Hi guys, The LFS server is definitely getting old and slow and nobody would complain if something gets done about it. So, something is about to be done about it.
I am going through some upgrades here at work. Our main router is going to be replaced as part of our network upgrade roll-outs. I decided the main router can be a Linux machine so it can also perform other duties such as being the main mail sever, web server, things like that. This box is already installed in our upstream ISPs colocation facility and connects to the province wide fiber optic network (called the Alberta Supernet for those who are familiar with the area). This new work server is quite powerful compared to the LFS server (stats will follow below) and what I'm going to have it do is not nearly going to max out its performance (both hardware and bandwidth). A quick chat with my boss and he agreed that I can make it the new LFS server too. I just have to pay the bandwidth it generates. The montly bandwidth costs are much lower than what i pay for LFS' colocation right now. I'll be cutting the bill literally in half. A good deal all around. So there will be a bit of a transition of course when I get around making the switch. For comparison, the current server specs: PIII-750 MHz, 512 MB RAM, 2x 9 GB SCSI disks New server specs: P4-2.8 GHz Xeon, 1 GB RAM, 2x 36 GB SCSI in RAID1 config. This province wide fiber network isn't tested by me yet. As soon as the router is online and I've tested the IP block assigned by Arin to be working properly, I'll give the server a week or two as a shake-down. Then of course, corporate stuff first and make all our customers move over to it. If all is stable, start moving LFS stuff to it. The only thing that will be downgraded is the bandwidth. Right now the server is sharing a 155 Mbit pipe but the NIC in Belgarath is only a 10/100 so we're on the 100 pipe. I've never gotten more than 5 Mbit out of it anyways seeing the nature of the Internet itself. But really, the pipe is overkill for LFS anyways seeing it's current usage. For the last months we have been averaging a sustained rate of 22 KB/sec (combined downloads and uploads) with the occasional spike up to maybe 400 or 500 KB/sec. I've not even had enough of a sustained burst to show up on MRTG that exceeded 3 Mbit. So seeing our bandwidth usage is so low with all our good mirrors out there, a downgrade in available link speed isn't going to hurt us. And if it does, well, use more mirrors and keep the main server as a master repository, SVN, etc. That truly wouldn't be a bad thing anyway. Anyways, this router will serve a segment of our customers and will get a 10 Mbit connection. As our network demands on this segment increase we'll get more bandwidth and the LFS server will get more as well than. By the way, as an aside in case you are wondering how a 10 meg link can serve a large group of customers without problems: our wireless hardware works with GPS sharing. Basically every connected module gets a GPS timeslot during which it gets maximum bandwidth (the full 10 Mbit or whatever a customer module can support. Some people can only pull down 3 Mbit max as they are capped to lower bandwidth on their contracts). A few milliseconds later the next module gets the full speed, and so forth. Thus far this customer segment we are upgrading has been served by a simple 4 Mbit DSL link to over a hundred people. I can max out the 4 Mbit by downloading something (like a kernel, they have great bandwidth and makes for a good sustained data rate test) and other people can do the same. The effective throughput goes down a little due to overhead, but I've seen a dozen people download at pretty much 3 Mbit simultaneously over one single DSL modem. Try that over a regular LAN. Anyways, upgrades pending. I just have to put the server through its paces and test the new connections since I don't to move LFS stuff over only to find out that a few days later there's something wrong of course. Moving the services over once is more than I'd like to do in a span of a few days. I'll keep you guys updated. It's one of the reasons I've been quiet on the list, getting this upgrade and pre-testing completed. -- Gerard Beekmans /* If Linux doesn't have the solution, you have the wrong problem */ -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
