested that, and then I bought blades with real CPUs.
I guess the main reason is that one terminal server consolidates 50 PCs and therefore has less unused CPU cycles that can be given to other virtual servers.
RE: Re: This is why Vmware is a NOT good solution in SBC
Thursday, March 03, 2005 10:39:41 AM
Dev environments are still the best fit for VMWare IMHO.

Shawn
You're full of crap!
Sunday, December 12, 2004 1:30:33 PM

Guest
This message was originally posted by Ron on July 26, 2004
Brian, I met with Egenera last week and pointed out your comments in this article during our meeting and let's just say that they are very surprised.
You're full of crap - take 2 ;)
Sunday, December 12, 2004 1:30:33 PM

Guest
This message was originally posted by Ron on July 26, 2004
At this point in time the bladeframe platform acts as 24 individual servers and not a one superhost (single system image). If you lose a blade you lose the work (users) that are on that blade. Engenera does have some very nice technology but it no way shape our form does it act like "Windows would logically see a single processor and 32GB of memory. (Of course that processor would run as if it were a 50GHz Xeon Penitum 4.) " Sorry but your story does not hold up. btw, we are going to bring them in for a tco/roi study ;) but at 2x what it costs for dl580 of similar config, it might be a hard sell.
This is news to me!
Sunday, December 12, 2004 1:30:34 PM

Guest
This message was originally posted by Brian Madden on July 27, 2004
I was first introduced to eGenera at UtilCompWorld 2004 in NYC. As I said in the article, that was my fist exposure to them. I talked to the guy for awhile and asked several times about how their technology works, but at this time it looks like I'll have to revisit that. Thanks for the heads up Ron!
Please edit the article, or post new information...
Sunday, December 12, 2004 1:49:42 PM

Guest
This message was originally posted by Matthew on September 7, 2004
I suppose it is my fault for jumping without reading all the comments, but I almost made a big error bringing the Egenera solution to the attention of management here. If what Ron says is true, please edit your article above to reflect the errors so that people who don't read all the comments don't make similar mistakes after reading your article. If what Ron says is not true, please make a comment here saying that as well...
I have modified the article...
Sunday, December 12, 2004 1:50:02 PM

Guest
This message was originally posted by Brian Madden on September 15, 2004
...since these first comments were made. In genereal, it looks like this eGenera stuff is not quite as mature as I was hoping. It definitely shows promise, and you can use it today in very cool ways with virtual machines.
eGenera maybe not... But IBM can do it!
Sunday, December 12, 2004 1:52:31 PM

Guest
This message was originally posted by Andy Meakin on October 18, 2004
At IBM we have been building High Availability solutions based on IBM xSeries and BladeCenter solutions with IBM SAN Storage and VMware ESX Virtualisation for over two years now.

We do have a NUMA architecture designed Server called the x445 which adds multiple "quads" into a single Server instance scaling from 2 to 32 processors. It can also be split back up virtually into 4 processor chunks, and VMware is the finishing touch that allows fine granulation between these chunks. I think that this is a lot closer to what you describe than the eGenera product.

This technology in itself does not address true HA though. If a Node fails, the instances running on them will die too. We can fail-over and re-activate (reboot) a VM on another system, but that causes connection loss and risks data corruption. The answer here is to Cluster Virtual Machines on separate physical ESX Server Nodes giving online failover.

Much of this is detailed in the following document for further reading: http://www.redbooks.ibm.com/abstracts/sg246434.html?Open

Going forward, don't expect to see anyone doing much more in hardware. Lock-step concurrent processing is extremely dif