Progress reports, of sorts
Some images from the recent progress of the work towards the “2.0” implementation of the cluster. Things are in many aspects of the matter coming along nicely, yet there is so much work to do. Getting the new generation of the software working for the completion of the hardware would be really nice, but at the present moment that seems but a faint idea and lucid dream.
Front shot, 128GB ram, with poor mans 8GB sticks utilized, and still only half-populated. With 32GB sticks this thing supports 1TB, that is a long ways away though.. Adding that quantity will be a project for next year, I think.. anyway, it won’t happen today 🙂
New firewall! Turns out its actually not entirely easy to just push a couple of gigabits reliably from WAN.. While keeping the internal routing to a minimum, a few network cards are needed for CARP and other miscellaneous. A few upgrades are planned down the line, it will be interesting to see how it all turns out in the end.
An entirely overexposed shot of the server racks.. A little before-shot, as the network had to be almost entirely redone. Left hand rack is mostly processing, while right hand rack deals mostly with storage. Already filled two racks. The mad part says to get a third. But it would probably be wise to wait it out 🙂 Stage one was all about building the cluster out, stage two is left for expansions..
Tearing the network apart, not without tears.. Old Juniper layer is retired and adding a restructure of the network layout for the opportunity. The old network has served well, but as the Juniper firewall layer only manages to push a couple of 100’s of megabits, and the new WAN is a couple gigabits, an upgrade is pretty much required. A redesign of the underlying architecture is generally never a bad thing. Time will tell how many of the components will end up being used.
Also tearing the server-side networking apart, redesigning the layout of the server-core network. All servers are now LACP’ing over the stack, to allow for component failures, as well as increased bandwidth. The underlying goal is to be able to loose half of either side and still remain in some form of operational condition, will be interesting to see how LACP and RSTP handles the job.
Tearing it apart. Not without shedding a tear. It has been a great routing and firewall layer, nevertheless, Juniper gear that handles a couple giabit of throughput is way out of budget 🙁
To be continued.