Increasing Availability and Throughput for VMware

This is a live update of VMworld 2010 session TA8133-
Increasing Availability and Throughput for VMware

Chad Sakac, EMC
Vaughn Stewart, NetApp

Chad and Vaughn started out by mentioning that they wanted the presentation to be non-vendor related, and basic blocking and tackling for VMware and storage. Kudos to them…

Key theme (no surprise)- design is critical to performance and availability. It is important to understand the relationship between performance and availability. How does this change as the environment grows, and what is the impact of a storage outage,planned or otherwise?

As for protocols, they all have strengths and weaknesses, and all the VMware features work across all protocols. Most customers use multiple protocols- start with what works for your environment and where your company is on their virtualization journey.

From personal experience, I have found that as customers grow in their expertise and maturity with virtualization, no matter which storage protocol they started with, they implement additional protocols for easier management or better performance.

Whatever your choice, follow the recommendation of your storage vendor.

What about file system alignment? File system misalignment basically refers to when the file system (VMFS, NTFS, etc) does not line up correctly with the chunks or blocks on the storage. This affects every protocol and every storage control,controller and every vendor. While it may not create a great deal of degradation on a single vm, it will on hundreds, since it imposes a drag on the controller.

How to fix this? First of all, Best practices from EMC, NetApp, and VMW all agree you should align your partitions. Beginning with Windows Vista, MSFT introduced a 1MB starting alignment for NTFS (this aligns optimally wi all storage vendors array technology). vCenter 3 and later will automatically align the VMFS. What about deploying older operating systems?
Well, you will have to use a tool to align the guest partition (you should probably do this as part of the P2V, otherwise you will have to take the vm down later).

Protocol considerations:

iSCSI

Well, beginning with vSphere 4, you can have multiple iScsi sessions per target. VMworld session TA7743 drills into this topic in detail. ‘Nuf said..

EMC Clariion Caveats-

Flare 29 and earlier- only allows one iSCSI session per host (even if ESX wa trying to open multiple). Workaround- put multiple vmkernel NICs on different VLANs. Flare 30 fixes this, which might reduce your admin work.

Also, on 10GB Ethernet Clariion has a known issue that might impose slower writes than expected. (NetApp does as well). There are patches for this (Flare 29/30 patch3), but you can expect 2.5 Gbps per port, and up to 16 Gbps on a high end unit.

Block Multipathing considerations

Asymmetric Logical Unit Access- allows for paths to be profiled, allowing for the optimal path to be selected by vSphere PSP and third party MPPs. Essentially, the optimized path is adverted to the ESX host and displayed in the vCenter GUI. requires vSphere 4 or later.

How to enable ALUA?

For existing storage, you will need to craeate a new LUN mask leveraging ALUA on the storage controller. Put your VI3 host in maintenance mode, migrate to a new storage path, and then upgrade to vSphere. I will try to drill down on ALUA more later as well.

MPIO

Beginning with vSphere, VMW introduced a pluggable storage architecture. This allows for different MPIO or PSP modules, including third party multipathing agents.

Out of the box in VMW, you can use MRU (all traffic goes down a given path, if fails then it cuts over to another path. Even if the original returns, it remains on the MRU!) Fixed uses a preferred path, so that if a path fails and comes back, the fixed path will return to its preferred path.

Round Robin in vSphere is the preferred using ALUA… In this case, all traffic is distributed across all available paths, using preferred optimal paths.

PowerPath is a multipathing plugin (MPP). Beginning with vSphere 4.1 Enterprise Plus licenses are no longer required. Chad put some recommendations for using PowerPathVE up on the slide, but it went by too fast. I will link to it once the deck is posted. He also mentioned that EMC’s VSI plugin will take care of the storage settings for the VI admin.

NetApp MPIO settings

Virtual Console plugin will take care of all the settings for all protocols.

NFS Considerations

Of course, start wi the vendor best practices, use the plugins… Use multiple NFS data stores and 10GbE. NFSv3 requires 1ip address for each export, and only one port will be active. The workaround is to create multiple 1GbE paths.

Using a traditional Ethernet switch (where each operates independently of the other), you need multiple IP networks across multiple switches. Using more modern switches, use multiswitch link aggregation. Also, use multiple VMKernel ports and ip subnets, the ESX routing table, and assign multiple sequential IP addresses with your storage.

Jumbo frames? Keep it simple, don’t worry about it. Hardware fast enough these days you don’t need it.

Flow control? Yes, turn it on, follow the recommendations in the documentation from your vendor(s).

Things to be aware of – while iSCSI is limited to specific vmkernel nic ports, NFS is not. NFS can go out over any NIC port.

summary-

– take the time to understand the design implications, whether you are the storage admin or the VI admin.
– use 10GbE!!!
– engage all the teams involved.

– Posted using BlogPress…

Location:SW Temple,Salt Lake City,United States

Advertisements

About Tom

Just a guy, mostly a father and husband, pretending to be a technologist, constantly underestimating how much there is to learn.... content is mine
This entry was posted in TechInfo, Virtualization - Server, VMworld and tagged , , , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s