Demitasse

I’m on the flight back from Las Vegas and trying to record some sense of what VMworld was for me.  I went to be part of a community event, as much as a technical event. I certainly found a community in the vExperts, every one I met was friendly and welcoming, all there to help out other community members.  I met way too many cool people to name, apart from the biggest names, I enjoyed meeting some contributors like Cody Bunch, Calvin Zito, Greg Shulz and Sean Crookston who were all very modest about their contribution.  I also got to meet back up with people like Andre Leibovici and Geoff O’Connor that I know from past vForums and a few Kiwi’s from vBeers and past courses, Paul, David, Chithirin, Jamie and the rest.

Of course there was a day long workshop for VMware Certified Instructors where I met more instructors in five minutes than I have met in five years of teaching.  And then there were the random people in lines or at parties. I tried to talk to people around me whenever there was a queue and had a lot of good chats to people from all over, Roy from Texas was a standout as a great guy.
Parties, there are a few at VMworld, ranging from a few hundred people to thousands.  I like to meet people and talk at a party, so the main party and another I went to in a noisy nightclub weren’t my favorites. I really likes @CXI’s party, no music and lots of interesting people.  I also had the fortune of going to the VMware CTO party, where Stephen Herrod knew my blog! And of course there was the unforgettable VMUnderground party, thanks to Stephen Foskett for getting me into that and sorry to the Kiwi’s I didn’t see at vOdgeball.

I hear there were a heap of technical sessions, I only made a few and didn’t always choose the right ones.  There we also a lot of labs, both the awesome VMware labs and the EMC labs were busy. I made the mistake of trying to do a vShield lab after a few beers at the nightclub party, not a wise choice.

Finally there was the location, Las Vegas, not my kind of place. I realized just how much I like peace and quiet.  The strip is exactly as it appears on TV only ten times more so, bright lights and huge buildings.  What you don’t realize from TV is that the casinos are huge. The Cosmopolitan, where I stayed, is 45 floors high. That’s taller than the largest office building in Auckland, and it doesn’t tower above the other casinos.  I wanted to go to Las Vegas once in my life & now I have. I don’t expect to go back to the strip again but there is plenty more to Nevada if I happen to be passing with some time to spare.

VMworld was awesome, I’m already thinking about how to get to 2012, which will be back in San Francisco. I think I’ll like San Fran better, not least because good coffee is easier to find there.

Finally a huge thanks to John Troyer for helping make it possible for me to be at VMworld, for creating the whole vExpert community and for fostering the wider VMware Community.  The community is a fabulous thing and all who are involved appreciate the hard work you and the team put in.

What would the impact be if VMware went all in with vRAM pricing?
What if you didn’t buy ESX sockets, just vRAM?
What if you didn’t buy ESX editions, just vRAM?

Customers are unhappy with the idea of buying extra sockets of CPU capacity just to get vRAM, even though very few actually need to buy them. Some customers are also reluctant to buy the top stock code of vSphere as it costs more than the lower codes. It certainly means that a lot of customers in the very cost conscious New Zealand market don’t get to use some of the most valuable features of vSphere. So how about not having feature differentiated editions of vSphere? Every ESX server gets the feature set of Enterprise Plus, even the smallest customer can use DRS and host profiles, rather than buying oversize ESX servers and avoiding server rebuilds because it’s a manual process.

I still see a need for some price differentiation, small to medium business has a much lower budget than Enterprise. A limited edition of vCentre (like the current Essentials) which only supports three hosts and a hard limit on vRAM would suit. Enterprise customers would want more scalability but apart from scalability the feature set would be the same. Then maybe two stock codes for vSphere, an 8GB vRAM for Essentials and a 32GB vRAM for Enterprise, again all features in all stock codes.

If you are as old as me and followed the same career progression as me, then cast your mind back to when Citrix moved from per server pricing to per user. I know it was a long time ago but it’s a similar thing. suddenly there was no penalty to standing up a new server for a low utilization workload or at another location. Suddenly a whole lot of use cases for Metaframe were workable. The result was more use of Metaframe and more profit for Citrix.

Would this be fairer and enable more use of vSphere without taking away VMware’s profitability?

Andre Leibovici has posted the second part of his linked clone IO study and it makes interesting reading.  The first part is here and the second here they are well worth the read.

One of the things Andre calls out is that having Windows cache the application files saves a lot of IO on the replica and that with sufficient memory in the VM windows doesn’t touch it’s page file.  This highlights a trend, giving VMs lots of RAM reduces disk IO load. 

This is true for server workloads as well as desktop, this post on the VROOM! blog calls out Exchange 2010 as benefiting from more RAM and I’ve had recent discussions about database applications on Linux that benefit from large amounts of OS RAM as disk cache.  For the Linux VM halving the configured RAM on the VM multiplied the IO rate by three, a change that was quickly backed out.  The message was repeated in the podcast I recorded with Michael Webster where we talked about virtualising Tier One applications.

When I started “doing” virtualisation it was all about how much you could do with how little, but that doesn’t always answer as a way to work.  Some workloads need a lot of RAM to do a lot of work.  The RAM is often cheaper than buying a huge amount of spindles of disk or expensive SSD and cache for the SAN.

Newly posted podcast talking with Michael Webster.

Live from Vegas baby.

All the details are on the podcast web site.

Years ago one of my managers said she treated her technical teams like a tribe, they need to meet together and have open forums to discuss technical things.  Often infrastructure geeks can’t discuss their work with the people they love without the loved ones falling asleep or telling us to shut up.

The vBeers movement aims to bring the  virtualisation tribe together for these forums.  I’ve had the pleasure of starting this in my home country of New Zealand with vBeers events in both Auckland and Wellington.  Each event attracted about a dozen people who “do” virtualisation and like talking to others who do the same.

I plan to keep organising these events, probably every couple of months in each city.  The hard part is finding a date when I’m sure I’ll be in town.  What this blog, follow @DemitasseNZ on twitter or watch the vBeers web site for updates.

Mike Laverick will return to the podcast to discuss Site Recovery Manager 5.0 new features.

All the details are at the Podcast blog

The podcast with Frank and Duncan is also, rather belatedly, up on Talkshoe now.

The week after next I will be teaching the vSphere Advanced FastTrack in Auckland.  As far as I’m aware this is the first time the course has been run in NZ.

This FastTrack course is made up of the Manage for Performance course and the VMware vSphere: Troubleshooting course combined into an extended hours five day course.

This course suits virtualisation professionals who are preparing for the VCAP-DCA exam or students with plenty of vSphere experience who wants to sit the VCP exam.

There are currently only three seats left on the course so if you’d like to join me there it’s time to book.

In keeping with this year’s theme of View being a big part of my work I spent some time with a customer this week, looking at a view implementation.  In emails with the customer an odd thing had come up, the View Composer Replica disks were being provisioned fully allocated, i.e. 30GB rather than the 10GB size of the data inside the VM.

You may want to revisit how View Composer disks work. You also may recall an earlier post about replica disks being bloated because deleted files remain on the disk, that was not the cause.  And you should definitely look at Duncan’s article about different datamovers used to copy disks on ESX.

The issue my customer had was that their Master VM (the one that is copied to the replica) was on a VAAI enabled array.  The same array holds the SSD backed datastore where the replica’s are held.  Since the two datastores both have a block size of 8MB the hardware offload datamover is used.  This datamover does not reclaim zeroed blocks in the copy, so the replica is the same size as the master, 30GB.  This was not desirable as a lot of expensive SSD was holding empty blocks.

The quick fix was to move the master to a datastore not on the VAAI array, in this case the local datastore in the host.   Since this has proved that the datamover was the cause of the issue (and that neither I nor the client was going mad) we will have to reformat one of the datastores with a different block size.  This will force the use of the zero reclaiming datamover.

I’d like to stress that VAAI is great for View Composer, hardware offload of locking makes VMFS perform better for the growing delta disks.  Do not disable VAAI to resolve this issue.  Also VAAI is great for storage copies and clones, but it behaves differently to non-VAAI when it comes to zero reclaim.  The solution is to have the Master VMs on datastores with different block size to the datastores that contain replicas.

My customer was using tiered storage, so only one replica per pool.  Without tiered storage there is one replica per datastore which makes fully allocated replicas far more painful.

Another side note is that the SSD backed datastore must have enough free space for the full size master VM in order to provision the replica, despite the fact that the replica is smaller than the master.  This is because before the clone we don’t know how much space will be saved by thin provisioning the replica.

If you are in Wellington, mark your calendar. vBeers in Wellington on Thursday 11 August.

http://www.vbeers.org/2011/07/28/vbeers-wellington-new-zealand-thursday-11-august/