Votos - 0, Puntuación media: 0
(
)
)
|
Otros manuales para este modelo:
Resumen del manual
• Bandwidth – one of the key aspects to replication is a good understanding of bandwidth on
the intersite link. If your arrays are within the same data center, building, site location, or city
replication is possible over the SAN Fibre Channel link, which typically supports 1, 2, or 4
Gb/sec. For greater distances, FCIP may be necessary. It is also important to understand the
link between your two sites. Do you have full access on that link? Perhaps your department is
limited to 20% utilization. Is there other traffic on that link? Possibly another application is
already using 50% of the capacity.
• Latency – round-trip delay on the replication link is also a critical factor. Lower latencies are of
course more desirable. A 1 to 2 ms latency is usually possible only over Fibre Channel. Such
low latency makes synchronous replication solutions appropriate. Higher latencies might drive
a customer towards an asynchronous solution. For example, 100 ms latency would bring most
applications almost to a stand-still in synchronous replication. In such a situation, the customer
might prefer asynchronous solutions. Some applications are very sensitive to the network latency
and have recommended maximum thresholds.
• Workload – You must get a clear picture of your application workload before planning a
replication solution. This is typical y done over a minimum of 48 hours during heavy usage, such
as end-of-the-month reporting. When planning for replication, you need look only at the writes
per second, since read requests are not replicated. Also, characterize your workload in terms of
high and low volumes over time. When are the peaks in the workload? How high do they get?
When are the valleys (downtime) in your workload, and how long do they last?
2 Choose the correct replication technology
The EVA series supports three types of replication: synchronous, asynchronous, and enhanced
asynchronous. The first-generation EVAs (3000/5000) support synchronous and asynchronous
replication (the lat er also referred to as standard asynchronous replication). The next-generation
EVAs (4x00/6x00/8x00) support both synchronous replication and a new feature known as
enhanced asynchronous replication.
• During synchronous replication, the array acknowledges input/output (I/O) completion after the
data is cached on both the local and remote arrays. When the response is received from the
destination array, the source array sends an acknowledgement to the host server. Only then can
the host send the next I/O request.
• During standard asynchronous replication, the array acknowledges I/O completion when data is
cached on the local array but before it is replicated to the remote array. When the bandwidth is
available, the data is replicated over the link to the destination array in the order it was received;
otherwise, it queues up in the cache of the source array. Up to 64-KB or 128-KB I/Os can be
queued in the cache, depending on EVA firmware version.
• Enhanced asynchronous replication was introduced with the EVA4x00/6x00/8x00 series.
Enhanced asynchronous replication uses a write history log created on disk, which makes it
possible to support a greater number of requests than standard asynchronous replication. The
incoming requests are writ en to cache and the write history log and then are acknowledged
back to the host server.
• Synchronous replication provides a consistent state between the source and destination LUNs and
can support an RPO of zero when in failsafe mode; however, the performance impact on the
server application can be significant if the replication link has high latencies.
• Asynchronous replication methods have minimal impact on the performance of the host
application, since replication takes place in the background. Its drawback is the potential for
data loss. If the replication requests are queuing up on the source array and a disaster strikes
during that time, then all unreplicated data is lost, possibly including commit ed transactions on
the primary array. Conversely, if the link latency is low enough and the bandwidth sufficient,
asynchronous replication will be able to keep up with the data workload, resulting in a near-zero
RPO.
2
...Otros modelos de este manual:
datos de dispositivos de procesamiento - HP EVA4000 Starter Kit (115.65 kb)
datos de dispositivos de procesamiento - HP EVA4000/6000/8000 Disk Arrays (115.65 kb)
datos de dispositivos de procesamiento - HP EVA4100 SAN Starter Kit (115.65 kb)
datos de dispositivos de procesamiento - HP EVA4100 Starter Kit (115.65 kb)