Author Topic: how wanos datastore built and synced between each other  (Read 4687 times)

peter

  • Contributor
  • Full Member
  • *****
  • Posts: 51
    • View Profile
how wanos datastore built and synced between each other
« on: March 01, 2015, 03:57:09 PM »
Case:

Simple bridge deployment of wanos does not require ip on each wanos appliance.

Question:

1) How does datastore on each site get built and synced for deduplication ?
 
2) How long is datastore time to live or gets refreshed for better tuning

Thanks

ahenning

  • Team Wanos
  • Administrator
  • Full Member
  • *****
  • Posts: 629
    • View Profile
Re: how wanos datastore built and synced between each other
« Reply #1 on: March 01, 2015, 06:06:36 PM »
Hi Peter,

1) Since the traffic is inline and must be for optimization to work, each side keeps track of the database. Sync messages happens within encapsulated packets (if needed). If there is no packetloss, theoretically syncing should be zero to minimal.

2) Data in the store has no expiry time. When the datastore is 100% it runs at peak efficiency. In other words no more room for more data. At this point oldest data is swapped out in a first in first out queue.
CCIE RS, CCIE SP, Mnet&sys

Note: Forum posts may be outdated. Please see the latest documentation at wanos.co/docs

peter

  • Contributor
  • Full Member
  • *****
  • Posts: 51
    • View Profile
Re: how wanos datastore built and synced between each other
« Reply #2 on: March 02, 2015, 03:29:33 PM »
Since datastore is only around 13G for free version, generally when will datastore  reach 100% and swapping out happen (after how much data(GB) pass through the wanos appliance).

Thanks

ahenning

  • Team Wanos
  • Administrator
  • Full Member
  • *****
  • Posts: 629
    • View Profile
Re: how wanos datastore built and synced between each other
« Reply #3 on: March 02, 2015, 03:56:56 PM »
Express has a limited number of remote peers, which limits the total datastore size. The datastore is 13GB per peer. For each peer FIFO happens when the datastore is full. In other words when the peer datastore is full, new traffic will swap out old traffic data.

Size of the database is actually not crucial. A 4GB datastore can provide better compression than a 40GB store. If the 40GB has 400 x 100 Mb files that can be matched, optimization is going to be close to 0% because it is unlikely that these files will be transferred multiple times. Now if the 4GB had 32 Million 128 byte patterns, then there are constantly hits.
CCIE RS, CCIE SP, Mnet&sys

Note: Forum posts may be outdated. Please see the latest documentation at wanos.co/docs