VMware Communities > Blogs > Virtualization Frontier > 2008

Blog Posts

Virtualization Frontier : May 2008

Previous Next
0

A recent thread here on delltechcenter asked for a power consumption comparison between the PowerEdge 1950 1U rack mount server vs the PowerEdge M600 blade. It turns out that there are a few whitepapers that have been done on 2950 power consumption and a big study on the blades power consumption - but nothing I can find that directly compares the two. The best way to get a comparison between just about any current or recent Dell server is to use the Dell Datacenter Capacity planner. It allows you to build out a rack of servers and get an estimate of power consumption.

If you use similar configurations for 16 PowerEdge 1950s and a full M1000e chassis (16 blades) then the blades configuration will consume less power. So you can save a significant amount of power by going with blades over similar traditional rack mount servers.

If you combine the efficiencies of the blades with the efficiencies of running lots of VMs on a single physical server - you might have the ultimate in power efficiency. This is one reason to use blades for virtualization - we'll be chatting about this and other reasons next week in our TechTuesday chat.

Todd

0 Comments Permalink
0

As many of you have probably figured out by now, I've spent a lot of time working with Exchange in VMs. Some of that time was spent explaining to peers, users, customers, and family members just exactly why in the world anybody would want to put Exchange into a VM. (To be honest - I don't think I was able to convince my mom.)

Early this week I ran across an entry titled Eight Good Uses for VMware and Other Virtual Machine Software on the TechnologySpeak blog. Number three on the list - Running an Exchange Server to Keep Outlook Synchronized. Turns out that having a local Exchange Server on your home network can make it much easier and faster to sync up Outlook across the 4 or 5 systems that you have at your house. If you are reading this entry, then I would bet that you have at least that many.

The only caveats that the blog calls out:

"The initial setup is a little bit tricky if you're not familiar with installing servers and server software, but once you get it going, it works very smoothly. Also, for many, the initial cost of buying the software might be prohibitive."

Meaning that you have to be a geek or a professional Exchange admin or both. On the cost perspective I initially thought that it might be really really prohibitive, but a little digging around and it looks like Microsoft Small Business Server licenses aren't too bad. The list price for SBS R2 is $599, which includes the Windows Server OS, Exchange Server, and 5 CALs.

With the memorial holiday weekend coming up, I think that everybody should take a break from the grill and spend some time getting your own virtualized Exchange Server setup for the house.

Todd

0 Comments Permalink
0

Massive IOPS on ESX

Posted by ToddMuirhead May 22, 2008

They performance guys over at VMware have announced some really impressive IOPS testing results. Using three CX3-80 storage arrays they were able to push over 100,000 IO Operations per second from a single 4-socket server running ESX 3.5. Additionally, they reported the latency numbers which look very good as well.

I find this very interesting because it is in line with results that Kong and I got in our Exchange 2007 on ESX testing we did last year. We found that you could size storage for Exchange VMs the same as you size storage for Exchange on physical servers (up to the 2000 users in a single VM that we tested). Exchange is I/O intensive and is sensitive to any increased latencies. So while we didn't test a workload of 100,000 IOPS, we did see good storage performance with an application that is very sensitive to these kinds of things.

The other thing that I like about this test is that it is cool. As evidence I point to the cool graphs they included and the 495 15K RPM disks that were used.

Todd

0 Comments Permalink
0

On the TechTuesday chat yesterday the topic was Selecting a Server for Virtualization, and we had an excellent discussion. One of the points that came up was the importance of understanding VMotion compatibility between different servers. On easy rule is that it is not possible to VMotion between AMD and Intel processors. As these are completely different processors in many respects it is easy to understand how it is not possible to move a running VM from one to the other with no downtime. The next aspect that VMotion will not work across different generations of processors. The difficulty here is that processor generations do not always line up with Dell server generations. So even if you have all Dell 9G servers, there are cases where VMotion will not work.

It really comes down to the instruction set that each processors is using. A running VM has identified the processor that it is running on and is expecting a certain instruction set to be available. If this were to suddenly change in the middle the OS would not be able to cope or adapt and would most likely crash. There has been some work done to improve the situation - but as of today it is still an issue.

The answer to this problem is the compatibility matrix that our virtualization engineering team has put together and updates as new servers are released. It is important to consider this matrix when selecting a server if VMotion is in use.

Todd

0 Comments Permalink
0

Bumping Into It on VMTN

Posted by ToddMuirhead May 20, 2008

I just ran into a thread on VMTN that is exactly what we are hosting a chat on later today. The VMTN thread is titled Physical Hardware Recommendation, but I am calling the same topic Selecting a Virtualization Server. Very coincidental that this thread was started today and it even specifically asks about 2950s and R900s. So for the record - this VMTN thread was not a planted thread by me or some secret Dell conspiracy. I do have to confess that it was Scott that spotted this thread first and he deserves the credit (Again!).

Todd

0 Comments Permalink
0

The Best Server for You

Posted by ToddMuirhead May 19, 2008

For the past three weeks, we at the Dell TechCenter have been focused on the decoder ring for systems management. Starting today we are going to tell you what the best server is for virtualization. The reason it is going to take three weeks is that the answer for everybody is different. So we are going to talk about key factors, advantages of one type of server over another, and learn from the decisions and thought processes of each other. We are simply hosting the conversation and I do not have a "favorite server" -- although I must admit that I used to lean heavily towards 2-socket servers.

The decision for most seems to come down to 2-socket, 4-socket, or blades servers. I hope that we end up expanding the conversation and talk about lots of other possibilities including storage options, hypervisor options, and who knows what else.

In order to get things going are going to have some chat sessions, there is a topic home page, and a server selection matrix page to specifically lay out the facts about each type of server. I've started the page off with 2-socket, 4-socket, and blades as server categories with some basic tech specs and advantages for each type. This page will grow as additional pros, cons, and others ideas come up.

Todd

0 Comments Permalink
0

New VMmark benchmark results highlight that the best performance for a single server is a 4-socket (or 16 cores with Quad-Core processors). This means that the R900 and R905 are at the top of the list, with the Intel based R900 slightly ahead of the AMD based R905. So if you absolutely need the most performance possible from a single server - then this is the way to go.

A little bit of analysis could lead you to believe that the 2-socket (or 8 core) servers are actually better performing. The VMmark score for the 4-socket R900 is 14.23 with 10 tiles, but the VMark score for the 2-socket 2950 III is 8.47 with 6 tiles. So on a per socket basis, the two-socket 2950 III is actually providing more performance. The same holds true for the two-socket M600 blade and two-socket R805 2U server.

There is much more to choosing a server than the results of a single benchmark, but I think that these results are fair barometers of performance. In general performance does not scale in a linear fashion when moving from 2-socket servers to 4-socket servers, and this seems to hold true with virtualization.

Todd

0 Comments Permalink
0

With the release of the 2U, 2-socket R805 and the 4U, 4-socket R905 (both AMD based servers) and the already existing 4U, 4-socket R900 (Intel based server) now is a really good time to be looking for a server to run your virtualization farm on. All three are excellent choices and could be the right server. To determine which one depends on lots of things. I'm looking for reasons to choose one over the other (let me know your thoughts). We will be putting together some information to help people decide here on TechCenter and hosting some chat sessions to enable discussion.

Specifically, I ran across this article from The Register that I think has a pretty good take on the R805/R905 Virtualization announcement from a couple of days ago. They point out all of the server options as well as the virtualizaiton services that were a part of the announcement on Tues.

Todd

0 Comments Permalink
0

The R805 is Here

Posted by ToddMuirhead May 8, 2008

I leave for vacation for a few days and all kinds of cool stuff is announced (although I did make it into the office a couple of times while on the beach). What I think is a really great server for virtualization is now on the website where you can configure and order it. The new PowerEdge r805 is not just a new version of the 2950 or 2970. It was designed with virtualization in mind and has the features to back it up. It has 16 memory DIMM slots, 4 embedded NICs, plus four more PCI slots, and is still only 2U.

I put together a couple of configurations on the website and the price for an r805 system with two top bin quad-core processors and 64 GB of RAM seems like a great deal to me. I guess that I'm still thinking about some of the old PowerEdge 8450 8-way servers that I tested with back in the day with a list price of over 50K, which was a good deal at the time.

Todd

0 Comments Permalink