Why
do you and how to cook it
In the first post we mentioned about CloudLITE
virtual store VMware Virtual SAN (VSAN). Today, we focus on the technology and
more will tell you what to consider when creating a VSAN for the project.
Why
do you VSAN
Traditional architecture consists of three key
components:
1. Servers,
2. Storage
system (DSS),
3. Storage
Area Network that connects storage to servers through a block (FC, FCoE, ISCSI)
or file protocols (NFS, SMB) using the appropriate switches.
To manage this economy need three different
interface, three different competences and, in an amicable way, three different
specialist. The deployment of this
architecture takes a long time, and rapid scaling is also quite a trivial task.
If your project involves predictable and systematic scale, to add a new store
has a week, and in the state there are experts who will be engaged in the
design, the traditional architecture - your choice. When you have a project (for example, public
cloud) is growing leaps, adding a new repository with minimal automation
capabilities it will take a lot of time.
This is where it comes in a converged architecture
VSAN, which allows to combine computing functions in the server and storage
functions.
VSAN
as a converged solution
The convergent solution enables you to create the
infrastructure of the typical units that combine several functions (e.g.,
computing, storage). Management of such infrastructure through a single
interface, and scalability - by adding a block.
In the case of VSAN each unit - a server. Not all, of course, but more
on that later. How the server does performs the functions of storage? VSAN
collects from local drives virtual servers "external" storage
available to all computing nodes of the cluster virtualization. In this program
part of VSAN runs on the same servers as the compute nodes. Thus, on one and
the same machine and arranged calculator (compute node), and a part of the
storage system (storage node) - all in one vial.
How
it works
Each server has 1 to 5 disk groups. In each group -
at least one SSD-drive (a necessary condition for building VSAN) and from 1 to
7 HDD-drives.
SSD-drives in the disk group is made up of a
common pool of data caching. VSAN first reads data from the cache; if data is
not in cache, VSAN is sent to the HDD-drives.
For each virtual machine you can configure your FTT (failures to
tolerate). The default is 2, ie. E., all data are written once virtual machines
on 2 different server cluster. If one server fails, we will have a synchronous
replica on another, and all the I/O operation will go to the second copy.
What
to look for when designing VSAN
Relative ease of deployment does not negate careful
design architecture VSAN. Here are a few points on which we would like to
dwell:
1. Compatible
hardware. Although VSAN and gives a certain freedom in the choice of
"iron", it is wise to stay within the list to ensure compatibility
with VMware VSAN equipment. So you do not have an educated bet select
compatible controllers, adapters and so on. In the case of CloudLITE on set of
technical and economic parameters, we chose Huawei FusionServer RH5885 V3. This
model has on board more efficient PCIe flash card (in comparison with the
already became a classic SSD-drives), which, by the way, saves "slots for
drives" and create more disk groups. In the near future it will arrange
unboxing. Stay tuned :).
2. Network. In the configuration of a VSAN VM can work in
one place, and stored - in the other. It makes quite high demands on the
network: you must be at least 10 GB network.
3. Performance
disk controllers. The disk controller must provide a buffer to surround a large
queue. Load it is significant: the controller will give the data necessary not
only to that server, but to the entire cluster. For example, the reduction of
the disposed disk group to the new group you want to record large amounts of
data in a short time. Recording speed just will depend on the performance of
the controller.
4. The
volume of discs. In this situation no longer is better. Quite the contrary.
Although currently available disks 4, 6 TB, VSAN is better to build from 1TB
drives. Let's imagine an emergency situation when the cache we do not gets
(replacing "a melted" disk group, backup or vosstanavleniju backup):
6 TB drives will recover up to 6 times longer than the 1 TB drives (if
ottalkivatcya the ratio of the speed of reading to the volume Stored data -
IOPS / GB). Here we are, of course, talking about the worst case, but the
situation is not out of the realm of fantasy. And the desire to use VSAN volume
wheels completely fell off, just imagine how many will recover the data on the
hard disk 7 to 6 TB.
5. The
ratio of the volume of SSD to the hard disk. It will directly affect the final
performance of the disk group: the higher capacity SSD (the more data is in the
cache), the better the performance. In CloudLITE used for caching PCIe flash
cards - they have less latency compared to the SSD. Incidentally, in the VSAN
6. Supports
disk groups consisting only of SSD. 6. The ratio of computing power disk space.
When designing VSAN should be all carefully bed: calculate the ratio of
processors, memory, and number of disk groups, and calculate how much of the
increase computing power that it was cost effective. When running the decision
can no longer be on the fly add disk space for VSAN (storage node), not adding
a new server, and therefore processor and memory. Alternatively, when the
server is used only as a storage (t. E. Compute node of the server is idle), it
is possible, but uneconomical: it is actually a return to the traditional
configuration and the refusal of the benefits of a converged solution.
