[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Reply to: [list | sender only]
Re: Crystallography Journals Online
- To: Multiple recipients of list <epc-l@iucr.org>
- Subject: Re: Crystallography Journals Online
- From: Brian McMahon <bm@iucr.org>
- Date: Thu, 1 Jul 1999 10:10:21 +0100 (BST)
Thanks to everyone for your comments so far. Many identify points on which we also have concerns, and many others open interesting lines of discussion that will be explored in due course. Here I just respond to Lachlan's inquiries for factual information about the connectivity issues. > What is the estimated cost of getting the IUCr a "top class" fast > link? > - Cost of connection > - Cost of hardware to support connection > - Admin/staff/time costs and issues > - Yearly recurrent/maintenance costs. These are costings that will be produced for the Finance Committee. We are looking at two possibilities to enhance direct connectivity: (1) upgrade the IUCr direct JANET connection to 2 Mbps. The line rental cost is high (about 37,000 pounds p.a.). We could currently support this with existing hardware and software, but it might be prudent to upgrade the network software and, in the longer term, provide a dedicated router machine. (2) lease a "facilities managed" server from an ISP. Pipex offer a deal hosting a Sun Ultra 10 on a 10 Mbps backbone for about 23,000 p.a. They provide daily backups and manage hardware downtime. All software would be installed and managed by us. Given that we would need to maintain our existing JANET link, the overall costs of schemes (1) and (2) are similar. The notional higher bandwidth of (2) is shared across far more users, so it's hard to know in advance which would provide the faster service to end-users. In practice a slower direct link to JANET (say 256 kbps) might offer the same throughput. My current thinking is however to favour (2) on the following grounds: (a) Redundancy - there is nothing to stop us running our own server at Chester for (i) our own use, (ii) possible use by Coeditors or other relevant people for whom it might be more convenient, (iii) redirection in the event of prolonged downtime at Pipex (or the demise of that company). Probably redundancy of international connectivity is not greatly helped in the event of a major disaster at the central routing stations: both JANET and Pipex access the international networks through London docklands sites. (b) Security - both of hardware through the facilities management at Pipex, and unauthorised access through separating the server from our own network. Further, replication of access controls at both sites would ease recovery from a compromised system. (c) Extensibility - Pipex is an MCI company, and offers similar deals on other networks world wide. Leasing of identical server hardware under our direct control will simplify mirroring. (3) A third approach to improve connectivity is to host the service through a third party. We are also exploring this avenue. In terms of costs directly attributable to bandwidth, it is a much cheaper option; but there are of course other costs against which that will be offset. It is interesting to note that one of the possible third parties we are considering uses the Pipex facilities-managed service for their own equipment. It would help to assess the likely improvement in speed if you could look at the following sites and let me have your views on the speed of response from each: www.rsc.org 2 Mbps leased connection through Pipex www.ioppublishing.co.uk " www.turpin-distribution.com 10Mbps Pipex backbone facilities managed > Other issue is how easy would it possible be to mirror these remotely > and is this practical if accounts could be obtained on remote servers? > And the machines were configured with security in mind (mainly secure > telnetting using Secure Shell)? I see the mirroring of the journals service proceeding through the acquisition of dedicated facilities-managed servers, rather than as an extension of the existing arrangements. The need for higher capacity (of order gigabytes) and reliable access control would demand more from the mirror site administrators and would likely be much more complex to administer. Comments welcome. Brian
Reply to: [list | sender only]
- Prev by Date: Re: Crystallography Journals Online
- Next by Date: Re: Crystallography Journals Online
- Prev by thread: Re: Crystallography Journals Online
- Next by thread: Re: Crystallography Journals Online
- Index(es):