Efficient Transport
Products > Realtime Data Integration > Efficient Transport
HVR's transport is fast because it uses the network efficiently and because the replicated changes hardly touch disk before being applied to the target database.
But replication only has to move changes, not the data itself. For a 10 GB source database the change rate could be just 100 changes per second. If each row is average 100 bytes that means 10Kbyte/sec which is 100 Kbit per second. This would only be 1% of a 10 Mbit line. It could almost manage on a 64 Kbit/sec line.
Also, a network link always has two properties; bandwidth (discussed above) and network round-trip latency. HVR only does a few round trips per second, even if it is moving thousands of changes per second.
HVR can replicate over any slow network. You only need a bigger network if;
HVR's transport is fast because it uses the network efficiently and because the replicated changes hardly touch disk before being applied to the target database.Extreme Network Compression
When moving data across the network, HVR uses powerful compression. These compression ratios are extremely high, because HVR uses tailor-made compression algorithms. These exploit specific information about each replicated change, such as the table width and column datatypes.Example
HVR’s compression ratio varies between 85% to 99%, depending on the data being compressed. It is especially high if there is a lot work queued up and during refresh and compare. For example if the database is 10 GB then refresh and compare needs to move the whole thing across the network, so if the compression ratio was 95% then HVR would need to move 500 MB. On a 10 Mbit line that would take 500 seconds. But HVR has to move the data over the WAN and also load in into the target, which may take longer than 10 minutes (depending on how fast the target machine’s disks are).But replication only has to move changes, not the data itself. For a 10 GB source database the change rate could be just 100 changes per second. If each row is average 100 bytes that means 10Kbyte/sec which is 100 Kbit per second. This would only be 1% of a 10 Mbit line. It could almost manage on a 64 Kbit/sec line.
Also, a network link always has two properties; bandwidth (discussed above) and network round-trip latency. HVR only does a few round trips per second, even if it is moving thousands of changes per second.
HVR can replicate over any slow network. You only need a bigger network if;
- HVR needs to refresh or compare, OR
- End-users need to login to the disaster recovery database in an emergency.
Avoid Saturating Network Bandwidth
HVR also has ‘network throttle’ feature. This lets HVR be configured to never use more than a certain percentage of the bandwidth of network link, so that response times for end-users always conform to Service Level Agreements.Minimal Replication Hops
Another explanation for HVR's speed is that it minimizes ‘replication hops’ using its streaming architecture. A replication hop is when changes touch disk between the capture database and replication target. Multiple hops incur disk I/O which adds to overhead. HVR does need to queue changes on disk, so the system can keep running if one target database is offline. But HVR only does this once and the queues are highly compressed. So for every megabyte of captured data HVR will perhaps use a 100 kilobyte hop.Network Encryption
HVR also has Secure Socket Layer (SSL) to encrypt data send over a public network.Customers
Ticker tape
News
30/11/2009 • HVR Software; new name
PSB High Volume Replication announces name change to HVR Software.
30/10/2009 • Ingres reseller agreement
HVR Software Adds Realtime Data Integration Capability to Ingres Database
...
06/10/2009 • Windows Cluster Awareness
HVR for Windows on 32-bit and 64-bit Intel platforms is now Windows Cluster Aware.