I once was a huge fan of HP’s Virtual Connect Flex-10 10Gb Ethernet Modules but with the new enhancements to VMware vSphere 5, I don’t think I would recommend for virtual environments anymore. The ability to divide the two onboard network cards up to 8 NICS was a great feature and still is, if you have to do physical deployments of servers. I do realize that there is the HP Virtual Connect FlexFabric 10Gb/24-port Module but I live in the land of iSCSI and NFS so that is off the table for me.

With vSphere 5.0, VMware improved on its VMware’s Virtual Distributed Switch (VDS) functionality and overall networking ability, so now it’s time to recoup some of that money on the hardware side. The way I see is most people with a chassis full of blade servers probably already have VMware Enterprise Plus licenses, so they are already entitled to VDS, however what you may not have known is that customers with VMware View Premier licenses are also entitled to use VDS. Some of the newest features found in VMware VDS 5 are:

The last feature is the one that makes me think I don’t need to use HP’s Flex-10 anymore. Network I/O control (NIOC) allows you to assign shares to your network interfaces set priority, limits control congestions all in a dynamic fashion. What I particularly like about NIOC as compared to Flex-10 is the wasted bandwidth with hard limits. In the VDI world, the workload becomes very bursty. One example can be seen when using vMotion. When I’m performing maintenance work in a virtual environment I think it sure would be nice to have more than 2 GB/s a link to move the desktops off – however when you have to move 50+ desktops per blade you have to sit there and wait awhile. Of course, when this is your design, you wait because you wouldn’t want to suffer performance problems during the day by lack of bandwidth on services.

A typical Flex-10 configuration may break down the on board nic (LOM) something like this

To get a similar setup with NIOC it may look something like this.

Total shares from above would be: 5 + 50 + 40 + 20 = 115

In this example FT, iSCSI and Replication don’t have to be counted as they will not be used. The shares only kick if there is contention. The shares are also only applied if the traffic type exists on the link. I think it would best practice to limit vMotion traffic as multiple vMotions kicking off could easily exceed the bandwidth. I think 8000 Mbps would be reasonable limit with this sort of setup.

Management: 5 shares; (5/115) X 10 Gb = 434.78 Mbps

NFS: 50 shares; (50/115) X 10 Gb = 4347.83Mbps

Virtual Machine: 40 shares; (40/115) * 10 Gb = 3478.26Mbps

vMotion: 20 shares; (20/115) X 10 Gb = 1739.13Mbps

I think the benefits plus the cost saving is worth moving ahead with a 10GB design with NIOC. Below are some list prices taken on November 28, 2011. Which one are you going to choose?

Flex-10

HP_6120G-XG

Dwayne is the newest Contributor to BladesMadeSimple.com and is the author of IT Blood Pressure (http://itbloodpressure.com/) where he provides tips on Virtual Desktops and gives advice on best practices in the IT industry with a particular focus in Healthcare.  In his day job, Dwayne is an Infrastructure Specialist in the Healthcare and Energy Sector in Western Canada.

With all that is made of the competition between blade server manufacturers and the growth of the blade server market in general, is there room for another type of condensed computing in the data center? Have we been going about things all wrong with regard to architecture design?

Nutanix thinks so.

Nutanix is a start-up company geared towards delivering a simplified virtualization infrastructure with a strong focus towards eliminating the need for a SAN. Their clustered solution brings storage and compute together which theoretically reduces expense, reduces complexity, and improves performance. On its own it doesn’t really seem that innovative but the secret sauce is how they make the cluster scale and tier/span data across all nodes without sacrificing performance. Each node has the usual compute resources plus a mix of local SSD and SATA hard disks. There are 4 nodes per 2u enclosure called a “block”. Add more blocks and you have a Nutanix cluster. The software stack scales and balances everything between the nodes and blocks. The technology originated from the architecture that companies like Google and Facebook employ in their data centers. Assuming that can be taken at face value, the scalability potential is phenomenal.

So what’s the big deal?

Well my thinking is that if you can eliminate the need for a SAN (for virtualization) then you can definitely eliminate the need for an enclosure of blade servers. No interconnects. No Enclosure. Simplified network architecture. No SAN. What’s not to love?

Eliminate the SAN. Really?

The 37 million dollar question that comes to mind is: Can I really eliminate my SAN? For me the answer is a quick and resounding, “No I cannot!” My SAN does a heck of lot more than deliver servers through a hypervisor. There are still application, database and file servers depending on a direct connection to that centralized beast of disk technology. Many of those servers may be virtualized but some of their data volumes are simply best served as LUNS and not ridiculously large VMDK files. Granted, the topic of VM data placement is debatable but it becomes moot if I have any physical servers around needing centralized storage. In my environment, and I feel confident I’m not alone in this, the SAN can’t really go away.

Specific to virtualized workloads, Nutanix claims that there is plenty of capacity for just about any scenario. Personally, I still doubt the clustered Nutanix nodes have the local capacity to absorb both the VMs and their data with redundancy. A single block (made up of 4 nodes) has some solid state goodness and 20TB worth of storage (5TB per node). As the cluster scales by adding blocks of nodes the increased pool of storage is available to all VMs. For some environments that may be more than enough capacity to contain everything related every virtual server. For other environments this may not even be close to the needed capacity. I also wonder what happens when I drink the Nutanix kool-aid and then later hit the capacity wall? I’d feel pretty silly running out of space and having to buy a full compute block just to increase storage capacity.

Regardless, it does seem possible that Nutanix could displace the SAN completely if your environment is 100% virtualized and capacity requirements aren’t too drastic. Realistically though, since most businesses are not 100% virtualized, this technology simply reduces the need for capacity expansion and the worry of performance tuning the SAN for VM workloads. That definitely has some value but how much depends on how virtualized you are and how much data your VMs are connected to (among other things).

Ok. So how about compute?

Today we have stacks of blade enclosures full of ESX hosts driving maximum efficiency of network and power. Or are they? Let’s say you have a HP c7000 full of half height BL460 servers. For every 10u of rack space you get 16 dual socket ESX hosts. With the Nutanix solution there are 4 hosts per every 2u block so that same 10u can hold 20 similarly configured physical hosts. Within that same 10u there is also a collective 6TB of Fusion IO, 6TB of SATA SSD, and 100TB of SATA disk. The density win clearly goes to the Nutanix solution.

From a networking perspective things are a bit less clear. Within the HP enclosure, using Flex-10 virtual connect, there would be two or more 10Gb uplinks per enclosure (16 physical hosts). For each pair of uplinks one would be active with the other ready for fail-over. You could get away with a single pair of 10Gb uplinks but that would be silly for a full enclosure of hosts. I’d argue that more than four uplinks would likely be needed but that’s a different conversation. The main point is that there is some flexibility and design decisions as to how the network is configured.

By contrast the Nutanix has four 10GB uplinks per 2u block (one for each node within the block). Comparing the same 10u rack space you end up with 20 10Gb uplinks. Nearest as I can tell there is no aggregation of the 10Gb interfaces. From a pure performance perspective you can’t argue with the simplicity and potential of a dedicated 10Gb connection per host. On the other hand that is a huge number of 10Gb ports and certainly increases the cost of the networking infrastructure. The single interface per node also doesn’t give any option for fail-over. At this point I’m not entirely clear on what all of the networking capabilities are because Nutanix literature focuses purely on all of the SAN oriented benefits and I haven’t yet connected with them to follow up. I’d like to learn a bit more about their networking architecture before I declare an advantage to one or the other. I’ll report back what I find out.

Bottom Line

Nutanix has a pretty good start on a new way of delivering virtualization. The idea of consolidating compute and storage isn’t entirely new (Pivot3, for example, has had VMs running on their storage arrays for a couple of years now) but Nutanix has absolutely evolved the concept. If they can deliver on all of their promises for less money than “the big guys” they stand a reasonable chance at being disruptive to both storage and blade center markets in the niche of virtualization.

One area where I can see this strategy really taking off is with VDI deployments. Who wouldn’t want a dedicated stack of gear running at high performance with plenty of capacity to spare? The total value of this technology,however, depends on a number of unanswered questions both technical and financial. Assuming it is as cost competitive as it is claimed to be, certainly green field environments should take a serious look and determine if a dedicated purpose built virtualization cluster has more value than a more complex but more versatile traditional storage and compute approach. Those that have existing storage and compute stacks will need to do a much more comprehensive evaluation because there’s a lot more to the value of “traditional” data center technology than consolidated server workloads.

Nutanix is cool, but I’m not replacing all of my blade servers just yet.

To find out more information about Nutanix click here.

To find out more about HP blade servers click here.

Chris is a contributor and author for BladesMadeSimple.com and in his day job is a Systems Architect for Clackamas County in Oregon and an HP customer.  The opinions expressed in his writings do not necessarily reflect those of BladesMadeSimple or Chris’ employer.  Follow Chris on Twitter @sysgeekguy.

I always find it interesting to hear how customers are using blade server technology “in the real world.”   Brickworks, a $1.7 billion Australian brick company with over 60 locations and 1,000 employees, recently provided The Australian with details about how Citrix XenApp Enterprise combined with VMware vSphere and Dell rack and blade servers helped develop a more efficient workplace for its employees.  Read the full article at:

http://www.theaustralian.com.au/australian-it/brickworks-builds-solid-system-using-citrix/story-e6frgakx-1226129943367

Kevin Houston is the founder of BladesMadeSimple.com.  He has over 14 plus years of experience in the x86 server marketplace.  Since 1997 Kevin has worked at several resellers in the Atlanta area, and has a vast array of competitive x86 server knowledge and certifications as well as an in-depth understanding of VMware and Citrix virtualization.    Kevin works for Dell as a Server Sales Engineer covering the Global 500 market. 

#IWorkForDell

Since my last post, I heard lots of rumors that VMware would make some changes in their new licensing model. Well, this week VMware made an official announcement of changes and posted here.

VMware made very clear that all the changes were motivated by the feedbacks they have received:

“These changes generated debate in the blogosphere, across the VMware communities, and in conversations with customers and partners.  Some of the discussion had to do with confusion around the changes.  We have been watching the commentaries on the blogs very carefully, and we have been listening in the customer conversations very intently. We got a ton of feedback that probed the impact of the new licensing model on every possible use case and scenario, and equally important, reflected our customers’ intense passion for VMware.” Said the Product management VP, Bogomil Balkansky.

This shows how important feedback is! We can say that changes come from us.

So, let´s see the new comparison of vSphere 4.x licensing vs the vSphere 5 model:

To see the original document, click here.

The main changes:

• Increased vRAM entitlements for all vSphere editions, including the doubling of the entitlements for vSphere Enterprise and Enterprise Plus.
• Capped the amount of vRAM  in any given VM by 96GB.
• Compliance policies will not be measured by high watermark. Instead, a 12 months rolling average of daily high watermark. That´s very important for testing and developing environments.

Plus: An official monitoring tool will replace all the scripts that have been circulating around the community to calculate how much “vRAM” customers are using now.

VMware confirmed that the new licensing model was made to keep up with the market changes and that changes are unlike, but necessary.

So, to compare apples to apples, let’s use the same example of last post: an IBM HX5 blade server with MAX5 maxed out with 640GB of memory with the NEW (I mean the newest) licensing model a user would need 7 VMware vSphere 5 Enterprise Plus Licenses to be “legal” vs 14 of the previous vSphere 5 licensing vs 2 with the current VMware vSphere 4.x licensing model.

That´s more reasonable. Especially if keep in mind that customers would not use all the physical RAM in the vRAM pool. They must have some free physical RAM to be able to do a vMotion for example.

We have to give VMware some credit to change so quickly and listen the community. Would any other vendor do something like this? I don´t know.

About the Author
Thales Osterne is a contributor for BladesMadeSimple.com.  He has over six years in the IT field with four years of experience in IBM BladeCenter and System X.  When he is not blogging, Thales works as a product manager for IBM System X & Bladecenter at Lanlink Informatica, a major business partner in Brazil.  He is fluent in Portuguese and English.

As you no doubt heard by now, VMware has announced a new version of vSphere along with some new or improved features however, this post will not highlight these features.  In this post, I want to talk about what did not improve – the licensing. 
 
With the announcement of the VMware vSphere 5, the biggest thing that startles me is the way that VMware has found to a way to collect more revenue from users with new server technologies. This seems to be totally going against the flow of x86 market since the main message of the major server manufacturers (in the past) has been how they can save users $$ on VMware licensing by offering the ability to stack your servers full of memory.  

For example – take a look at the IBM HX5 blade server combined with the MAX 5 which provides users with a total of 40 memory DIMM slots and a current maximum of 640GB using only two processors.  Taking a look at the comparison, with the previous vSphere 4.x licensing, you would only need two VMware vSphere 4.x Enterprise Plus licenses. (With 640GB of RAM, one can only imagine how many virtual machines were possible with just 2 licenses.)
 
Since more users were moving to at the architecture of larger 2 socket systems with A LOT of memory, I believe that VMware realized the vSphere licensing  opportunities were than they expected so they decided to change their licensing method  to  a “per CPU and memory” licensing model creating the concept of a processor and vRAM pooling of entitlements.
 
Here is a look at a comparison of vSphere 4.x licensing vs the vSphere 5 model:

 
So, if we use the example of a the IBM HX5 blade with MAX5 mentioned above maxed out with 640GB of memory with the NEW licensing model a user would need 14 VMware vSphere 5 Enterprise Plus Licenses to be “legal” vs 2 with the current VMware vSphere 4.x licensing model. 
 
According to VMware, the good news is that there is “no limit to the number of cores per processor”, but in reality  – I don’t see that that as a big deal. vSphere 4.X already supported 12 cores per processor with vSphere Enterprise Plus licenses which is the maximum of the current market  – even the next generation Intel 2 socket CPU will only be 8 cores – so there is no advantage to changing the licensing.  This is why many people are calling this new licensing model “vRAM Tax“.
 
We all agree that VMware has the best products and is the market leader for virtualization, BUT this new licensing model is giving ammo to Microsoft and Citrix to strike at the cost issue (which was already an issue in VS 4.x).

About the Author
Thales Osterne is a contributor for BladesMadeSimple.com.  He has over six years in the IT field with four years of experience in IBM BladeCenter and System X.  When he is not blogging, Thales works as a product manager for IBM System X & Bladecenter at Lanlink Informatica, a major business partner in Brazil.  He is fluent in Portugese and English.

Over the past couple of weeks, I’ve been researching what are the “ideal” configurations for a home VMware vSphere lab.  Ideally, it would be small enough to run a few virtual machines and demonstrate high availability features like vMotion, DRS, HA or even FT.  The problem that I’ve run into is that I haven’t found a really good site that has a detailed list of parts and pieces to buy.  The problem that I’ve run into is that some people recommend rack servers, others recommend you use tower servers and all of the designs includes external switching, connectivity to storage.  As you can imagine, this leads to a mess. 

Sure, it’s probably okay to have a pile of cables, switches and servers generating noise and heat in your home office, but what if you didn’t have to?  One of the big value propositions that blade servers offer is the ability to reduce your cable, management and power complexity – so what if there was a blade infrastructure that would be cost efficient for a home lab and was robust enough to server your virtual needs.  I’m working on finding a blade solution to be able to write about that will solve your lab needs, but this is where I need your help.  I need to know what your ideal lab environment would look like.  Take a few minutes and help me out by answering the following questions (either via comments below, or email me at kevin AT bladesmadesimple.com):

• What’s the MINIMUM amount of servers would you need for your lab?
• What’s the MINIMUM memory you’d want for your lab?
• What’s the MINIMUM storage you’d need for your lab?
• What’s the MAXIMUM price you’d pay for a bundled blade solution (servers/storage/networking)?

Thanks for taking the time to answer – feel free to comment about what your needs would be for a Citrix or Microsoft lab as well (no bias here.)

As VMworld 2010 is right around the corner, I wanted to take a few minutes to make a plea to all attendees. 

This year, IF you receive a bag or backpack that you just don’t want, please don’t throw it away, but instead take it home, go to the dollar store and fill  the backpack with pencils, crayons, paper and erasers and donate it to your local school system.  You would be AMAZED to find out the numbers of children who don’t get backpacks and whose familes can not afford the costly school supplies that are required each year.  You will be making some family happy and you’ll get the name “VMware” marketed throughout the schools, getting the next generation of techno geeks ready to learn all about virtualization.

Thanks for the consideration!

(Updated 7/27/2010 – 11 am EST – added info on power and tower options)

 When you think about blade servers, you probably think, “they are too expensive.” When you think about doing a VMware project, you probably think, “my servers are too old” or “I can’t afford new servers.” For $8 per GB, you can have blade servers preloaded with VMware ESXi 4.1 AND 4TB of storage! Want to know how? Keep reading. 

No, I’m not smoking something. I’ve done the configuration and I can show you how to achieve a 4 TB SAN and 3 ESX hosts on blade servers with the IBM BladeCenter S. Before I can explain what I’ve done, let me give you the basics of the IBM BladeCenter S. 

Overview of the IBM BladeCenter S 

At 7U high, the chassis of the IBM BladeCenter S is the same height as the original IBM BladeCenter (now called the BladeCenter E). The chassis uses the same blade servers as the rest of the IBM blade chassis family, but the chassis holds only 6 blade servers – primarily due to the addition of locally attached storage drives. In addition, the chassis has the option to add a DVD drive for access to local media. 

In contrast, the SAS RAID Controller Module (IBM part #43W3584) allows for you to pool the storage and offer access to these arrays to each blade server that has the SAS Connectivity Card installed. Volumes that are created can be assigned to a specific blade or shared by several blade servers. The IBM SAS RAID Controller Module supports RAID levels 0, 1, 5 and 10 and each module also comes with RAID battery backup module. There are some caveats to be aware of: only SAS or NL SAS drives are supported (no SATA); the maximum volume size is currently limited to 2TB and the maximum number volumes each blade server can have is 8 (for a total of 48 volumes per chassis.) Another important thing to take note is that you must have 2 x SAS RAID Controller Modules, which sit in I/O Bays #3 and #4. This provides a redundant connection for each blade server with the SAS Connectivity Card. In fact, since I brought it up, let’s take a closer look at how the modules work in the IBM BladeCenter S. 

BladeCenter S a Look From BehindWhen you look at the back of the IBM BladeCenter S, it may look confusing, but don’t worry – it’s pretty straight forward. The chassis comes standard with 2 x 1450w power supplies standard and a single Advanced Management Module. If you are using high wattage blade servers or using the second DSM, you probably will need the 2nd set of power supplies (IBM part # 46C7438). If your budget is tight and you can’t afford to pony up the additional $599 U.S. list I recommend you take advantage of IBM’s Power Calculator prior to purchase to see if you need the 2nd set of power supplies. Following the design of the other IBM BladeCenter chassis, the IBM BladeCenter is cooled by a set of 4 redundant hot-swap blower modules. Don’t bother looking for any other fans or cooling devices, because you won’t find them. These four blowers cool the entire chassis, modules and blade servers. 

Management 
The Advanced Management Module (AMM) is the device that provides you with LOCAL keyboard, video and mouse connectivity (although only USB for keyboard and mouse) as well as an ethernet port to connect into your management network. The AMM gives you the ability to manage / monitor all of the chassis’ thermals as well as remotely control the blade servers and the I/O modules. In all honesty, the AMM is feature rich, so if you want to take a peek at what it can do, take at look at this IBM BladeCenter S Advanced Management Module simulator. Unlike the other IBM blade chassis, there is not an option for a redundant AMM, however in the event of a failure your blade servers, I/O modules, fans and blowers will continue to function without penalty. 

I/O Architecture 

The biggest technical confusion I see from engineers and customers alike is around the I/O layout of the blade chassis. The IBM BladeCenter S is a bit dissimilar to the other IBM chassis in the BladeCenter family so let me explain how it works. There are 4 I/O bays in the IBM BladeCenter S. The 1st I/O bay maps to the NICs that come on the motherboard of each blade server. If you are familiar with rack-mount servers, you know they typically have 2 x 1Gb Ethernet ports. The IBM Blade servers are no different – they also have 2 x 1Gb Ethernet NICs. In order for them to be “lit up” you need to have a module in bay 1 that can allow the signal from the blade server to extend out of the chassis. To simplify things – think of having a power outlet in the wall at home and connecting an extension cord to it so you can turn on a light that is a few feet away. The same rudimentary concept is how it works in the blade infrastructure. The only difference is, with the IBM BladeCenter S, both NIC ports 0 and 1 go to the I/O Module Bay #1. This means if that module has an issue, then those 2 NICs located on the motherboard of each blade server will be dead. There is no redundancy with the onboard NICs in the IBM BladeCenter S (unlike the other IBM BladeCenter chassis.) Why did IBM design it this way? Well, the original target market for the IBM BladeCenter S was small businesses and remote offices. When you look at those environments – how many have redundant NICs for the rack / tower server NICs? Odds are none. With that in mind, IBM designed the BladeCenter S to only have a single I/O module for both onboard NICs. Never fear, though. After a few months, IBM revised the design to allow for I/O Module #2 to provide an additional 2 NICs, using the 2/4 port Ethernet adapter (IBM part # 44W4479) on each blade server. The card is designed to provide 4 Ethernet Ports, however with the BladeCenter S, only 2 ports are connected. Therefore with network modules in I/O Module Bay 1 and 2 you can get 4 NICs. Add this to the 2 x SAS storage cards that we mentioned in the sections above and you “should” have adequate architecture to provide a VMware environment. 

The $32,000 Design 

Now that I’ve spent a few moments telling you what the IBM BladeCenter S is all about, perhaps you understand the potential. So how did I get to the $32,000 design that enables you to have 4 TB and 3 ESX hosts? I won’t devulge in the actual bill of materials, but here’s what I came up with: 

• 1 x IBM BladeCenter S chassis
• 1 x Disk Storage Module
• 4 x 1 TB Near Line Storage Disk Drives
• 1 set of 1450W Power Supplies
• 2 x Server Connectivity (Ethernet) Modules
• 2 x SAS RAID Connectivity Modules
• 1 x DVD
• 3 x HS22 blade servers – each with 2 x Intel E5620 Xeon Processors, 24GB RAM, SAS Connectivity Card, ESXi 4.1 USB Memory Key

Total U.S. List Price (as of 7/26/2010): $30,768.00 

Yes, I also know that HP has an offering (the BladeSystem C3000) that could compete to this design – however IBM is the only blade server vendor that offers dedicated local disk storage. HP’s design takes up a blade server slot. Perhaps I’ll write up something on this in the future. 

THIS SECTION ADDED ON 7/27/2010 – I’ve added this section to cover a couple of pieces that I left off in the original post.

Power
A really valuable feature of the IBM BladeCenter S is the ability to run on 110 v or 220 v.  Use of 110 v is ideal for remote or small offices.  The BladeCenter S power supplies are auto-sensing so you can use the same power supplies.  There are a few power categories to choose from with the BladeCenter S:

• Redundant AC Power Source – in this policy, the power limit is set to equal the capacity of N power modules.  According to IBM’s Implementing the IBM
  BladeCenter S Chassis Redbook, this policy is the most conservative approach and is recommended when all four power modules are installed. When the chassis is correctly wired with dual AC power sources, one AC power source can fail without affecting your blade server operation. 
• Redundant Power Module Policy – in this policy, the power limit equals the capacity of one less than the number of power modules installed (more than one power module must be present). One power module can fail without affecting blade server operation.  If a single power module fails, all the blade servers that are powered on will continue to operate at normal performance levels.
• No Redundancy – all power modules are used, there is no redundancy and if you lose a power supply and the power demands exceed the capacity of the available power modules, the chassis will power down… Not recommended.

Office Enablement Kit
For those environments where a standard server rack is not ideal, IBM offers the Office Enablement Kit (part # 201886X).  This adds an additional few hundred dollars, but it gives you an 11u rack enclosure, complete with front and rear locking doors and wheels.  It also comes with an acoustic attenuation module that helps muffle the sound.  (YouTube video on this can be seen here.)  As mentioned above, the IBM BladeCenter S is only 7U tall, so the additional 4U can be used for an optional Flat Panel Monitor kit (shown in the image to the left) or perhaps additional storage or networking.  This kit really helps to finalize an “all-in-one” solution for small or remote environments.

Yes, I know it’s only 4 x 1TB drives, and I know it’s Near-Line SAS drives, however it is enough resources to help YOU create that VMware infrastructure that you need. Of course, the VMware licensing will be extra, but I just saved you a ton of money – now you can afford it… 

So – what do you think? Is this appealing, or is this just a pipe dream? Let me know your thoughts. I’m really interested in getting an idea of whether this design would really work in your world.

(Updated to include links to results)

I’ve had a few questions lately about “the best” blade server to use for virtualization – specifically VMware virtualization. While the obvious answer is “it depends”, I thought it would be an interesting approach to identify the blade servers that ranked in the top 5 in VMware’s VMmark benchmark.  Before I begin, let me explain what the VMmark testing is about.   VMmark enables equipment manufacturers, software vendors, system integrators and other organizations to:

• Measure virtual machine performance accurately and reliably
• Determine the performance of different hardware and virtualization platforms
• Make appropriate hardware decisions for your virtual infrastructure

VMware developed VMmark as a standard methodology for comparing virtualized systems. According to VMware’s VMmark website, the benchmark system in VMmark is comprised of a series of “sub-tests” that are derived from commonly used load-generation tools, as well as from benchmarks developed by the Standard Performance Evaluation Corporation (SPEC®). In parallel to VMmark, VMware is a member of the SPEC Virtualization subcommittee and is working with other SPEC members to create the next generation virtualization benchmark.

In testing the terms, a “tile” is simply a collection of virtual machines (VM’s) that are executing a set of diverse workloads designed to represent a natural work environment.   The total number of tiles that a server can handle provides a detailed measurement of that server’s consolidation capacity.  The more tiles, the better.  The faster the performance, the better.

THE RESULTS (as of 6/2/2010)…click on the link to open the details

24 Cores (4 Sockets)
HP ProLiant BL685c G6 (Six Core AMD Opteron 8435) running VMware ESX v4.0 – 29.19@20 tiles (published 7/14/2009)
HP ProLiant BL680c G5 (Six Core Intel Xeon E7458) running VMware ESX v3.5.0 Update 3 - 18.64@14 tiles (published 3/30/2009)

16 Cores (4 Sockets)
Dell PowerEdge M905 (Four Core AMD Opteron 8393 SE) running VMware ESX v4.0 – 22.90@17 tiles (published 6/19/2009)
HP ProLiant BL685 G6 (Four Core AMD Opteron 8389) running VMware ESX v4.0 – 20.87@14 tiles (published 4/24/2009)

12 Cores (2 Sockets)
Cisco UCS B250 M2 (Six Core Intel Xeon X5680) running VMware ESX v4.0 Update 1 – 35.83@26 tiles (published 4/6/2010)
Fujitsu BX922 S2 (Six Core Intel Xeon X5680) running VMware ESX v4.0 Update 1 – 32.89@24 tiles (published 4/6/2010)

8 Cores (2 Sockets)
Fujitsu BX922 S2 (Four Core Intel Xeon X5677) running VMware ESX v4.0 Update 1 – 27.99@18tiles(published 5/10/2010)
HP ProLiant BL490c G6 (Four Core Intel Xeon X5570) runningVMware ESX v4.0 – 25.27@17tiles (published 4/20/2010)

THE WINNER IS…
Cisco UCS B250 M2 running VMware ESX v4.0 Update 1 – 35.83 with 26 tiles

Cisco’s Winning Configuration
So – how did Cisco reach the top server spot?  Here’s the configuration:

server config:

• 2 x Intel Xeon X5680 Processors
• 192GB of RAM (48 x 4GB)
• 1 x Converged Network Adapter (Cisco UCS VIC M81KR)

storage config:

• EMC CX4-240
• Cisco MDS 9134
• 1173.48GB Used Disk Space
• 1024MB Array Cache
• 50 disks used on 5 enclosures/shelves (1 with 14 disk, 4 with 9 disks)
• 55 LUNs used
  *21 at 38GB (file server + mail server) over 20 x 73GB SSDs
  *5 at 38GB (file server + mail server) over 20 x 73GB SSDs
  *21 at 15GB (database) + 2 LUNs at 400GB (Standby, Webserver, Javaserver) over 16 x 450GB 15k disks
  *5 at 15GB (database)  over 16 x 450GB 15k disks
  * 1 LUN at 20GB (boot) over 5 x 300GB 15k disks
• RAID 0 for VMs, RAID 5 for VMware ESX 4.0 O/S

As you can see from the information above, the Cisco UCS B250 M2 is the clear winner above all of the blade server offerings.  As you can see, none of the Xeon 7500 blade servers have yet to be tested but when they do, I’ll be sure to let you know.

Cisco is getting some (more) recognition with their UCS blade server product, as they recently achieved the top position for “8 Core Server” on VMware’s VMmark benchmark tool.  VMmark is the industry’s first (and only credible) virtualization benchmark for x86-based computers.  According to the VMmark website, the Cisco UCS B200 blade server reached a score of 25.06 @ 17 tiles.  A “tile” is simple a collection of virtual machines (VM’s) that are executing a set of diverse workloads designed to represent a natural work environment.   The total number of tiles that a server can handle provides a detailed measurement of that server’s consolidation capacity.

Cisco’s Winning Configuration
So – how did Cisco reach the top server spot?  Here’s the configuration:

server config:

• 2 x Intel Xeon X5570 Processors
• 96GB of RAM (16 x 8GB)
• 1 x Converged Network Adapter (Cisco UCS M71KR-Q)

storage config:

• EMC CX4-240
• Cisco MDS 9130
• 1154.27GB Used Disk Space
• 1024MB Array Cache
• 41 disks used on 4 enclosures/shelves (1 with 14 disk, 3 with 9 disks)
• 37 LUNs used
  *17 at 38GB (file server + mail server) over 20 x 73GB SSDs
  *17 at 15GB (database) + 2 LUNs at 400GB (Misc) over 16 x 450GB 15k disks
  * 1 LUN at 20GB (boot) over 5 x 300GB 15k disks
• RAID 0 for VMs, RAID 5 for VMware ESX 4.0 O/S

While first place on the VMmark page (8 cores) shows Fujitsu’s RX300, it’s important to note that it was reached using Intel’s W5590 processor – a processor that is designed for “workstations” – not servers.  Second place, of server processors, currently shows HP’s BL490 with 24.54 (@ 17 tiles)

Thanks to Omar Sultan (@omarsultan) for Tweeting about this and to Harris Sussman for blogging about it.