The future of IT?

16 09 2009

Well, how about this?  The GSA has started an “app store” for U.S. government image agencies.  This is not your everyday, fart app.  No, these are major applications like Salesforce offered up in the cloud.  The agency picks what they want, how many users, etc., and it gets built.  I think this is where true virtualization really kicks off.  All I have to do is pick my application that I want, how many people are going to use it, and maybe some SLA’s, and I am off and running.  Does the agency care about how much storage or redundancy is required on the back end?  Nope, they just order it…


VMWorld 2009 Session Notes: Virtualizing Exchange 2007

2 09 2009

Exchange 2007 on vSphere 4…
1. 2K7 has reduced IO with larger cache, etc. This makes it easier to virtualize.
2. Must run it on server 2008 to get MS support.
3. Use 1vcpu 4.5 GB ram, and 500 heavy users as a building block.
4. Use 2vCPU 7 GB RAM, and 1000 heavy users as a larger building block.
5. Only 10 ms send latency enabling FT.
6. Disaster recovery…ha, FT, lcr and scr does not require microsoft clustering.

VMWorld 2009 Session Notes: vCenter AppSpeed

1 09 2009

vCenter AppSpeed
1. Monitors apps performance based on inspecting traffic on the vswitch.
2. You must be able to measure end user performance to troubleshoot it.
3. Appspeed let’s you monitor as you virtualize (assured migration).
4. Appspeed server discovers, maps and monitors performance vs SLA’s. It puts a small vm on each host.
5. Environment map automatically created. Shows interdependencies. This gives you a deeper understanding of how everything is connected.
6. AppSpeed then monitors performance letting you drill down and identify bottlenecks. Alerts automatically. You can trend over time.
7. AppSpeed can then analyze root causes.

VMWorld 2009: VMWare FT Session nodes

1 09 2009

1. Zero downtime if a host drops.
2. Transparent to applications and os.
3. Basic concepts of FT…zero downtime when a host is lost, it will create another duplicate on a third host when the first goes down.
4. Virtual lockstep– all input is recorded and replayed to mirrored instance it is replayed via FT logging.
5. Hardware and software req’s. Latest processors, shared disk needed, hosts must be patched to same level, must be in HA cluster, minimum of 2 nics for ft logging nics. Only one processor (vcpu).
6. You can use the vmware site survey tool to check compliance for FT. You can also use profile compliance on vcenter.
7. HV must be enabled in bios on host.
8. Cannot help with mutiple host outages.
9. Best practices…(avg disk reads bw+avg network input) x 1.2 to get approximate lgging traffic needed. Isolate your traffic for logging. Keep the cpus within 400mhz between hosts. Max of four FT vms per host! Mix and match primary and secondary across hosts. Less than 1ms network latency.

VMWorld 2009 Notes: Availability Solutions and Futures

1 09 2009

Unplanned outages…
1. Expectations of availability are increasing.
2. RTO is decreasing. Less than 12 hours now.
3. Vmware can protect at each level: component, server, storage, and site.
4. HA– you can now do ha maintenance mode, individual vm health monitoring, priority of vm boot up (admission control).
5. How many vms can you fit on your hosts? You can make memory and cpu reservations per slot.
6. If a host fails…set your restart priority.
7. If you experience an isolated response where a host loses network connectivity, you can set it to turn off the vms or leave them on.
8. Vm monitoring let’s you reset vms if a guest blue screens. You can set the sensitivity of this. It uses a heartbeat to vmware tools to do this. You can set max resets.
9. VMWare FT. Vms run in lockstep, mirroring everything. Uses common disk. Fails over to shadow copy. You need a ft logging nic on each host. You can’t snapshot with FT on!!! What if I want to back it up?
10. Futures…protection for single component failures (like a san, or network), pre-empt failure (host retirement), stretched clusters let’s you set affinity on hosts to a certain site,
11. Netapp metrocluster can use FT to keep everything running across sites.
12. Futures…can dive deeper into the vm at application level to restart services etc based on application level things.
13. Host retirement. Can monitor things that can cause failure, such as hardware cpu and mem, and if it fails, it can put host into maintenance mode.

vStorage Storage for Cloud OS

1 09 2009

Storage Integrations for Cloud OS

VSphere can now handle 300,000 IOPS.
Can handle 3X the web traffic of ebay with one host.
Storage is key to encapsulating vms. Everything about the vm is stored in a few files.
Storage vmotion is now built into the gui and vcenter.
Storage vm creates a new folder, and creates a new file, then creates a changed block tracking bitmap, then copies the vmdk, adds the modified bitmap to the target and makes it active.
Linked clones let’s you use one copy of a vm and use it with multiple vms that can add changes to each vm to make them unique.
Disk options– VMFS, RDM, and NFS.
Things added to vstorage–thin provisioning, fcoe, vmfs volume grow, paravirtualized scsi adapter, enhanced storage vmotion, better management visibility, vstorage apis.
Vstorage apis add multipathing, array based storage vmotions, provisioning from a template, snapshots integration right on the san.
In the future, they will be able to pause a vm when it runs out of space, can accelerate lock intensive operations.
In order to get all these into the external cloud, is being able to integrate all these storage things to send it all external.

VMWorld 2009 Lab notes:VSphere Advanced Topics

31 08 2009

New features, best of, advanced features…
1. You can link virtual centers as one.
2. Licensing is now via just a key.
3. Look and feel is changed. It is much more icon driven.
4. Host profiles have been added. You can take a baseline of a host, and then deploy the profile.
5. You can now create a distributed virtual switch. Network state and vlans stay with the vm as it is vmotioned. You can only get this with enterprise plus.