Quantcast
Channel: Datacentre Management . org » storage hardware
Viewing all articles
Browse latest Browse all 6

Server virtualization storage opening metrics: Which ones matter?

0
0

When it comes to assessing and monitoring how good a storage systems that support your server
virtualization sourroundings are doing around storage
performance metrics
, there are many layers to consider.

Storage for practical servers can be served adult by a hypervisor or served adult from block- or
file
-based storage inclination over a network connection. When storage is served adult by a hypervisor, the
hypervisor controls a entrance and as such is a some-more engaging of a dual given what impacts
the hypervisor impacts all regulating on a hypervisor; metrics—as good as bargain how
they impact a altogether workloads on a hypervisor— matter.

Storage served adult by a hypervisor looks only like a SCSI device to a practical machine, while
network-served storage might need specialized drivers, such as iSCSI.

Hypervisor-served storage can be in a form of Fibre


Channel, iSCSI, NFS (and,
in a box of Hyper-V, CIFS)
or internal storage, though by a time a practical appurtenance attaches to a storage device, it acts just
like a normal SCSI device and
therefore uses a normal SCSI motorist from within a guest handling system. The hypervisor uses
binary interpretation of a customary practical appurtenance SCSI motorist commands into those that can be
handled by a other technologies, possibly that is Fibre Channel, iSCSI, NFS or a internal SCSI
device. Binary interpretation happens possibly within a hypervisor or within a CPU regulating Intel VT-x
or AMD RVI authority structures. In possibly case, a practical appurtenance sees a storage as SCSI, while
the hypervisor sees a storage as something else entirely.

Metrics and how to appreciate them

Given these dynamics, there are mixed sets of metrics to cruise associated to virtualization
storage
:

  • Storage opening metrics seen by a guest
    operating system
  • Storage opening metrics seen by a hypervisor
  • Storage opening metrics seen by a storage hardware

Each set of metrics is critical for specific reasons, though we can’t count on them to always
guide we to a right decision. Some of these metrics could tell an untruth and therefore by
untrustworthy.

The slightest profitable metric is a information seen by a guest handling system, given the
virtual appurtenance does not indispensably accept full CPU cycles, in that box a information within the
virtual appurtenance is suspect. The metrics involving CPU cycles for practical machines are always
suspect given a practical appurtenance might or might not accept full CPU cycles. But, some virtual
machine metrics don’t have anything to do with CPU cycles, and those non-CPU VM metrics are
trustworthy.

Storage opening metrics seen by a hypervisor are valuable, and a many frequently viewed,
but these can also be dubious as they’re formed on information that could be cached or queued by the
hypervisor.

That leaves metrics as seen by a storage hardware, and they’re a best ones to use, given the
hardware covering provides fine-grained data, down to a shaft being used. In many cases, this data
is a same as seen by a hypervisor, though in rarely implicit subsystems, this might not be a case.
Unfortunately, not all hardware subsystems concede this information to be seen, in that box we need to
concentrate on a metrics accessible to a hypervisor given metrics from within a VM are not
completely reliable.

From a virtualization storage hardware layer, a many critical metrics are a review and
write latency values, or how prolonged a information took to be review from a hoop or created to a disk
once perceived by a specific layer. Next adult in significance is a series of IOPS. You can't only
look during  IOPS though bargain a review and write kilobits per second, or Kbps. IOPS
refers to a operations; Kbps refers to a tangible information review or created by a system. IOPS is the
most-looked-at metric possibly from a hypervisor or storage device. However, latency is a better
metric given it indicates possibly there are issues with a storage. The IOPS series varies with
the series of blocks to be created and, with NFS (and CIFS), can’t be simply tracked given latency
metrics are not fundamental within a protocol.

Which collection can do a job?

To accumulate all these storage opening metrics, demeanour to collection from companies such as NetApp
(Balance), SolarWinds (Storage Manager), Quest (vFoglight Storage) and others that speak directly to
the hardware. Such products can check a storage hardware covering regulating a Storage Management Initiative
Specification
(SMI-S) program covering or directly by storage manufacturer protocols.

For a hypervisor level, collection such as VMware vCenter Operations, vKernel, VMturbo and Quest
vFoglight, to name a few, check a hypervisor covering by querying a hypervisor directly or
indirectly regulating a hypervisor’s executive government console (such as VMware vCenter).

Finally, guest handling systems yield their possess collection for entertainment storage performance
metrics.

A mixed of SMI-S and hypervisor-based collection yield a best brew of functionality for
determining a latency, IOPS and bytes created or read. That’s especially given a numbers produced
by these collection tend to be in sync unless a hypervisor is intensely busy. In that case, the
hardware numbers are best for pristine storage performance, though given all resources hold one another
within a hypervisor, ensuring that a hypervisor does not strech that impassioned bustling state is
always a good thing.

If a storage-level collection do not exist for your environment, such as with some iSCSI servers,
the hypervisor-based collection turn a many important.

Edward L. Haletky is a author of mixed books about a VMware virtualization platform
and a CEO of AstroArch Consulting Inc. and The Virtualization Practice.

This was initial published in Sep 2011

Article source: http://www.pheedcontent.com/click.phdo?i=70f246f7ded68dcf6328d4deae3d3122


Viewing all articles
Browse latest Browse all 6

Latest Images

Trending Articles





Latest Images